* [PATCH v4 0/9] snic:initial submission of snic driver for Cisco SCSI HBA
@ 2015-04-09 11:49 Narsimhulu Musini
2015-04-09 11:49 ` [PATCH v4 1/9] snic: snic module infrastructure Narsimhulu Musini
` (8 more replies)
0 siblings, 9 replies; 24+ messages in thread
From: Narsimhulu Musini @ 2015-04-09 11:49 UTC (permalink / raw)
To: JBottomley, linux-scsi, hare, hch; +Cc: Narsimhulu Musini
The following patch series adds snic driver for Cisco SCSI HBA
Cisco has developed a new PCI HBA interface called sNIC, which stands for
SCSI NIC. This is a new storage feature supported on specialized network
adapter. The new PCI function provides a uniform host interface and abstracts
backend storage.
* v4
- Incorporated Christoph Hellwig's comments.
* v3
- Incorporated Dr. Hannes Reinecke's review comments.
* v2
- Incorporated Dr. Hannes Reinecke's review comments.
Thanks
Narsimhulu Musini & Sesidhar Baddela
Narsimhulu Musini (9):
snic: snic module infrastructure
snic:Add interrupt, resource firmware interfaces
snic:Add meta request, handling of meta requests.
snic:Add snic target discovery
snic:add SCSI handling, AEN, and fwreset handling
snic:Add low level queuing interfaces
snic:Add sysfs entries to list stats and trace data
snic:Add event tracing to capture IO events.
snic:Add Makefile, patch Kconfig, MAINTAINERS
MAINTAINERS | 7 +
drivers/scsi/Kconfig | 17 +
drivers/scsi/Makefile | 1 +
drivers/scsi/snic/Makefile | 21 +
drivers/scsi/snic/cq_desc.h | 76 ++
drivers/scsi/snic/cq_enet_desc.h | 38 +
drivers/scsi/snic/snic.h | 415 ++++++
drivers/scsi/snic/snic_attrs.c | 79 ++
drivers/scsi/snic/snic_ctl.c | 276 ++++
drivers/scsi/snic/snic_debugfs.c | 562 ++++++++
drivers/scsi/snic/snic_disc.c | 602 +++++++++
drivers/scsi/snic/snic_disc.h | 124 ++
drivers/scsi/snic/snic_fwint.h | 525 ++++++++
drivers/scsi/snic/snic_io.c | 519 ++++++++
drivers/scsi/snic/snic_io.h | 118 ++
drivers/scsi/snic/snic_isr.c | 209 +++
drivers/scsi/snic/snic_main.c | 1039 +++++++++++++++
drivers/scsi/snic/snic_res.c | 297 +++++
drivers/scsi/snic/snic_res.h | 96 ++
drivers/scsi/snic/snic_scsi.c | 2638 +++++++++++++++++++++++++++++++++++++
drivers/scsi/snic/snic_stats.h | 123 ++
drivers/scsi/snic/snic_trc.c | 182 +++
drivers/scsi/snic/snic_trc.h | 121 ++
drivers/scsi/snic/vnic_cq.c | 86 ++
drivers/scsi/snic/vnic_cq.h | 120 ++
drivers/scsi/snic/vnic_cq_fw.h | 62 +
drivers/scsi/snic/vnic_dev.c | 749 +++++++++++
drivers/scsi/snic/vnic_dev.h | 140 ++
drivers/scsi/snic/vnic_devcmd.h | 270 ++++
drivers/scsi/snic/vnic_intr.c | 59 +
drivers/scsi/snic/vnic_intr.h | 119 ++
drivers/scsi/snic/vnic_resource.h | 68 +
drivers/scsi/snic/vnic_snic.h | 54 +
drivers/scsi/snic/vnic_stats.h | 68 +
drivers/scsi/snic/vnic_wq.c | 236 ++++
drivers/scsi/snic/vnic_wq.h | 187 +++
drivers/scsi/snic/wq_enet_desc.h | 91 ++
37 files changed, 10394 insertions(+)
create mode 100644 drivers/scsi/snic/Makefile
create mode 100644 drivers/scsi/snic/cq_desc.h
create mode 100644 drivers/scsi/snic/cq_enet_desc.h
create mode 100644 drivers/scsi/snic/snic.h
create mode 100644 drivers/scsi/snic/snic_attrs.c
create mode 100644 drivers/scsi/snic/snic_ctl.c
create mode 100644 drivers/scsi/snic/snic_debugfs.c
create mode 100644 drivers/scsi/snic/snic_disc.c
create mode 100644 drivers/scsi/snic/snic_disc.h
create mode 100644 drivers/scsi/snic/snic_fwint.h
create mode 100644 drivers/scsi/snic/snic_io.c
create mode 100644 drivers/scsi/snic/snic_io.h
create mode 100644 drivers/scsi/snic/snic_isr.c
create mode 100644 drivers/scsi/snic/snic_main.c
create mode 100644 drivers/scsi/snic/snic_res.c
create mode 100644 drivers/scsi/snic/snic_res.h
create mode 100644 drivers/scsi/snic/snic_scsi.c
create mode 100644 drivers/scsi/snic/snic_stats.h
create mode 100644 drivers/scsi/snic/snic_trc.c
create mode 100644 drivers/scsi/snic/snic_trc.h
create mode 100644 drivers/scsi/snic/vnic_cq.c
create mode 100644 drivers/scsi/snic/vnic_cq.h
create mode 100644 drivers/scsi/snic/vnic_cq_fw.h
create mode 100644 drivers/scsi/snic/vnic_dev.c
create mode 100644 drivers/scsi/snic/vnic_dev.h
create mode 100644 drivers/scsi/snic/vnic_devcmd.h
create mode 100644 drivers/scsi/snic/vnic_intr.c
create mode 100644 drivers/scsi/snic/vnic_intr.h
create mode 100644 drivers/scsi/snic/vnic_resource.h
create mode 100644 drivers/scsi/snic/vnic_snic.h
create mode 100644 drivers/scsi/snic/vnic_stats.h
create mode 100644 drivers/scsi/snic/vnic_wq.c
create mode 100644 drivers/scsi/snic/vnic_wq.h
create mode 100644 drivers/scsi/snic/wq_enet_desc.h
--
1.8.5.4
^ permalink raw reply [flat|nested] 24+ messages in thread
* [PATCH v4 1/9] snic: snic module infrastructure
2015-04-09 11:49 [PATCH v4 0/9] snic:initial submission of snic driver for Cisco SCSI HBA Narsimhulu Musini
@ 2015-04-09 11:49 ` Narsimhulu Musini
2015-04-09 12:40 ` Hannes Reinecke
2015-04-09 11:49 ` [PATCH v4 2/9] snic:Add interrupt, resource firmware interfaces Narsimhulu Musini
` (7 subsequent siblings)
8 siblings, 1 reply; 24+ messages in thread
From: Narsimhulu Musini @ 2015-04-09 11:49 UTC (permalink / raw)
To: JBottomley, linux-scsi, hare, hch; +Cc: Narsimhulu Musini, Sesidhar Baddela
snic_main.c contains module load and unload, global driver context,
PCI Registration, PCI probe and remove, SCSI ML registration functionality.
snic.h contains snic structure definition, snic global context, and
prototypes.
snic_attrs.c contains device attributes to list snic state, link state,
and driver version under /sys/class/scsi_host/host<id>/
Signed-off-by: Narsimhulu Musini <nmusini@cisco.com>
Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
---
* v4
- Replaced scratch pad usage with private data structure for internal use.
- Changed cmd_per_lun to default queue depth.
- Removed snic_os.h file and replaced the wrappers with direct linux API.
* v3
- Cleaned up unused functions, structure members, and redundant comments.
- Fixed a typo in time computation.
* v2
- Added Compile time macro for debugfs dependent functionality.
drivers/scsi/snic/snic.h | 415 ++++++++++++++++
drivers/scsi/snic/snic_attrs.c | 79 +++
drivers/scsi/snic/snic_main.c | 1039 ++++++++++++++++++++++++++++++++++++++++
3 files changed, 1533 insertions(+)
create mode 100644 drivers/scsi/snic/snic.h
create mode 100644 drivers/scsi/snic/snic_attrs.c
create mode 100644 drivers/scsi/snic/snic_main.c
diff --git a/drivers/scsi/snic/snic.h b/drivers/scsi/snic/snic.h
new file mode 100644
index 0000000..533abf0
--- /dev/null
+++ b/drivers/scsi/snic/snic.h
@@ -0,0 +1,415 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef _SNIC_H_
+#define _SNIC_H_
+
+#include <linux/module.h>
+#include <linux/netdevice.h>
+#include <linux/workqueue.h>
+#include <linux/bitops.h>
+#include <linux/mempool.h>
+#include <scsi/scsi_cmnd.h>
+#include <scsi/scsi.h>
+#include <scsi/scsi_host.h>
+
+#include "snic_disc.h"
+#include "snic_io.h"
+#include "snic_res.h"
+#include "snic_trc.h"
+#include "snic_stats.h"
+#include "vnic_dev.h"
+#include "vnic_wq.h"
+#include "vnic_cq.h"
+#include "vnic_intr.h"
+#include "vnic_stats.h"
+#include "vnic_snic.h"
+
+#define SNIC_DRV_NAME "snic"
+#define SNIC_DRV_DESCRIPTION "Cisco SCSI NIC Driver"
+#define SNIC_DRV_VERSION "0.0.1.18"
+#define PFX SNIC_DRV_NAME ":"
+#define DFX SNIC_DRV_NAME "%d: "
+
+#define DESC_CLEAN_LOW_WATERMARK 8
+#define SNIC_UCSM_DFLT_THROTTLE_CNT_BLD 16 /* UCSM default throttle count */
+#define SNIC_MAX_IO_REQ 50 /* scsi_cmnd tag map entries */
+#define SNIC_MIN_IO_REQ 8 /* Min IO throttle count */
+#define SNIC_IO_LOCKS 64 /* IO locks: power of 2 */
+#define SNIC_DFLT_QUEUE_DEPTH 32 /* Default Queue Depth */
+#define SNIC_MAX_QUEUE_DEPTH 64 /* Max Queue Depth */
+#define SNIC_DFLT_CMD_TIMEOUT 90 /* Extended tmo for FW */
+
+/*
+ * Tag bits used for special requests.
+ */
+#define SNIC_TAG_ABORT BIT(30) /* Tag indicating abort */
+#define SNIC_TAG_DEV_RST BIT(29) /* Tag for device reset */
+#define SNIC_TAG_IOCTL_DEV_RST BIT(28) /* Tag for User Device Reset */
+#define SNIC_TAG_MASK (BIT(24) - 1) /* Mask for lookup */
+#define SNIC_NO_TAG -1
+
+/*
+ * Command flags to identify the type of command and for other future use
+ */
+#define SNIC_NO_FLAGS 0
+#define SNIC_IO_INITIALIZED BIT(0)
+#define SNIC_IO_ISSUED BIT(1)
+#define SNIC_IO_DONE BIT(2)
+#define SNIC_IO_REQ_NULL BIT(3)
+#define SNIC_IO_ABTS_PENDING BIT(4)
+#define SNIC_IO_ABORTED BIT(5)
+#define SNIC_IO_ABTS_ISSUED BIT(6)
+#define SNIC_IO_TERM_ISSUED BIT(7)
+#define SNIC_IO_ABTS_TIMEDOUT BIT(8)
+#define SNIC_IO_ABTS_TERM_DONE BIT(9)
+#define SNIC_IO_ABTS_TERM_REQ_NULL BIT(10)
+#define SNIC_IO_ABTS_TERM_TIMEDOUT BIT(11)
+#define SNIC_IO_INTERNAL_TERM_PENDING BIT(12)
+#define SNIC_IO_INTERNAL_TERM_ISSUED BIT(13)
+#define SNIC_DEVICE_RESET BIT(14)
+#define SNIC_DEV_RST_ISSUED BIT(15)
+#define SNIC_DEV_RST_TIMEDOUT BIT(16)
+#define SNIC_DEV_RST_ABTS_ISSUED BIT(17)
+#define SNIC_DEV_RST_TERM_ISSUED BIT(18)
+#define SNIC_DEV_RST_DONE BIT(19)
+#define SNIC_DEV_RST_REQ_NULL BIT(20)
+#define SNIC_DEV_RST_ABTS_DONE BIT(21)
+#define SNIC_DEV_RST_TERM_DONE BIT(22)
+#define SNIC_DEV_RST_ABTS_PENDING BIT(23)
+#define SNIC_DEV_RST_PENDING BIT(24)
+#define SNIC_DEV_RST_NOTSUP BIT(25)
+#define SNIC_SCSI_CLEANUP BIT(26)
+#define SNIC_HOST_RESET_ISSUED BIT(27)
+
+#define SNIC_ABTS_TIMEOUT 30000 /* msec */
+#define SNIC_LUN_RESET_TIMEOUT 30000 /* msec */
+#define SNIC_HOST_RESET_TIMEOUT 30000 /* msec */
+
+
+/*
+ * These are protected by the hashed req_lock.
+ */
+#define CMD_SP(Cmnd) \
+ (((struct snic_internal_io_state *)scsi_cmd_priv(Cmnd))->rqi)
+#define CMD_STATE(Cmnd) \
+ (((struct snic_internal_io_state *)scsi_cmd_priv(Cmnd))->state)
+#define CMD_ABTS_STATUS(Cmnd) \
+ (((struct snic_internal_io_state *)scsi_cmd_priv(Cmnd))->abts_status)
+#define CMD_LR_STATUS(Cmnd) \
+ (((struct snic_internal_io_state *)scsi_cmd_priv(Cmnd))->lr_status)
+#define CMD_FLAGS(Cmnd) \
+ (((struct snic_internal_io_state *)scsi_cmd_priv(Cmnd))->flags)
+
+#define SNIC_INVALID_CODE 0x100 /* Hdr Status val unused by firmware */
+
+#define SNIC_MAX_TARGET 256
+#define SNIC_FLAGS_NONE (0)
+
+/* snic module params */
+extern unsigned int snic_max_qdepth;
+
+/* snic debugging */
+extern unsigned int snic_log_level;
+
+#define SNIC_MAIN_LOGGING 0x1
+#define SNIC_SCSI_LOGGING 0x2
+#define SNIC_ISR_LOGGING 0x8
+#define SNIC_DESC_LOGGING 0x10
+
+#define SNIC_CHECK_LOGGING(LEVEL, CMD) \
+do { \
+ if (unlikely(snic_log_level & LEVEL)) \
+ do { \
+ CMD; \
+ } while (0); \
+} while (0)
+
+#define SNIC_MAIN_DBG(host, fmt, args...) \
+ SNIC_CHECK_LOGGING(SNIC_MAIN_LOGGING, \
+ shost_printk(KERN_INFO, host, fmt, ## args);)
+
+#define SNIC_SCSI_DBG(host, fmt, args...) \
+ SNIC_CHECK_LOGGING(SNIC_SCSI_LOGGING, \
+ shost_printk(KERN_INFO, host, fmt, ##args);)
+
+#define SNIC_DISC_DBG(host, fmt, args...) \
+ SNIC_CHECK_LOGGING(SNIC_SCSI_LOGGING, \
+ shost_printk(KERN_INFO, host, fmt, ##args);)
+
+#define SNIC_ISR_DBG(host, fmt, args...) \
+ SNIC_CHECK_LOGGING(SNIC_ISR_LOGGING, \
+ shost_printk(KERN_INFO, host, fmt, ##args);)
+
+#define SNIC_HOST_ERR(host, fmt, args...) \
+ shost_printk(KERN_ERR, host, fmt, ##args)
+
+#define SNIC_HOST_INFO(host, fmt, args...) \
+ shost_printk(KERN_INFO, host, fmt, ##args)
+
+#define SNIC_INFO(fmt, args...) \
+ pr_info(PFX fmt, ## args)
+
+#define SNIC_DBG(fmt, args...) \
+ pr_info(PFX fmt, ## args)
+
+#define SNIC_ERR(fmt, args...) \
+ pr_err(PFX fmt, ## args)
+
+#ifdef DEBUG
+#define SNIC_BUG_ON(EXPR) \
+ ({ \
+ if (EXPR) { \
+ SNIC_ERR("SNIC BUG(%s)\n", #EXPR); \
+ BUG_ON(EXPR); \
+ } \
+ })
+#else
+#define SNIC_BUG_ON(EXPR) \
+ ({ \
+ if (EXPR) { \
+ SNIC_ERR("SNIC BUG(%s) at %s : %d\n", \
+ #EXPR, __func__, __LINE__); \
+ WARN_ON_ONCE(EXPR); \
+ } \
+ })
+#endif
+
+/* Soft assert */
+#define SNIC_ASSERT_NOT_IMPL(EXPR) \
+ ({ \
+ if (EXPR) {\
+ SNIC_INFO("Functionality not impl'ed at %s:%d\n", \
+ __func__, __LINE__); \
+ WARN_ON_ONCE(EXPR); \
+ } \
+ })
+
+
+extern const char *snic_state_str[];
+
+enum snic_intx_intr_index {
+ SNIC_INTX_WQ_RQ_COPYWQ,
+ SNIC_INTX_ERR,
+ SNIC_INTX_NOTIFY,
+ SNIC_INTX_INTR_MAX,
+};
+
+enum snic_msix_intr_index {
+ SNIC_MSIX_WQ,
+ SNIC_MSIX_IO_CMPL,
+ SNIC_MSIX_ERR_NOTIFY,
+ SNIC_MSIX_INTR_MAX,
+};
+
+struct snic_msix_entry {
+ int requested;
+ char devname[IFNAMSIZ];
+ irqreturn_t (*isr)(int, void *);
+ void *devid;
+};
+
+enum snic_state {
+ SNIC_INIT = 0,
+ SNIC_ERROR,
+ SNIC_ONLINE,
+ SNIC_OFFLINE,
+ SNIC_FWRESET,
+};
+
+#define SNIC_WQ_MAX 1
+#define SNIC_CQ_IO_CMPL_MAX 1
+#define SNIC_CQ_MAX (SNIC_WQ_MAX + SNIC_CQ_IO_CMPL_MAX)
+
+/* firmware version information */
+struct snic_fw_info {
+ u32 fw_ver;
+ u32 hid; /* u16 hid | u16 vnic id */
+ u32 max_concur_ios; /* max concurrent ios */
+ u32 max_sgs_per_cmd; /* max sgls per IO */
+ u32 max_io_sz; /* max io size supported */
+ u32 hba_cap; /* hba capabilities */
+ u32 max_tgts; /* max tgts supported */
+ u16 io_tmo; /* FW Extended timeout */
+ struct completion *wait; /* protected by snic lock*/
+};
+
+/*
+ * snic_work item : defined to process asynchronous events
+ */
+struct snic_work {
+ struct work_struct work;
+ u16 ev_id;
+ u64 *ev_data;
+};
+
+/*
+ * snic structure to represent SCSI vNIC
+ */
+struct snic {
+ /* snic specific members */
+ struct list_head list;
+ char name[IFNAMSIZ];
+ atomic_t state;
+ spinlock_t snic_lock;
+ struct completion *remove_wait;
+ bool in_remove;
+ bool stop_link_events; /* stop processing link events */
+
+ /* discovery related */
+ struct snic_disc disc;
+
+ /* Scsi Host info */
+ struct Scsi_Host *shost;
+
+ /* vnic related structures */
+ struct vnic_dev_bar bar0;
+
+ struct vnic_stats *stats;
+ unsigned long stats_time;
+ unsigned long stats_reset_time;
+
+ struct vnic_dev *vdev;
+
+ /* hw resource info */
+ unsigned int wq_count;
+ unsigned int cq_count;
+ unsigned int intr_count;
+ unsigned int err_intr_offset;
+
+ int link_status; /* retrieved from vnic_dev_link_status() */
+ u32 link_down_cnt;
+
+ /* pci related */
+ struct pci_dev *pdev;
+ struct msix_entry msix_entry[SNIC_MSIX_INTR_MAX];
+ struct snic_msix_entry msix[SNIC_MSIX_INTR_MAX];
+
+ /* io related info */
+ mempool_t *req_pool[SNIC_REQ_MAX_CACHES]; /* (??) */
+ ____cacheline_aligned spinlock_t io_req_lock[SNIC_IO_LOCKS];
+
+ /* Maintain snic specific commands, cmds with no tag in spl_cmd_list */
+ ____cacheline_aligned spinlock_t spl_cmd_lock;
+ struct list_head spl_cmd_list;
+
+ unsigned int max_tag_id;
+ atomic_t ios_inflight; /* io in flight counter */
+
+ struct vnic_snic_config config;
+
+ struct work_struct link_work;
+
+ /* firmware information */
+ struct snic_fw_info fwinfo;
+
+ /* Work for processing Target related work */
+ struct work_struct tgt_work;
+
+ /* Work for processing Discovery */
+ struct work_struct disc_work;
+
+ /* stats related */
+ unsigned int reset_stats;
+ atomic64_t io_cmpl_skip;
+ struct snic_stats s_stats; /* Per SNIC driver stats */
+
+ /* platform specific */
+#ifdef SNIC_DEBUG_FS
+ struct dentry *stats_host; /* Per snic debugfs root */
+ struct dentry *stats_file; /* Per snic debugfs file */
+ struct dentry *reset_stats_file;/* Per snic reset stats file */
+#endif
+
+ /* completion queue cache line section */
+ ____cacheline_aligned struct vnic_cq cq[SNIC_CQ_MAX];
+
+ /* work queue cache line section */
+ ____cacheline_aligned struct vnic_wq wq[SNIC_WQ_MAX];
+ spinlock_t wq_lock[SNIC_WQ_MAX];
+
+ /* interrupt resource cache line section */
+ ____cacheline_aligned struct vnic_intr intr[SNIC_MSIX_INTR_MAX];
+}; /* end of snic structure */
+
+/*
+ * SNIC Driver's Global Data
+ */
+struct snic_global {
+ struct list_head snic_list;
+ spinlock_t snic_list_lock;
+
+ struct kmem_cache *req_cache[SNIC_REQ_MAX_CACHES];
+
+ struct workqueue_struct *event_q;
+
+#ifdef SNIC_DEBUG_FS
+ /* debugfs related global data */
+ struct dentry *trc_root;
+ struct dentry *stats_root;
+
+ struct snic_trc trc ____cacheline_aligned;
+
+#endif
+};
+
+extern struct snic_global *snic_glob;
+
+int snic_glob_init(void);
+void snic_glob_cleanup(void);
+
+extern struct workqueue_struct *snic_event_queue;
+extern struct device_attribute *snic_attrs[];
+
+int snic_queuecommand(struct Scsi_Host *, struct scsi_cmnd *);
+int snic_abort_cmd(struct scsi_cmnd *);
+int snic_device_reset(struct scsi_cmnd *);
+int snic_host_reset(struct scsi_cmnd *);
+int snic_reset(struct Scsi_Host *, struct scsi_cmnd *);
+void snic_shutdown_scsi_cleanup(struct snic *);
+
+
+int snic_request_intr(struct snic *);
+void snic_free_intr(struct snic *);
+int snic_set_intr_mode(struct snic *);
+void snic_clear_intr_mode(struct snic *);
+
+int snic_fwcq_cmpl_handler(struct snic *, int);
+int snic_wq_cmpl_handler(struct snic *, int);
+void snic_free_wq_buf(struct vnic_wq *, struct vnic_wq_buf *);
+
+
+void snic_log_q_error(struct snic *);
+void snic_handle_link_event(struct snic *);
+void snic_handle_link(struct work_struct *);
+
+int snic_queue_exch_ver_req(struct snic *);
+int snic_io_exch_ver_cmpl_handler(struct snic *, struct snic_fw_req *);
+
+int snic_queue_wq_desc(struct snic *, void *os_buf, u16 len);
+
+void snic_handle_untagged_req(struct snic *, struct snic_req_info *);
+void snic_release_untagged_req(struct snic *, struct snic_req_info *);
+void snic_free_all_untagged_reqs(struct snic *);
+int snic_get_conf(struct snic *);
+void snic_set_state(struct snic *, enum snic_state);
+int snic_get_state(struct snic *);
+const char *snic_state_to_str(unsigned int);
+void snic_hex_dump(char *, char *, int);
+void snic_print_desc(const char *fn, char *os_buf, int len);
+const char *show_opcode_name(int val);
+#endif /* _SNIC_H */
diff --git a/drivers/scsi/snic/snic_attrs.c b/drivers/scsi/snic/snic_attrs.c
new file mode 100644
index 0000000..88822ad
--- /dev/null
+++ b/drivers/scsi/snic/snic_attrs.c
@@ -0,0 +1,79 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ * [Insert appropriate license here when releasing outside of Cisco]
+ *
+ */
+
+#include <linux/string.h>
+#include <linux/device.h>
+
+#include "snic.h"
+
+static ssize_t
+snic_show_sym_name(struct device *dev,
+ struct device_attribute *attr,
+ char *buf)
+{
+ struct snic *snic = shost_priv(class_to_shost(dev));
+
+ return snprintf(buf, PAGE_SIZE, "%s\n", snic->name);
+}
+
+static ssize_t
+snic_show_state(struct device *dev,
+ struct device_attribute *attr,
+ char *buf)
+{
+ struct snic *snic = shost_priv(class_to_shost(dev));
+
+ return snprintf(buf, PAGE_SIZE, "%s\n",
+ snic_state_str[snic_get_state(snic)]);
+}
+
+static ssize_t
+snic_show_drv_version(struct device *dev,
+ struct device_attribute *attr,
+ char *buf)
+{
+ return snprintf(buf, PAGE_SIZE, "%s\n", SNIC_DRV_VERSION);
+}
+
+static ssize_t
+snic_show_link_state(struct device *dev,
+ struct device_attribute *attr,
+ char *buf)
+{
+ struct snic *snic = shost_priv(class_to_shost(dev));
+
+ if (snic->config.xpt_type == SNIC_DAS)
+ snic->link_status = vnic_dev_link_status(snic->vdev);
+
+ return snprintf(buf, PAGE_SIZE, "%s\n",
+ (snic->link_status) ? "Link Up" : "Link Down");
+}
+
+static DEVICE_ATTR(snic_sym_name, S_IRUGO, snic_show_sym_name, NULL);
+static DEVICE_ATTR(snic_state, S_IRUGO, snic_show_state, NULL);
+static DEVICE_ATTR(drv_version, S_IRUGO, snic_show_drv_version, NULL);
+static DEVICE_ATTR(link_state, S_IRUGO, snic_show_link_state, NULL);
+
+struct device_attribute *snic_attrs[] = {
+ &dev_attr_snic_sym_name,
+ &dev_attr_snic_state,
+ &dev_attr_drv_version,
+ &dev_attr_link_state,
+ NULL,
+};
diff --git a/drivers/scsi/snic/snic_main.c b/drivers/scsi/snic/snic_main.c
new file mode 100644
index 0000000..a815377
--- /dev/null
+++ b/drivers/scsi/snic/snic_main.c
@@ -0,0 +1,1039 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/module.h>
+#include <linux/mempool.h>
+#include <linux/string.h>
+#include <linux/slab.h>
+#include <linux/errno.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/skbuff.h>
+#include <linux/interrupt.h>
+#include <linux/spinlock.h>
+#include <linux/workqueue.h>
+#include <scsi/scsi_host.h>
+#include <scsi/scsi_tcq.h>
+
+#include "snic.h"
+#include "snic_fwint.h"
+
+#define PCI_DEVICE_ID_CISCO_SNIC 0x0046
+
+/* Supported devices by snic module */
+static struct pci_device_id snic_id_table[] = {
+ {PCI_DEVICE(0x1137, PCI_DEVICE_ID_CISCO_SNIC) },
+ { 0, } /* end of table */
+};
+
+unsigned int snic_log_level = 0x0;
+module_param(snic_log_level, int, S_IRUGO|S_IWUSR);
+MODULE_PARM_DESC(snic_log_level, "bitmask for snic logging levels");
+
+#ifdef SNIC_DEBUG_FS
+unsigned int snic_trace_max_pages = 16;
+module_param(snic_trace_max_pages, uint, S_IRUGO|S_IWUSR);
+MODULE_PARM_DESC(snic_trace_max_pages,
+ "Total allocated memory pages for snic trace buffer");
+
+#endif
+unsigned int snic_max_qdepth = SNIC_DFLT_QUEUE_DEPTH;
+module_param(snic_max_qdepth, uint, S_IRUGO | S_IWUSR);
+MODULE_PARM_DESC(snic_max_qdepth, "Queue depth to report for each LUN");
+
+/*
+ * snic_slave_alloc : callback function to SCSI Mid Layer, called on
+ * scsi device initialization.
+ */
+static int
+snic_slave_alloc(struct scsi_device *sdev)
+{
+ struct snic_tgt *tgt = starget_to_tgt(scsi_target(sdev));
+
+ if (!tgt || snic_tgt_chkready(tgt))
+ return -ENXIO;
+
+ return 0;
+}
+
+/*
+ * snic_slave_configure : callback function to SCSI Mid Layer, called on
+ * scsi device initialization.
+ */
+static int
+snic_slave_configure(struct scsi_device *sdev)
+{
+ struct snic *snic = shost_priv(sdev->host);
+ u32 qdepth = 0, max_ios = 0;
+ int tmo = SNIC_DFLT_CMD_TIMEOUT * HZ;
+
+ /* Set Queue Depth */
+ max_ios = snic_max_qdepth;
+ qdepth = min_t(u32, max_ios, SNIC_MAX_QUEUE_DEPTH);
+ scsi_change_queue_depth(sdev, qdepth);
+
+ if (snic->fwinfo.io_tmo > 1)
+ tmo = snic->fwinfo.io_tmo * HZ;
+
+ /* FW requires extended timeouts */
+ blk_queue_rq_timeout(sdev->request_queue, tmo);
+
+ return 0;
+}
+
+static int
+snic_change_queue_depth(struct scsi_device *sdev, int qdepth)
+{
+ int qsz = 0;
+
+ qsz = min_t(u32, qdepth, SNIC_MAX_QUEUE_DEPTH);
+ scsi_change_queue_depth(sdev, qsz);
+ SNIC_INFO("QDepth Changed to %d\n", sdev->queue_depth);
+
+ return sdev->queue_depth;
+}
+
+static struct scsi_host_template snic_host_template = {
+ .module = THIS_MODULE,
+ .name = SNIC_DRV_NAME,
+ .queuecommand = snic_queuecommand,
+ .eh_abort_handler = snic_abort_cmd,
+ .eh_device_reset_handler = snic_device_reset,
+ .eh_host_reset_handler = snic_host_reset,
+ .slave_alloc = snic_slave_alloc,
+ .slave_configure = snic_slave_configure,
+ .change_queue_depth = snic_change_queue_depth,
+ .this_id = -1,
+ .cmd_per_lun = SNIC_DFLT_QUEUE_DEPTH,
+ .can_queue = SNIC_MAX_IO_REQ,
+ .use_clustering = ENABLE_CLUSTERING,
+ .sg_tablesize = SNIC_MAX_SG_DESC_CNT,
+ .max_sectors = 0x800,
+ .shost_attrs = snic_attrs,
+ .use_blk_tags = 1,
+ .track_queue_depth = 1,
+ .cmd_size = sizeof(struct snic_internal_io_state),
+ .proc_name = "snic_scsi",
+};
+
+/*
+ * snic_handle_link_event : Handles link events such as link up/down/error
+ */
+void
+snic_handle_link_event(struct snic *snic)
+{
+ unsigned long flags;
+
+ spin_lock_irqsave(&snic->snic_lock, flags);
+ if (snic->stop_link_events) {
+ spin_unlock_irqrestore(&snic->snic_lock, flags);
+
+ return;
+ }
+ spin_unlock_irqrestore(&snic->snic_lock, flags);
+
+ queue_work(snic_glob->event_q, &snic->link_work);
+} /* end of snic_handle_link_event */
+
+/*
+ * snic_notify_set : sets notification area
+ * This notification area is to receive events from fw
+ * Note: snic supports only MSIX interrupts, in which we can just call
+ * vnic_dev_notify_set directly
+ */
+static int
+snic_notify_set(struct snic *snic)
+{
+ int ret = 0;
+ enum vnic_dev_intr_mode intr_mode;
+
+ intr_mode = vnic_dev_get_intr_mode(snic->vdev);
+
+ if (intr_mode == VNIC_DEV_INTR_MODE_MSIX) {
+ ret = vnic_dev_notify_set(snic->vdev, SNIC_MSIX_ERR_NOTIFY);
+ } else {
+ SNIC_HOST_ERR(snic->shost,
+ "Interrupt mode should be setup before devcmd notify set %d\n",
+ intr_mode);
+ ret = -1;
+ }
+
+ return ret;
+} /* end of snic_notify_set */
+
+/*
+ * snic_dev_wait : polls vnic open status.
+ */
+static int
+snic_dev_wait(struct vnic_dev *vdev,
+ int (*start)(struct vnic_dev *, int),
+ int (*finished)(struct vnic_dev *, int *),
+ int arg)
+{
+ unsigned long time;
+ int ret, done;
+ int retry_cnt = 0;
+
+ ret = start(vdev, arg);
+ if (ret)
+ return ret;
+
+ /*
+ * Wait for func to complete...2 seconds max.
+ *
+ * Sometimes schedule_timeout_uninterruptible take long time
+ * to wakeup, which results skipping retry. The retry counter
+ * ensures to retry at least two times.
+ */
+ time = jiffies + (HZ * 2);
+ do {
+ ret = finished(vdev, &done);
+ if (ret)
+ return ret;
+
+ if (done)
+ return 0;
+ schedule_timeout_uninterruptible(HZ/10);
+ ++retry_cnt;
+ } while (time_after(time, jiffies) || (retry_cnt < 3));
+
+ return -ETIMEDOUT;
+} /* end of snic_dev_wait */
+
+/*
+ * snic_cleanup: called by snic_remove
+ * Stops the snic device, masks all interrupts, Completed CQ entries are
+ * drained. Posted WQ/RQ/Copy-WQ entries are cleanup
+ */
+static int
+snic_cleanup(struct snic *snic)
+{
+ unsigned int i;
+ int ret;
+
+ vnic_dev_disable(snic->vdev);
+ for (i = 0; i < snic->intr_count; i++)
+ vnic_intr_mask(&snic->intr[i]);
+
+ for (i = 0; i < snic->wq_count; i++) {
+ ret = vnic_wq_disable(&snic->wq[i]);
+ if (ret)
+ return ret;
+ }
+
+ /* Clean up completed IOs */
+ snic_fwcq_cmpl_handler(snic, -1);
+
+ snic_wq_cmpl_handler(snic, -1);
+
+ /* Clean up the IOs that have not completed */
+ for (i = 0; i < snic->wq_count; i++)
+ vnic_wq_clean(&snic->wq[i], snic_free_wq_buf);
+
+ for (i = 0; i < snic->cq_count; i++)
+ vnic_cq_clean(&snic->cq[i]);
+
+ for (i = 0; i < snic->intr_count; i++)
+ vnic_intr_clean(&snic->intr[i]);
+
+ /* Cleanup snic specific requests */
+ snic_free_all_untagged_reqs(snic);
+
+ /* Cleanup Pending SCSI commands */
+ snic_shutdown_scsi_cleanup(snic);
+
+ for (i = 0; i < SNIC_REQ_MAX_CACHES; i++)
+ mempool_destroy(snic->req_pool[i]);
+
+ return 0;
+} /* end of snic_cleanup */
+
+
+static void
+snic_iounmap(struct snic *snic)
+{
+ if (snic->bar0.vaddr)
+ iounmap(snic->bar0.vaddr);
+}
+
+/*
+ * snic_vdev_open_done : polls for vnic_dev_open cmd completion.
+ */
+static int
+snic_vdev_open_done(struct vnic_dev *vdev, int *done)
+{
+ struct snic *snic = vnic_dev_priv(vdev);
+ int ret;
+ int nretries = 5;
+
+ do {
+ ret = vnic_dev_open_done(vdev, done);
+ if (ret == 0)
+ break;
+
+ SNIC_HOST_INFO(snic->shost, "VNIC_DEV_OPEN Timedout.\n");
+ } while (nretries--);
+
+ return ret;
+} /* end of snic_vdev_open_done */
+
+/*
+ * snic_add_host : registers scsi host with ML
+ */
+static int
+snic_add_host(struct Scsi_Host *shost, struct pci_dev *pdev)
+{
+ int ret = 0;
+
+ ret = scsi_add_host(shost, &pdev->dev);
+ if (ret) {
+ SNIC_HOST_ERR(shost,
+ "snic: scsi_add_host failed. %d\n",
+ ret);
+
+ return ret;
+ }
+
+ SNIC_BUG_ON(shost->work_q != NULL);
+ snprintf(shost->work_q_name, sizeof(shost->work_q_name), "scsi_wq_%d",
+ shost->host_no);
+ shost->work_q = create_singlethread_workqueue(shost->work_q_name);
+ if (!shost->work_q) {
+ SNIC_HOST_ERR(shost, "Failed to Create ScsiHost wq.\n");
+
+ ret = -ENOMEM;
+ }
+
+ return ret;
+} /* end of snic_add_host */
+
+static void
+snic_del_host(struct Scsi_Host *shost)
+{
+ if (!shost->work_q)
+ return;
+
+ destroy_workqueue(shost->work_q);
+ shost->work_q = NULL;
+ scsi_remove_host(shost);
+}
+
+int
+snic_get_state(struct snic *snic)
+{
+ return atomic_read(&snic->state);
+}
+
+void
+snic_set_state(struct snic *snic, enum snic_state state)
+{
+ SNIC_HOST_INFO(snic->shost, "snic state change from %s to %s\n",
+ snic_state_to_str(snic_get_state(snic)),
+ snic_state_to_str(state));
+
+ atomic_set(&snic->state, state);
+}
+
+/*
+ * snic_probe : Initialize the snic interface.
+ */
+static int
+snic_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
+{
+ struct Scsi_Host *shost;
+ struct snic *snic;
+ mempool_t *pool;
+ unsigned long flags;
+ u32 max_ios = 0;
+ int ret, i;
+
+ /* Device Information */
+ SNIC_INFO("snic device %4x:%4x:%4x:%4x: ",
+ pdev->vendor, pdev->device, pdev->subsystem_vendor,
+ pdev->subsystem_device);
+
+ SNIC_INFO("snic device bus %x: slot %x: fn %x\n",
+ pdev->bus->number, PCI_SLOT(pdev->devfn),
+ PCI_FUNC(pdev->devfn));
+
+ /*
+ * Allocate SCSI Host and setup association between host, and snic
+ */
+ shost = scsi_host_alloc(&snic_host_template, sizeof(struct snic));
+ if (!shost) {
+ SNIC_ERR("Unable to alloc scsi_host\n");
+ ret = -ENOMEM;
+
+ goto prob_end;
+ }
+ snic = shost_priv(shost);
+ snic->shost = shost;
+
+ snprintf(snic->name, sizeof(snic->name) - 1, "%s%d", SNIC_DRV_NAME,
+ shost->host_no);
+
+ SNIC_HOST_INFO(shost,
+ "snic%d = %p shost = %p device bus %x: slot %x: fn %x\n",
+ shost->host_no, snic, shost, pdev->bus->number,
+ PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
+#ifdef SNIC_DEBUG_FS
+ /* Per snic debugfs init */
+ ret = snic_stats_debugfs_init(snic);
+ if (ret) {
+ SNIC_HOST_ERR(snic->shost,
+ "Failed to initialize debugfs stats\n");
+ snic_stats_debugfs_remove(snic);
+ }
+#endif
+
+ /* Setup PCI Resources */
+ pci_set_drvdata(pdev, snic);
+ snic->pdev = pdev;
+
+ ret = pci_enable_device(pdev);
+ if (ret) {
+ SNIC_HOST_ERR(shost,
+ "Cannot enable PCI Resources, aborting : %d\n",
+ ret);
+
+ goto err_free_snic;
+ }
+
+ ret = pci_request_regions(pdev, SNIC_DRV_NAME);
+ if (ret) {
+ SNIC_HOST_ERR(shost,
+ "Cannot obtain PCI Resources, aborting : %d\n",
+ ret);
+
+ goto err_pci_disable;
+ }
+
+ pci_set_master(pdev);
+
+ /*
+ * Query PCI Controller on system for DMA addressing
+ * limitation for the device. Try 43-bit first, and
+ * fail to 32-bit.
+ */
+ ret = pci_set_dma_mask(pdev, DMA_BIT_MASK(43));
+ if (ret) {
+ ret = pci_set_dma_mask(pdev, DMA_BIT_MASK(32));
+ if (ret) {
+ SNIC_HOST_ERR(shost,
+ "No Usable DMA Configuration, aborting %d\n",
+ ret);
+
+ goto err_rel_regions;
+ }
+
+ ret = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(32));
+ if (ret) {
+ SNIC_HOST_ERR(shost,
+ "Unable to obtain 32-bit DMA for consistent allocations, aborting: %d\n",
+ ret);
+
+ goto err_rel_regions;
+ }
+ } else {
+ ret = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(43));
+ if (ret) {
+ SNIC_HOST_ERR(shost,
+ "Unable to obtain 43-bit DMA for consistent allocations. aborting: %d\n",
+ ret);
+
+ goto err_rel_regions;
+ }
+ }
+
+
+ /* Map vNIC resources from BAR0 */
+ if (!(pci_resource_flags(pdev, 0) & IORESOURCE_MEM)) {
+ SNIC_HOST_ERR(shost, "BAR0 not memory mappable aborting.\n");
+
+ ret = -ENODEV;
+ goto err_rel_regions;
+ }
+
+ snic->bar0.vaddr = pci_iomap(pdev, 0, 0);
+ if (!snic->bar0.vaddr) {
+ SNIC_HOST_ERR(shost,
+ "Cannot memory map BAR0 res hdr aborting.\n");
+
+ ret = -ENODEV;
+ goto err_rel_regions;
+ }
+
+ snic->bar0.bus_addr = pci_resource_start(pdev, 0);
+ snic->bar0.len = pci_resource_len(pdev, 0);
+ SNIC_BUG_ON(snic->bar0.bus_addr == 0);
+
+ /* Devcmd2 Resource Allocation and Initialization */
+ snic->vdev = vnic_dev_alloc_discover(NULL, snic, pdev, &snic->bar0, 1);
+ if (!snic->vdev) {
+ SNIC_HOST_ERR(shost, "vNIC Resource Discovery Failed.\n");
+
+ ret = -ENODEV;
+ goto err_iounmap;
+ }
+
+ ret = vnic_dev_cmd_init(snic->vdev, 0);
+ if (ret) {
+ SNIC_HOST_INFO(shost, "Devcmd2 Init Failed. err = %d\n", ret);
+
+ goto err_vnic_unreg;
+ }
+
+ ret = snic_dev_wait(snic->vdev, vnic_dev_open, snic_vdev_open_done, 0);
+ if (ret) {
+ SNIC_HOST_ERR(shost,
+ "vNIC dev open failed, aborting. %d\n",
+ ret);
+
+ goto err_vnic_unreg;
+ }
+
+ ret = vnic_dev_init(snic->vdev, 0);
+ if (ret) {
+ SNIC_HOST_ERR(shost,
+ "vNIC dev init failed. aborting. %d\n",
+ ret);
+
+ goto err_dev_close;
+ }
+
+ /* Get vNIC information */
+ ret = snic_get_vnic_config(snic);
+ if (ret) {
+ SNIC_HOST_ERR(shost,
+ "Get vNIC configuration failed, aborting. %d\n",
+ ret);
+
+ goto err_dev_close;
+ }
+
+ /* Configure Maximum Outstanding IO reqs */
+ max_ios = snic->config.io_throttle_count;
+ if (max_ios != SNIC_UCSM_DFLT_THROTTLE_CNT_BLD)
+ shost->can_queue = min_t(u32, SNIC_MAX_IO_REQ,
+ max_t(u32, SNIC_MIN_IO_REQ, max_ios));
+
+ snic->max_tag_id = shost->can_queue;
+
+ ret = scsi_init_shared_tag_map(shost, snic->max_tag_id);
+ if (ret) {
+ SNIC_HOST_ERR(shost,
+ "Unable to alloc shared tag map. %d\n",
+ ret);
+
+ goto err_dev_close;
+ }
+
+ shost->max_lun = snic->config.luns_per_tgt;
+ shost->max_id = SNIC_MAX_TARGET;
+
+ shost->max_cmd_len = MAX_COMMAND_SIZE; /*defined in scsi_cmnd.h*/
+
+ snic_get_res_counts(snic);
+
+ /*
+ * Assumption: Only MSIx is supported
+ */
+ ret = snic_set_intr_mode(snic);
+ if (ret) {
+ SNIC_HOST_ERR(shost,
+ "Failed to set intr mode aborting. %d\n",
+ ret);
+
+ goto err_dev_close;
+ }
+
+ ret = snic_alloc_vnic_res(snic);
+ if (ret) {
+ SNIC_HOST_ERR(shost,
+ "Failed to alloc vNIC resources aborting. %d\n",
+ ret);
+
+ goto err_clear_intr;
+ }
+
+ /* Initialize specific lists */
+ INIT_LIST_HEAD(&snic->list);
+
+ /*
+ * spl_cmd_list for maintaining snic specific cmds
+ * such as EXCH_VER_REQ, REPORT_TARGETS etc
+ */
+ INIT_LIST_HEAD(&snic->spl_cmd_list);
+ spin_lock_init(&snic->spl_cmd_lock);
+
+ /* initialize all snic locks */
+ spin_lock_init(&snic->snic_lock);
+
+ for (i = 0; i < SNIC_WQ_MAX; i++)
+ spin_lock_init(&snic->wq_lock[i]);
+
+ for (i = 0; i < SNIC_IO_LOCKS; i++)
+ spin_lock_init(&snic->io_req_lock[i]);
+
+ pool = mempool_create_slab_pool(2,
+ snic_glob->req_cache[SNIC_REQ_CACHE_DFLT_SGL]);
+ if (!pool) {
+ SNIC_HOST_ERR(shost, "dflt sgl pool creation failed\n");
+
+ goto err_free_res;
+ }
+
+ snic->req_pool[SNIC_REQ_CACHE_DFLT_SGL] = pool;
+
+ pool = mempool_create_slab_pool(2,
+ snic_glob->req_cache[SNIC_REQ_CACHE_MAX_SGL]);
+ if (!pool) {
+ SNIC_HOST_ERR(shost, "max sgl pool creation failed\n");
+
+ goto err_free_dflt_sgl_pool;
+ }
+
+ snic->req_pool[SNIC_REQ_CACHE_MAX_SGL] = pool;
+
+ pool = mempool_create_slab_pool(2,
+ snic_glob->req_cache[SNIC_REQ_TM_CACHE]);
+ if (!pool) {
+ SNIC_HOST_ERR(shost, "snic tmreq info pool creation failed.\n");
+
+ goto err_free_max_sgl_pool;
+ }
+
+ snic->req_pool[SNIC_REQ_TM_CACHE] = pool;
+
+ /* Initialize snic state */
+ atomic_set(&snic->state, SNIC_INIT);
+
+ atomic_set(&snic->ios_inflight, 0);
+
+ /* Setup notification buffer area */
+ ret = snic_notify_set(snic);
+ if (ret) {
+ SNIC_HOST_ERR(shost,
+ "Failed to alloc notify buffer aborting. %d\n",
+ ret);
+
+ goto err_free_tmreq_pool;
+ }
+
+ /*
+ * Initialization done with PCI system, hardware, firmware.
+ * Add shost to SCSI
+ */
+ ret = snic_add_host(shost, pdev);
+ if (ret) {
+ SNIC_HOST_ERR(shost,
+ "Adding scsi host Failed ... exiting. %d\n",
+ ret);
+
+ goto err_notify_unset;
+ }
+
+ spin_lock_irqsave(&snic_glob->snic_list_lock, flags);
+ list_add_tail(&snic->list, &snic_glob->snic_list);
+ spin_unlock_irqrestore(&snic_glob->snic_list_lock, flags);
+
+ snic_disc_init(&snic->disc);
+ INIT_WORK(&snic->tgt_work, snic_handle_tgt_disc);
+ INIT_WORK(&snic->disc_work, snic_handle_disc);
+ INIT_WORK(&snic->link_work, snic_handle_link);
+
+ /* Enable all queues */
+ for (i = 0; i < snic->wq_count; i++)
+ vnic_wq_enable(&snic->wq[i]);
+
+ ret = vnic_dev_enable_wait(snic->vdev);
+ if (ret) {
+ SNIC_HOST_ERR(shost,
+ "vNIC dev enable failed w/ error %d\n",
+ ret);
+
+ goto err_vdev_enable;
+ }
+
+ ret = snic_request_intr(snic);
+ if (ret) {
+ SNIC_HOST_ERR(shost, "Unable to request irq. %d\n", ret);
+
+ goto err_req_intr;
+ }
+
+ for (i = 0; i < snic->intr_count; i++)
+ vnic_intr_unmask(&snic->intr[i]);
+
+ snic_set_state(snic, SNIC_ONLINE);
+
+ /* Get snic params */
+ ret = snic_get_conf(snic);
+ if (ret) {
+ SNIC_HOST_ERR(shost,
+ "Failed to get snic io config from FW w err %d\n",
+ ret);
+
+ goto err_get_conf;
+ }
+
+ ret = snic_disc_start(snic);
+ if (ret) {
+ SNIC_HOST_ERR(shost, "snic_probe:Discovery Failed w err = %d\n",
+ ret);
+
+ goto err_get_conf;
+ }
+
+ SNIC_HOST_INFO(shost, "SNIC Device Probe Successful.\n");
+
+ return 0;
+
+err_get_conf:
+ snic_free_all_untagged_reqs(snic);
+
+ for (i = 0; i < snic->intr_count; i++)
+ vnic_intr_mask(&snic->intr[i]);
+
+ snic_free_intr(snic);
+
+err_req_intr:
+ vnic_dev_disable(snic->vdev);
+
+err_vdev_enable:
+ for (i = 0; i < snic->wq_count; i++) {
+ int rc = 0;
+
+ rc = vnic_wq_disable(&snic->wq[i]);
+ if (rc) {
+ SNIC_HOST_ERR(shost,
+ "WQ Disable Failed w/ err = %d\n", rc);
+
+ break;
+ }
+ }
+ snic_del_host(snic->shost);
+
+err_notify_unset:
+ vnic_dev_notify_unset(snic->vdev);
+
+err_free_tmreq_pool:
+ mempool_destroy(snic->req_pool[SNIC_REQ_TM_CACHE]);
+
+err_free_max_sgl_pool:
+ mempool_destroy(snic->req_pool[SNIC_REQ_CACHE_MAX_SGL]);
+
+err_free_dflt_sgl_pool:
+ mempool_destroy(snic->req_pool[SNIC_REQ_CACHE_DFLT_SGL]);
+
+err_free_res:
+ snic_free_vnic_res(snic);
+
+err_clear_intr:
+ snic_clear_intr_mode(snic);
+
+err_dev_close:
+ vnic_dev_close(snic->vdev);
+
+err_vnic_unreg:
+ vnic_dev_unregister(snic->vdev);
+
+err_iounmap:
+ snic_iounmap(snic);
+
+err_rel_regions:
+ pci_release_regions(pdev);
+
+err_pci_disable:
+ pci_disable_device(pdev);
+
+err_free_snic:
+#ifdef SNIC_DEBUG_FS
+ snic_stats_debugfs_remove(snic);
+#endif
+ scsi_host_put(shost);
+ pci_set_drvdata(pdev, NULL);
+
+prob_end:
+ SNIC_INFO("sNIC device : bus %d: slot %d: fn %d Registration Failed.\n",
+ pdev->bus->number, PCI_SLOT(pdev->devfn),
+ PCI_FUNC(pdev->devfn));
+
+ return ret;
+} /* end of snic_probe */
+
+
+/*
+ * snic_remove : invoked on unbinding the interface to cleanup the
+ * resources allocated in snic_probe on initialization.
+ */
+static void
+snic_remove(struct pci_dev *pdev)
+{
+ struct snic *snic = pci_get_drvdata(pdev);
+ unsigned long flags;
+
+ if (!snic) {
+ SNIC_INFO("sNIC dev: bus %d slot %d fn %d snic inst is null.\n",
+ pdev->bus->number, PCI_SLOT(pdev->devfn),
+ PCI_FUNC(pdev->devfn));
+
+ return;
+ }
+
+ /*
+ * Mark state so that the workqueue thread stops forwarding
+ * received frames and link events. ISR and other threads
+ * that can queue work items will also stop creating work
+ * items on the snic workqueue
+ */
+ snic_set_state(snic, SNIC_OFFLINE);
+ spin_lock_irqsave(&snic->snic_lock, flags);
+ snic->stop_link_events = 1;
+ spin_unlock_irqrestore(&snic->snic_lock, flags);
+
+ flush_workqueue(snic_glob->event_q);
+ snic_disc_term(snic);
+
+ spin_lock_irqsave(&snic->snic_lock, flags);
+ snic->in_remove = 1;
+ spin_unlock_irqrestore(&snic->snic_lock, flags);
+
+ /*
+ * This stops the snic device, masks all interrupts, Completed
+ * CQ entries are drained. Posted WQ/RQ/Copy-WQ entries are
+ * cleanup
+ */
+ snic_cleanup(snic);
+
+ spin_lock_irqsave(&snic_glob->snic_list_lock, flags);
+ list_del(&snic->list);
+ spin_unlock_irqrestore(&snic_glob->snic_list_lock, flags);
+
+ snic_tgt_del_all(snic);
+#ifdef SNIC_DEBUG_FS
+ snic_stats_debugfs_remove(snic);
+#endif
+ snic_del_host(snic->shost);
+
+ vnic_dev_notify_unset(snic->vdev);
+ snic_free_intr(snic);
+ snic_free_vnic_res(snic);
+ snic_clear_intr_mode(snic);
+ vnic_dev_close(snic->vdev);
+ vnic_dev_unregister(snic->vdev);
+ snic_iounmap(snic);
+ pci_release_regions(pdev);
+ pci_disable_device(pdev);
+ pci_set_drvdata(pdev, NULL);
+
+ /* this frees Scsi_Host and snic memory (continuous chunk) */
+ scsi_host_put(snic->shost);
+} /* end of snic_remove */
+
+
+struct snic_global *snic_glob;
+
+/*
+ * snic_global_data_init: Initialize SNIC Global Data
+ * Notes: All the global lists, variables should be part of global data
+ * this helps in debugging.
+ */
+static int
+snic_global_data_init(void)
+{
+ int ret = 0;
+ struct kmem_cache *cachep;
+ ssize_t len = 0;
+
+ snic_glob = kzalloc(sizeof(*snic_glob), GFP_KERNEL);
+
+ if (!snic_glob) {
+ SNIC_ERR("Failed to allocate Global Context.\n");
+
+ ret = -ENOMEM;
+ goto gdi_end;
+ }
+
+#ifdef SNIC_DEBUG_FS
+ /* Debugfs related Initialization */
+ /* Create debugfs entries for snic */
+ ret = snic_debugfs_init();
+ if (ret < 0) {
+ SNIC_ERR("Failed to create sysfs dir for tracing and stats.\n");
+ snic_debugfs_term();
+ /* continue even if it fails */
+ }
+
+ /* Trace related Initialization */
+ /* Allocate memory for trace buffer */
+ ret = snic_trc_init();
+ if (ret < 0) {
+ SNIC_ERR("Trace buffer init failed, SNIC tracing disabled\n");
+ snic_trc_free();
+ /* continue even if it fails */
+ }
+
+#endif
+ INIT_LIST_HEAD(&snic_glob->snic_list);
+ spin_lock_init(&snic_glob->snic_list_lock);
+
+ /* Create a cache for allocation of snic_host_req+default size ESGLs */
+ len = sizeof(struct snic_req_info);
+ len += sizeof(struct snic_host_req) + sizeof(struct snic_dflt_sgl);
+ cachep = kmem_cache_create("snic_req_dfltsgl", len, SNIC_SG_DESC_ALIGN,
+ SLAB_HWCACHE_ALIGN, NULL);
+ if (!cachep) {
+ SNIC_ERR("Failed to create snic default sgl slab\n");
+ ret = -ENOMEM;
+
+ goto err_dflt_req_slab;
+ }
+ snic_glob->req_cache[SNIC_REQ_CACHE_DFLT_SGL] = cachep;
+
+ /* Create a cache for allocation of max size Extended SGLs */
+ len = sizeof(struct snic_req_info);
+ len += sizeof(struct snic_host_req) + sizeof(struct snic_max_sgl);
+ cachep = kmem_cache_create("snic_req_maxsgl", len, SNIC_SG_DESC_ALIGN,
+ SLAB_HWCACHE_ALIGN, NULL);
+ if (!cachep) {
+ SNIC_ERR("Failed to create snic max sgl slab\n");
+ ret = -ENOMEM;
+
+ goto err_max_req_slab;
+ }
+ snic_glob->req_cache[SNIC_REQ_CACHE_MAX_SGL] = cachep;
+
+ len = sizeof(struct snic_host_req);
+ cachep = kmem_cache_create("snic_req_maxsgl", len, SNIC_SG_DESC_ALIGN,
+ SLAB_HWCACHE_ALIGN, NULL);
+ if (!cachep) {
+ SNIC_ERR("Failed to create snic tm req slab\n");
+ ret = -ENOMEM;
+
+ goto err_tmreq_slab;
+ }
+ snic_glob->req_cache[SNIC_REQ_TM_CACHE] = cachep;
+
+ /* snic_event queue */
+ snic_glob->event_q = create_singlethread_workqueue("snic_event_wq");
+ if (!snic_glob->event_q) {
+ SNIC_ERR("snic event queue create failed\n");
+ ret = -ENOMEM;
+
+ goto err_eventq;
+ }
+
+ return ret;
+
+err_eventq:
+ kmem_cache_destroy(snic_glob->req_cache[SNIC_REQ_TM_CACHE]);
+
+err_tmreq_slab:
+ kmem_cache_destroy(snic_glob->req_cache[SNIC_REQ_CACHE_MAX_SGL]);
+
+err_max_req_slab:
+ kmem_cache_destroy(snic_glob->req_cache[SNIC_REQ_CACHE_DFLT_SGL]);
+
+err_dflt_req_slab:
+#ifdef SNIC_DEBUG_FS
+ snic_trc_free();
+ snic_debugfs_term();
+#endif
+ kfree(snic_glob);
+ snic_glob = NULL;
+
+gdi_end:
+ return ret;
+} /* end of snic_glob_init */
+
+/*
+ * snic_global_data_cleanup : Frees SNIC Global Data
+ */
+static void
+snic_global_data_cleanup(void)
+{
+ SNIC_BUG_ON(snic_glob == NULL);
+
+ destroy_workqueue(snic_glob->event_q);
+ kmem_cache_destroy(snic_glob->req_cache[SNIC_REQ_TM_CACHE]);
+ kmem_cache_destroy(snic_glob->req_cache[SNIC_REQ_CACHE_MAX_SGL]);
+ kmem_cache_destroy(snic_glob->req_cache[SNIC_REQ_CACHE_DFLT_SGL]);
+
+#ifdef SNIC_DEBUG_FS
+ /* Freeing Trace Resources */
+ snic_trc_free();
+
+ /* Freeing Debugfs Resources */
+ snic_debugfs_term();
+#endif
+ kfree(snic_glob);
+ snic_glob = NULL;
+} /* end of snic_glob_cleanup */
+
+static struct pci_driver snic_driver = {
+ .name = SNIC_DRV_NAME,
+ .id_table = snic_id_table,
+ .probe = snic_probe,
+ .remove = snic_remove,
+};
+
+static int __init
+snic_init_module(void)
+{
+ int ret = 0;
+
+ SNIC_INFO("%s, ver %s\n", SNIC_DRV_DESCRIPTION, SNIC_DRV_VERSION);
+
+ ret = snic_global_data_init();
+ if (ret) {
+ SNIC_ERR("Failed to Initialize Global Data.\n");
+
+ return ret;
+ }
+
+ ret = pci_register_driver(&snic_driver);
+ if (ret < 0) {
+ SNIC_ERR("PCI driver register error\n");
+
+ goto err_pci_reg;
+ }
+
+ return ret;
+
+err_pci_reg:
+ snic_global_data_cleanup();
+
+ return ret;
+}
+
+static void __exit
+snic_cleanup_module(void)
+{
+ pci_unregister_driver(&snic_driver);
+ snic_global_data_cleanup();
+}
+
+module_init(snic_init_module);
+module_exit(snic_cleanup_module);
+
+MODULE_LICENSE("GPL v2");
+MODULE_DESCRIPTION(SNIC_DRV_DESCRIPTION);
+MODULE_VERSION(SNIC_DRV_VERSION);
+MODULE_DEVICE_TABLE(pci, snic_id_table);
+MODULE_AUTHOR("Narsimhulu Musini <nmusini@cisco.com>, "
+ "Sesidhar Baddela <sebaddel@cisco.com>");
--
1.8.5.4
^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH v4 2/9] snic:Add interrupt, resource firmware interfaces
2015-04-09 11:49 [PATCH v4 0/9] snic:initial submission of snic driver for Cisco SCSI HBA Narsimhulu Musini
2015-04-09 11:49 ` [PATCH v4 1/9] snic: snic module infrastructure Narsimhulu Musini
@ 2015-04-09 11:49 ` Narsimhulu Musini
2015-04-09 12:46 ` Hannes Reinecke
2015-04-09 11:49 ` [PATCH v4 3/9] snic:Add meta request, handling of meta requests Narsimhulu Musini
` (6 subsequent siblings)
8 siblings, 1 reply; 24+ messages in thread
From: Narsimhulu Musini @ 2015-04-09 11:49 UTC (permalink / raw)
To: JBottomley, linux-scsi, hare, hch; +Cc: Narsimhulu Musini, Sesidhar Baddela
snic_fwint.h contains firmware interface structures.
snic_res.h contains firmware request initialization
snic_res.c contains retrieval of resource configuration, and allocation,
and initialization of HW Queues.
snic_isr.c contains interrupt request, release, and handling
Signed-off-by: Narsimhulu Musini <nmusini@cisco.com>
Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
---
* v4
- Updated comment on future changes when hardware supports multiple queues.
* v3
- Cleaned up unused structure definitions and functions.
drivers/scsi/snic/snic_fwint.h | 525 +++++++++++++++++++++++++++++++++++++++++
drivers/scsi/snic/snic_isr.c | 209 ++++++++++++++++
drivers/scsi/snic/snic_res.c | 297 +++++++++++++++++++++++
drivers/scsi/snic/snic_res.h | 96 ++++++++
4 files changed, 1127 insertions(+)
create mode 100644 drivers/scsi/snic/snic_fwint.h
create mode 100644 drivers/scsi/snic/snic_isr.c
create mode 100644 drivers/scsi/snic/snic_res.c
create mode 100644 drivers/scsi/snic/snic_res.h
diff --git a/drivers/scsi/snic/snic_fwint.h b/drivers/scsi/snic/snic_fwint.h
new file mode 100644
index 0000000..1142e52
--- /dev/null
+++ b/drivers/scsi/snic/snic_fwint.h
@@ -0,0 +1,525 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef __SNIC_FWINT_H
+#define __SNIC_FWINT_H
+
+#define SNIC_CDB_LEN 32 /* SCSI CDB size 32, can be used for 16 bytes */
+#define LUN_ADDR_LEN 8
+
+/*
+ * Command entry type
+ */
+enum snic_io_type {
+ /*
+ * Initiator request types
+ */
+ SNIC_REQ_REPORT_TGTS = 0x2, /* Report Targets */
+ SNIC_REQ_ICMND, /* Initiator command for SCSI IO */
+ SNIC_REQ_ITMF, /* Initiator command for Task Mgmt */
+ SNIC_REQ_HBA_RESET, /* SNIC Reset */
+ SNIC_REQ_EXCH_VER, /* Exchange Version Information */
+ SNIC_REQ_TGT_INFO, /* Backend/Target Information */
+ SNIC_REQ_BOOT_LUNS,
+
+ /*
+ * Response type
+ */
+ SNIC_RSP_REPORT_TGTS_CMPL = 0x12,/* Report Targets Completion */
+ SNIC_RSP_ICMND_CMPL, /* SCSI IO Completion */
+ SNIC_RSP_ITMF_CMPL, /* Task Management Completion */
+ SNIC_RSP_HBA_RESET_CMPL, /* SNIC Reset Completion */
+ SNIC_RSP_EXCH_VER_CMPL, /* Exchange Version Completion*/
+ SNIC_RSP_BOOT_LUNS_CMPL,
+
+ /*
+ * Misc Request types
+ */
+ SNIC_MSG_ACK = 0x80, /* Ack: snic_notify_msg */
+ SNIC_MSG_ASYNC_EVNOTIFY, /* Asynchronous Event Notification */
+}; /* end of enum snic_io_type */
+
+
+/*
+ * Header status codes from firmware
+ */
+enum snic_io_status {
+ SNIC_STAT_IO_SUCCESS = 0, /* request was successful */
+
+ /*
+ * If a request to the fw is rejected, the original request header
+ * will be returned with the status set to one of the following:
+ */
+ SNIC_STAT_INVALID_HDR, /* header contains invalid data */
+ SNIC_STAT_OUT_OF_RES, /* out of resources to complete request */
+ SNIC_STAT_INVALID_PARM, /* some parameter in request is not valid */
+ SNIC_STAT_REQ_NOT_SUP, /* req type is not supported */
+ SNIC_STAT_IO_NOT_FOUND, /* requested IO was not found */
+
+ /*
+ * Once a request is processed, the fw will usually return
+ * a cmpl message type. In cases where errors occurred,
+ * the header status would be filled in with one of the following:
+ */
+ SNIC_STAT_ABORTED, /* req was aborted */
+ SNIC_STAT_TIMEOUT, /* req was timed out */
+ SNIC_STAT_SGL_INVALID, /* req was aborted due to sgl error */
+ SNIC_STAT_DATA_CNT_MISMATCH, /*recv/sent more/less data than expec */
+ SNIC_STAT_FW_ERR, /* req was terminated due to fw error */
+ SNIC_STAT_ITMF_REJECT, /* itmf req was rejected by target */
+ SNIC_STAT_ITMF_FAIL, /* itmf req was failed */
+ SNIC_STAT_ITMF_INCORRECT_LUN, /* itmf req has incorrect LUN id*/
+ SNIC_STAT_CMND_REJECT, /* req was invalid and rejected */
+ SNIC_STAT_DEV_OFFLINE, /* req sent to offline device */
+ SNIC_STAT_NO_BOOTLUN,
+ SNIC_STAT_SCSI_ERR, /* SCSI error returned by Target. */
+ SNIC_STAT_NOT_READY, /* sNIC Subsystem is not ready */
+ SNIC_STAT_FATAL_ERROR, /* sNIC is in unrecoverable state */
+}; /* end of enum snic_io_status */
+
+/*
+ * snic_io_hdr : host <--> firmare
+ *
+ * for any other message that will be queued to firmware should
+ * have the following request header
+ */
+struct snic_io_hdr {
+ u32 hid;
+ u32 cmnd_id; /* tag here */
+ u64 init_ctx; /* initiator context */
+ u8 type; /* request/response type */
+ u8 status; /* header status entry */
+ u8 protocol; /* Protocol specific, may needed for RoCE*/
+ u8 flags;
+ u16 sg_cnt;
+ u16 resvd;
+};
+
+/* auxillary funciton for encoding the snic_io_hdr */
+static inline void
+snic_io_hdr_enc(struct snic_io_hdr *hdr, u8 typ, u8 status, u32 id, u32 hid,
+ u16 sg_cnt, u64 ctx)
+{
+ hdr->type = typ;
+ hdr->status = status;
+ hdr->protocol = 0;
+ hdr->hid = hid;
+ hdr->cmnd_id = id;
+ hdr->sg_cnt = sg_cnt;
+ hdr->init_ctx = ctx;
+ hdr->flags = 0;
+}
+
+/* auxillary funciton for decoding the snic_io_hdr */
+static inline void
+snic_io_hdr_dec(struct snic_io_hdr *hdr, u8 *typ, u8 *stat, u32 *cmnd_id,
+ u32 *hid, u64 *ctx)
+{
+ *typ = hdr->type;
+ *stat = hdr->status;
+ *hid = hdr->hid;
+ *cmnd_id = hdr->cmnd_id;
+ *ctx = hdr->init_ctx;
+}
+
+/*
+ * snic_host_info: host -> firmware
+ *
+ * Used for sending host information to firmware, and request fw version
+ */
+struct snic_exch_ver_req {
+ u32 drvr_ver; /* for debugging, when fw dump captured */
+ u32 os_type; /* for OS specific features */
+};
+
+/*
+ * os_type flags
+ * Bit 0-7 : OS information
+ * Bit 8-31: Feature/Capability Information
+ */
+#define SNIC_OS_LINUX 0x1
+#define SNIC_OS_WIN 0x2
+#define SNIC_OS_ESX 0x3
+
+/*
+ * HBA Capabilities
+ * Bit 1: Reserved.
+ * Bit 2: Dynamic Discovery of LUNs.
+ * Bit 3: Async event notifications on on tgt online/offline events.
+ * Bit 4: IO timeout support in FW.
+ * Bit 5-31: Reserved.
+ */
+#define SNIC_HBA_CAP_DDL 0x02 /* Supports Dynamic Discovery of LUNs */
+#define SNIC_HBA_CAP_AEN 0x04 /* Supports Async Event Noitifcation */
+#define SNIC_HBA_CAP_TMO 0x08 /* Supports IO timeout in FW */
+
+/*
+ * snic_exch_ver_rsp : firmware -> host
+ *
+ * Used by firmware to send response to version request
+ */
+struct snic_exch_ver_rsp {
+ u32 version;
+ u32 hid;
+ u32 max_concur_ios; /* max concurrent ios */
+ u32 max_sgs_per_cmd; /* max sgls per IO */
+ u32 max_io_sz; /* max io size supported */
+ u32 hba_cap; /* hba capabilities */
+ u32 max_tgts; /* max tgts supported */
+ u16 io_timeout; /* FW extended timeout */
+ u16 rsvd;
+};
+
+
+/*
+ * snic_report_tgts : host -> firmware request
+ *
+ * Used by the host to request list of targets
+ */
+struct snic_report_tgts {
+ u16 sg_cnt;
+ u16 flags; /* specific flags from fw */
+ u8 _resvd[4];
+ u64 sg_addr; /* Points to SGL */
+ u64 sense_addr;
+};
+
+enum snic_type {
+ SNIC_NONE = 0x0,
+ SNIC_DAS,
+ SNIC_SAN,
+};
+
+
+/* Report Target Response */
+enum snic_tgt_type {
+ SNIC_TGT_NONE = 0x0,
+ SNIC_TGT_DAS, /* DAS Target */
+ SNIC_TGT_SAN, /* SAN Target */
+};
+
+/* target id format */
+struct snic_tgt_id {
+ u32 tgt_id; /* target id */
+ u16 tgt_type; /* tgt type */
+ u16 vnic_id; /* corresponding vnic id */
+};
+
+/*
+ * snic_report_tgts_cmpl : firmware -> host response
+ *
+ * Used by firmware to send response to Report Targets request
+ */
+struct snic_report_tgts_cmpl {
+ u32 tgt_cnt; /* Number of Targets accessible */
+ u32 _resvd;
+};
+
+/*
+ * Command flags
+ *
+ * Bit 0: Read flags
+ * Bit 1: Write flag
+ * Bit 2: ESGL - sg/esg array contains extended sg
+ * ESGE - is a host buffer contains sg elements
+ * Bit 3-4: Task Attributes
+ * 00b - simple
+ * 01b - head of queue
+ * 10b - ordered
+ * Bit 5-7: Priority - future use
+ * Bit 8-15: Reserved
+ */
+
+#define SNIC_ICMND_WR 0x01 /* write command */
+#define SNIC_ICMND_RD 0x02 /* read command */
+#define SNIC_ICMND_ESGL 0x04 /* SGE/ESGE array contains valid data*/
+
+/*
+ * Priority/Task Attribute settings
+ */
+#define SNIC_ICMND_TSK_SHIFT 2 /* task attr starts at bit 2 */
+#define SNIC_ICMND_TSK_MASK(x) ((x>>SNIC_ICMND_TSK_SHIFT) & ~(0xffff))
+#define SNIC_ICMND_TSK_SIMPLE 0 /* simple task attr */
+#define SNIC_ICMND_TSK_HEAD_OF_QUEUE 1 /* head of qeuue task attr */
+#define SNIC_ICMND_TSK_ORDERED 2 /* ordered task attr */
+
+#define SNIC_ICMND_PRI_SHIFT 5 /* prio val starts at bit 5 */
+
+/*
+ * snic_icmnd : host-> firmware request
+ *
+ * used for sending out an initiator SCSI 16/32-byte command
+ */
+struct snic_icmnd {
+ u16 sg_cnt; /* Number of SG Elements */
+ u16 flags; /* flags */
+ u32 sense_len; /* Sense buffer length */
+ u64 tgt_id; /* Destination Target ID */
+ u64 lun_id; /* Destination LUN ID */
+ u8 cdb_len;
+ u8 _resvd;
+ u16 time_out; /* ms time for Res allocations fw to handle io*/
+ u32 data_len; /* Total number of bytes to be transferred */
+ u8 cdb[SNIC_CDB_LEN];
+ u64 sg_addr; /* Points to SG List */
+ u64 sense_addr; /* Sense buffer address */
+};
+
+
+/* Response flags */
+/* Bit 0: Under run
+ * Bit 1: Over Run
+ * Bit 2-7: Reserved
+ */
+#define SNIC_ICMND_CMPL_UNDR_RUN 0x01 /* resid under and valid */
+#define SNIC_ICMND_CMPL_OVER_RUN 0x02 /* resid over and valid */
+
+/*
+ * snic_icmnd_cmpl: firmware -> host response
+ *
+ * Used for sending the host a response to an icmnd (initiator command)
+ */
+struct snic_icmnd_cmpl {
+ u8 scsi_status; /* value as per SAM */
+ u8 flags;
+ u16 sense_len; /* Sense Length */
+ u32 resid; /* Residue : # bytes under or over run */
+};
+
+/*
+ * snic_itmf: host->firmware request
+ *
+ * used for requesting the firmware to abort a request and/or send out
+ * a task management function
+ *
+ * the req_id field is valid in case of abort task and clear task
+ */
+struct snic_itmf {
+ u8 tm_type; /* SCSI Task Management request */
+ u8 resvd;
+ u16 flags; /* flags */
+ u32 req_id; /* Command id of snic req to be aborted */
+ u64 tgt_id; /* Target ID */
+ u64 lun_id; /* Destination LUN ID */
+ u16 timeout; /* in sec */
+};
+
+/*
+ * Task Management Request
+ */
+enum snic_itmf_tm_type {
+ SNIC_ITMF_ABTS_TASK = 0x01, /* Abort Task */
+ SNIC_ITMF_ABTS_TASK_SET, /* Abort Task Set */
+ SNIC_ITMF_CLR_TASK, /* Clear Task */
+ SNIC_ITMF_CLR_TASKSET, /* Clear Task Set */
+ SNIC_ITMF_LUN_RESET, /* Lun Reset */
+ SNIC_ITMF_ABTS_TASK_TERM, /* Supported for SAN Targets */
+};
+
+/*
+ * snic_itmf_cmpl: firmware -> host resposne
+ *
+ * used for sending the host a response for a itmf request
+ */
+struct snic_itmf_cmpl {
+ u32 nterminated; /* # IOs terminated as a result of tmf */
+ u8 flags; /* flags */
+ u8 _resvd[3];
+};
+
+/*
+ * itmfl_cmpl flags
+ * Bit 0 : 1 - Num terminated field valid
+ * Bit 1 - 7 : Reserved
+ */
+#define SNIC_NUM_TERM_VALID 0x01 /* Number of IOs terminated */
+
+/*
+ * snic_hba_reset: host -> firmware request
+ *
+ * used for requesting firmware to reset snic
+ */
+struct snic_hba_reset {
+ u16 flags; /* flags */
+ u8 _resvd[6];
+};
+
+/*
+ * snic_hba_reset_cmpl: firmware -> host response
+ *
+ * Used by firmware to respond to the host's hba reset request
+ */
+struct snic_hba_reset_cmpl {
+ u8 flags; /* flags : more info needs to be added*/
+ u8 _resvd[7];
+};
+
+/*
+ * snic_notify_msg: firmware -> host response
+ *
+ * Used by firmware to notify host of the last work queue entry received
+ */
+struct snic_notify_msg {
+ u32 wqe_num; /* wq entry number */
+ u8 flags; /* flags, macros */
+ u8 _resvd[4];
+};
+
+
+#define SNIC_EVDATA_LEN 24 /* in bytes */
+/* snic_async_evnotify: firmware -> host notification
+ *
+ * Used by firmware to notify the host about configuration/state changes
+ */
+struct snic_async_evnotify {
+ u8 FLS_EVENT_DESC; /* TODO: ?? */
+ u8 vnic; /* vnic id */
+ u8 _resvd[2];
+ u32 ev_id; /* Event ID */
+ u8 ev_data[SNIC_EVDATA_LEN]; /* Event Data */
+ u8 _resvd2[4];
+};
+
+/* async event flags */
+enum snic_ev_type {
+ SNIC_EV_TGT_OFFLINE = 0x01, /* Target Offline, PL contains TGT ID */
+ SNIC_EV_TGT_ONLINE, /* Target Online, PL contains TGT ID */
+ SNIC_EV_LUN_OFFLINE, /* LUN Offline, PL contains LUN ID */
+ SNIC_EV_LUN_ONLINE, /* LUN Online, PL contains LUN ID */
+ SNIC_EV_CONF_CHG, /* Dev Config/Attr Change Event */
+ SNIC_EV_TGT_ADDED, /* TODO:Target Added, PL contains ?? */
+ SNIC_EV_TGT_DELTD, /* Target Del'd, PL contains TGT ID */
+ SNIC_EV_LUN_ADDED, /* TODO:LUN Added, PL contains ?? */
+ SNIC_EV_LUN_DELTD, /* LUN Del'd, PL cont. TGT & LUN ID */
+
+ SNIC_EV_DISC_CMPL = 0x10, /* Discovery Completed Event */
+};
+
+
+#define SNIC_HOST_REQ_LEN 128 /*Exp length of host req, wq desc sz*/
+/* Payload 88 bytes = 128 - 24 - 16 */
+#define SNIC_HOST_REQ_PAYLOAD ((int)(SNIC_HOST_REQ_LEN - \
+ sizeof(struct snic_io_hdr) - \
+ (2 * sizeof(u64))))
+
+/*
+ * snic_host_req: host -> firmware request
+ *
+ * Basic structure for all snic requests that are sent from the host to
+ * firmware. They are 128 bytes in size.
+ */
+struct snic_host_req {
+ u64 ctrl_data[2]; /*16 bytes - Control Data */
+ struct snic_io_hdr hdr;
+ union {
+ /*
+ * Entry specific space, last byte contains color
+ */
+ u8 buf[SNIC_HOST_REQ_PAYLOAD];
+
+ /*
+ * Exchange firmware version
+ */
+ struct snic_exch_ver_req exch_ver;
+
+ /* report targets */
+ struct snic_report_tgts rpt_tgts;
+
+ /* io request */
+ struct snic_icmnd icmnd;
+
+ /* task management request */
+ struct snic_itmf itmf;
+
+ /* hba reset */
+ struct snic_hba_reset reset;
+ } u;
+}; /* end of snic_host_req structure */
+
+
+#define SNIC_FW_REQ_LEN 64 /* Expected length of fw req */
+struct snic_fw_req {
+ struct snic_io_hdr hdr;
+ union {
+ /*
+ * Entry specific space, last byte contains color
+ */
+ u8 buf[SNIC_FW_REQ_LEN - sizeof(struct snic_io_hdr)];
+
+ /* Exchange Version Response */
+ struct snic_exch_ver_rsp exch_ver_cmpl;
+
+ /* Report Targets Response */
+ struct snic_report_tgts_cmpl rpt_tgts_cmpl;
+
+ /* scsi response */
+ struct snic_icmnd_cmpl icmnd_cmpl;
+
+ /* task management response */
+ struct snic_itmf_cmpl itmf_cmpl;
+
+ /* hba reset response */
+ struct snic_hba_reset_cmpl reset_cmpl;
+
+ /* notify message */
+ struct snic_notify_msg ack;
+
+ /* async notification event */
+ struct snic_async_evnotify async_ev;
+
+ } u;
+}; /* end of snic_fw_req structure */
+
+/*
+ * Auxillary macro to verify specific snic req/cmpl structures
+ * to ensure that it will be aligned to 64 bit, and not using
+ * color bit field
+ */
+#define VERIFY_REQ_SZ(x)
+#define VERIFY_CMPL_SZ(x)
+
+/*
+ * Access routines to encode and decode the color bit, which is the most
+ * significant bit of the structure.
+ */
+static inline void
+snic_color_enc(struct snic_fw_req *req, u8 color)
+{
+ u8 *c = ((u8 *) req) + sizeof(struct snic_fw_req) - 1;
+
+ if (color)
+ *c |= 0x80;
+ else
+ *c &= ~0x80;
+}
+
+static inline void
+snic_color_dec(struct snic_fw_req *req, u8 *color)
+{
+ u8 *c = ((u8 *) req) + sizeof(struct snic_fw_req) - 1;
+
+ *color = *c >> 7;
+
+ /* Make sure color bit is read from desc *before* other fields
+ * are read from desc. Hardware guarantees color bit is last
+ * bit (byte) written. Adding the rmb() prevents the compiler
+ * and/or CPU from reordering the reads which would potentially
+ * result in reading stale values.
+ */
+ rmb();
+}
+#endif /* end of __SNIC_FWINT_H */
diff --git a/drivers/scsi/snic/snic_isr.c b/drivers/scsi/snic/snic_isr.c
new file mode 100644
index 0000000..5d5cb0c
--- /dev/null
+++ b/drivers/scsi/snic/snic_isr.c
@@ -0,0 +1,209 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ * [Insert appropriate license here when releasing outside of Cisco]
+ *
+ */
+
+#include <linux/string.h>
+#include <linux/errno.h>
+#include <linux/pci.h>
+#include <linux/interrupt.h>
+
+#include "vnic_dev.h"
+#include "vnic_intr.h"
+#include "vnic_stats.h"
+#include "snic_io.h"
+#include "snic.h"
+
+
+/*
+ * snic_isr_msix_wq : MSIx ISR for work queue.
+ */
+
+static irqreturn_t
+snic_isr_msix_wq(int irq, void *data)
+{
+ struct snic *snic = data;
+ unsigned long wq_work_done = 0;
+
+ snic->s_stats.misc.last_isr_time = jiffies;
+ atomic64_inc(&snic->s_stats.misc.isr_cnt);
+
+ wq_work_done = snic_wq_cmpl_handler(snic, -1);
+ vnic_intr_return_credits(&snic->intr[SNIC_MSIX_WQ],
+ wq_work_done,
+ 1 /* unmask intr */,
+ 1 /* reset intr timer */);
+
+ return IRQ_HANDLED;
+} /* end of snic_isr_msix_wq */
+
+static irqreturn_t
+snic_isr_msix_io_cmpl(int irq, void *data)
+{
+ struct snic *snic = data;
+ unsigned long iocmpl_work_done = 0;
+
+ snic->s_stats.misc.last_isr_time = jiffies;
+ atomic64_inc(&snic->s_stats.misc.isr_cnt);
+
+ iocmpl_work_done = snic_fwcq_cmpl_handler(snic, -1);
+ vnic_intr_return_credits(&snic->intr[SNIC_MSIX_IO_CMPL],
+ iocmpl_work_done,
+ 1 /* unmask intr */,
+ 1 /* reset intr timer */);
+
+ return IRQ_HANDLED;
+} /* end of snic_isr_msix_io_cmpl */
+
+static irqreturn_t
+snic_isr_msix_err_notify(int irq, void *data)
+{
+ struct snic *snic = data;
+
+ snic->s_stats.misc.last_isr_time = jiffies;
+ atomic64_inc(&snic->s_stats.misc.isr_cnt);
+
+ vnic_intr_return_all_credits(&snic->intr[SNIC_MSIX_ERR_NOTIFY]);
+ snic_log_q_error(snic);
+
+ /*Handling link events */
+ snic_handle_link_event(snic);
+
+ return IRQ_HANDLED;
+} /* end of snic_isr_msix_err_notify */
+
+
+void
+snic_free_intr(struct snic *snic)
+{
+ int i;
+
+ /* ONLY interrupt mode MSIX is supported */
+ for (i = 0; i < ARRAY_SIZE(snic->msix); i++) {
+ if (snic->msix[i].requested) {
+ free_irq(snic->msix_entry[i].vector,
+ snic->msix[i].devid);
+ }
+ }
+} /* end of snic_free_intr */
+
+int
+snic_request_intr(struct snic *snic)
+{
+ int ret = 0, i;
+
+#ifdef SNIC_DEBUG
+ enum vnic_dev_intr_mode intr_mode;
+
+ intr_mode = vnic_dev_get_intr_mode(snic->vdev);
+ SNIC_BUG_ON(intr_mode != VNIC_DEV_INTR_MODE_MSIX);
+#endif
+
+ /*
+ * Currently HW supports single WQ and CQ. So passing devid as snic.
+ * When hardware supports multiple WQs and CQs, one idea is
+ * to pass devid as corresponding WQ or CQ ptr and retrieve snic
+ * from queue ptr.
+ * Except for err_notify, which is always one.
+ */
+ sprintf(snic->msix[SNIC_MSIX_WQ].devname,
+ "%.11s-scsi-wq",
+ snic->name);
+ snic->msix[SNIC_MSIX_WQ].isr = snic_isr_msix_wq;
+ snic->msix[SNIC_MSIX_WQ].devid = snic;
+
+ sprintf(snic->msix[SNIC_MSIX_IO_CMPL].devname,
+ "%.11s-io-cmpl",
+ snic->name);
+ snic->msix[SNIC_MSIX_IO_CMPL].isr = snic_isr_msix_io_cmpl;
+ snic->msix[SNIC_MSIX_IO_CMPL].devid = snic;
+
+ sprintf(snic->msix[SNIC_MSIX_ERR_NOTIFY].devname,
+ "%.11s-err-notify",
+ snic->name);
+ snic->msix[SNIC_MSIX_ERR_NOTIFY].isr = snic_isr_msix_err_notify;
+ snic->msix[SNIC_MSIX_ERR_NOTIFY].devid = snic;
+
+ for (i = 0; i < ARRAY_SIZE(snic->msix); i++) {
+ ret = request_irq(snic->msix_entry[i].vector,
+ snic->msix[i].isr,
+ 0,
+ snic->msix[i].devname,
+ snic->msix[i].devid);
+ if (ret) {
+ SNIC_HOST_ERR(snic->shost,
+ "MSI-X: requrest_irq(%d) failed %d\n",
+ i,
+ ret);
+ snic_free_intr(snic);
+ break;
+ }
+ snic->msix[i].requested = 1;
+ }
+
+ return ret;
+} /* end of snic_requrest_intr */
+
+int
+snic_set_intr_mode(struct snic *snic)
+{
+ unsigned int n = ARRAY_SIZE(snic->wq);
+ unsigned int m = SNIC_CQ_IO_CMPL_MAX;
+ unsigned int i;
+
+ /*
+ * We need n WQs, m CQs, and n+m+1 INTRs
+ * (last INTR is used for WQ/CQ errors and notification area
+ */
+
+ BUILD_BUG_ON((ARRAY_SIZE(snic->wq) + SNIC_CQ_IO_CMPL_MAX) >
+ ARRAY_SIZE(snic->intr));
+ SNIC_BUG_ON(ARRAY_SIZE(snic->msix_entry) < (n + m + 1));
+
+ for (i = 0; i < (n + m + 1); i++)
+ snic->msix_entry[i].entry = i;
+
+ if (snic->wq_count >= n && snic->cq_count >= (n + m)) {
+ if (!pci_enable_msix(snic->pdev,
+ snic->msix_entry,
+ (n + m + 1))) {
+ snic->wq_count = n;
+ snic->cq_count = n + m;
+ snic->intr_count = n + m + 1;
+ snic->err_intr_offset = SNIC_MSIX_ERR_NOTIFY;
+
+ SNIC_ISR_DBG(snic->shost,
+ "Using MSI-X Interrupts\n");
+ vnic_dev_set_intr_mode(snic->vdev,
+ VNIC_DEV_INTR_MODE_MSIX);
+
+ return 0;
+ }
+ }
+
+ vnic_dev_set_intr_mode(snic->vdev, VNIC_DEV_INTR_MODE_UNKNOWN);
+
+ return -EINVAL;
+} /* end of snic_set_intr_mode */
+
+void
+snic_clear_intr_mode(struct snic *snic)
+{
+ pci_disable_msix(snic->pdev);
+
+ vnic_dev_set_intr_mode(snic->vdev, VNIC_DEV_INTR_MODE_INTX);
+}
diff --git a/drivers/scsi/snic/snic_res.c b/drivers/scsi/snic/snic_res.c
new file mode 100644
index 0000000..4b2e4b1
--- /dev/null
+++ b/drivers/scsi/snic/snic_res.c
@@ -0,0 +1,297 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/errno.h>
+#include <linux/types.h>
+#include <linux/pci.h>
+
+#include "wq_enet_desc.h"
+#include "cq_enet_desc.h"
+#include "vnic_resource.h"
+#include "vnic_dev.h"
+#include "vnic_wq.h"
+#include "vnic_cq.h"
+#include "vnic_intr.h"
+#include "vnic_stats.h"
+#include "snic.h"
+
+int
+snic_get_vnic_config(struct snic *snic)
+{
+ struct vnic_snic_config *c = &snic->config;
+ int ret;
+
+#define GET_CONFIG(m) \
+ do { \
+ ret = vnic_dev_spec(snic->vdev, \
+ offsetof(struct vnic_snic_config, m), \
+ sizeof(c->m), \
+ &c->m); \
+ if (ret) { \
+ SNIC_HOST_ERR(snic->shost, \
+ "Error getting %s, %d\n", #m, ret); \
+ return ret; \
+ } \
+ } while (0)
+
+ GET_CONFIG(wq_enet_desc_count);
+ GET_CONFIG(maxdatafieldsize);
+ GET_CONFIG(intr_timer);
+ GET_CONFIG(intr_timer_type);
+ GET_CONFIG(flags);
+ GET_CONFIG(io_throttle_count);
+ GET_CONFIG(port_down_timeout);
+ GET_CONFIG(port_down_io_retries);
+ GET_CONFIG(luns_per_tgt);
+ GET_CONFIG(xpt_type);
+ GET_CONFIG(hid);
+
+ c->wq_enet_desc_count = min_t(u32,
+ VNIC_SNIC_WQ_DESCS_MAX,
+ max_t(u32,
+ VNIC_SNIC_WQ_DESCS_MIN,
+ c->wq_enet_desc_count));
+
+ c->wq_enet_desc_count = ALIGN(c->wq_enet_desc_count, 16);
+
+ c->maxdatafieldsize = min_t(u32,
+ VNIC_SNIC_MAXDATAFIELDSIZE_MAX,
+ max_t(u32,
+ VNIC_SNIC_MAXDATAFIELDSIZE_MIN,
+ c->maxdatafieldsize));
+
+ c->io_throttle_count = min_t(u32,
+ VNIC_SNIC_IO_THROTTLE_COUNT_MAX,
+ max_t(u32,
+ VNIC_SNIC_IO_THROTTLE_COUNT_MIN,
+ c->io_throttle_count));
+
+ c->port_down_timeout = min_t(u32,
+ VNIC_SNIC_PORT_DOWN_TIMEOUT_MAX,
+ c->port_down_timeout);
+
+ c->port_down_io_retries = min_t(u32,
+ VNIC_SNIC_PORT_DOWN_IO_RETRIES_MAX,
+ c->port_down_io_retries);
+
+ c->luns_per_tgt = min_t(u32,
+ VNIC_SNIC_LUNS_PER_TARGET_MAX,
+ max_t(u32,
+ VNIC_SNIC_LUNS_PER_TARGET_MIN,
+ c->luns_per_tgt));
+
+ c->intr_timer = min_t(u32, VNIC_INTR_TIMER_MAX, c->intr_timer);
+
+ SNIC_INFO("vNIC resources wq %d\n", c->wq_enet_desc_count);
+ SNIC_INFO("vNIC mtu %d intr timer %d\n",
+ c->maxdatafieldsize,
+ c->intr_timer);
+
+ SNIC_INFO("vNIC flags 0x%x luns per tgt %d\n",
+ c->flags,
+ c->luns_per_tgt);
+
+ SNIC_INFO("vNIC io throttle count %d\n", c->io_throttle_count);
+ SNIC_INFO("vNIC port down timeout %d port down io retries %d\n",
+ c->port_down_timeout,
+ c->port_down_io_retries);
+
+ SNIC_INFO("vNIC back end type = %d\n", c->xpt_type);
+ SNIC_INFO("vNIC hid = %d\n", c->hid);
+
+ return 0;
+}
+
+void
+snic_get_res_counts(struct snic *snic)
+{
+ snic->wq_count = vnic_dev_get_res_count(snic->vdev, RES_TYPE_WQ);
+ SNIC_BUG_ON(snic->wq_count == 0);
+ snic->cq_count = vnic_dev_get_res_count(snic->vdev, RES_TYPE_CQ);
+ SNIC_BUG_ON(snic->cq_count == 0);
+ snic->intr_count = vnic_dev_get_res_count(snic->vdev,
+ RES_TYPE_INTR_CTRL);
+ SNIC_BUG_ON(snic->intr_count == 0);
+}
+
+void
+snic_free_vnic_res(struct snic *snic)
+{
+ unsigned int i;
+
+ for (i = 0; i < snic->wq_count; i++)
+ vnic_wq_free(&snic->wq[i]);
+
+ for (i = 0; i < snic->cq_count; i++)
+ vnic_cq_free(&snic->cq[i]);
+
+ for (i = 0; i < snic->intr_count; i++)
+ vnic_intr_free(&snic->intr[i]);
+}
+
+int
+snic_alloc_vnic_res(struct snic *snic)
+{
+ enum vnic_dev_intr_mode intr_mode;
+ unsigned int mask_on_assertion;
+ unsigned int intr_offset;
+ unsigned int err_intr_enable;
+ unsigned int err_intr_offset;
+ unsigned int i;
+ int ret;
+
+ intr_mode = vnic_dev_get_intr_mode(snic->vdev);
+
+ SNIC_INFO("vNIC interrupt mode: %s\n",
+ ((intr_mode == VNIC_DEV_INTR_MODE_INTX) ?
+ "Legacy PCI INTx" :
+ ((intr_mode == VNIC_DEV_INTR_MODE_MSI) ?
+ "MSI" :
+ ((intr_mode == VNIC_DEV_INTR_MODE_MSIX) ?
+ "MSI-X" : "Unknown"))));
+
+ /* only MSI-X is supported */
+ SNIC_BUG_ON(intr_mode != VNIC_DEV_INTR_MODE_MSIX);
+
+ SNIC_INFO("wq %d cq %d intr %d\n", snic->wq_count,
+ snic->cq_count,
+ snic->intr_count);
+
+
+ /* Allocate WQs used for SCSI IOs */
+ for (i = 0; i < snic->wq_count; i++) {
+ ret = vnic_wq_alloc(snic->vdev,
+ &snic->wq[i],
+ i,
+ snic->config.wq_enet_desc_count,
+ sizeof(struct wq_enet_desc));
+ if (ret)
+ goto error_cleanup;
+ }
+
+ /* CQ for each WQ */
+ for (i = 0; i < snic->wq_count; i++) {
+ ret = vnic_cq_alloc(snic->vdev,
+ &snic->cq[i],
+ i,
+ snic->config.wq_enet_desc_count,
+ sizeof(struct cq_enet_wq_desc));
+ if (ret)
+ goto error_cleanup;
+ }
+
+#ifdef SNIC_DEBUG
+ SNIC_BUG_ON(snic->cq_count != 2 * snic->wq_count);
+#endif
+ /* CQ for FW TO host */
+ for (i = snic->wq_count; i < snic->cq_count; i++) {
+ ret = vnic_cq_alloc(snic->vdev,
+ &snic->cq[i],
+ i,
+ (snic->config.wq_enet_desc_count * 3),
+ sizeof(struct snic_fw_req));
+ if (ret)
+ goto error_cleanup;
+ }
+
+ for (i = 0; i < snic->intr_count; i++) {
+ ret = vnic_intr_alloc(snic->vdev, &snic->intr[i], i);
+ if (ret)
+ goto error_cleanup;
+ }
+
+ /*
+ * Init WQ Resources.
+ * WQ[0 to n] points to CQ[0 to n-1]
+ * firmware to host comm points to CQ[n to m+1]
+ */
+ err_intr_enable = 1;
+ err_intr_offset = snic->err_intr_offset;
+
+ for (i = 0; i < snic->wq_count; i++) {
+ vnic_wq_init(&snic->wq[i],
+ i,
+ err_intr_enable,
+ err_intr_offset);
+ }
+
+ for (i = 0; i < snic->cq_count; i++) {
+ intr_offset = i;
+
+ vnic_cq_init(&snic->cq[i],
+ 0 /* flow_control_enable */,
+ 1 /* color_enable */,
+ 0 /* cq_head */,
+ 0 /* cq_tail */,
+ 1 /* cq_tail_color */,
+ 1 /* interrupt_enable */,
+ 1 /* cq_entry_enable */,
+ 0 /* cq_message_enable */,
+ intr_offset,
+ 0 /* cq_message_addr */);
+ }
+
+ /*
+ * Init INTR resources
+ * Assumption : snic is always in MSI-X mode
+ */
+ SNIC_BUG_ON(intr_mode != VNIC_DEV_INTR_MODE_MSIX);
+ mask_on_assertion = 1;
+
+ for (i = 0; i < snic->intr_count; i++) {
+ vnic_intr_init(&snic->intr[i],
+ snic->config.intr_timer,
+ snic->config.intr_timer_type,
+ mask_on_assertion);
+ }
+
+ /* init the stats memory by making the first call here */
+ ret = vnic_dev_stats_dump(snic->vdev, &snic->stats);
+ if (ret) {
+ SNIC_HOST_ERR(snic->shost,
+ "vnic_dev_stats_dump failed - x%x\n",
+ ret);
+ goto error_cleanup;
+ }
+
+ /* Clear LIF stats */
+ vnic_dev_stats_clear(snic->vdev);
+ ret = 0;
+
+ return ret;
+
+error_cleanup:
+ snic_free_vnic_res(snic);
+
+ return ret;
+}
+
+void
+snic_log_q_error(struct snic *snic)
+{
+ unsigned int i;
+ u32 err_status;
+
+ for (i = 0; i < snic->wq_count; i++) {
+ err_status = ioread32(&snic->wq[i].ctrl->error_status);
+ if (err_status)
+ SNIC_HOST_ERR(snic->shost,
+ "WQ[%d] error status %d\n",
+ i,
+ err_status);
+ }
+} /* end of snic_log_q_error */
diff --git a/drivers/scsi/snic/snic_res.h b/drivers/scsi/snic/snic_res.h
new file mode 100644
index 0000000..f8f4e29
--- /dev/null
+++ b/drivers/scsi/snic/snic_res.h
@@ -0,0 +1,96 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef __SNIC_RES_H
+#define __SNIC_RES_H
+
+#include "snic_io.h"
+#include "wq_enet_desc.h"
+#include "vnic_wq.h"
+#include "snic_fwint.h"
+#include "vnic_cq_fw.h"
+
+static inline void
+snic_icmnd_init(struct snic_host_req *req, u32 cmnd_id, u32 host_id, u64 ctx,
+ u16 flags, u64 tgt_id, u8 *lun, u8 *scsi_cdb, u8 cdb_len,
+ u32 data_len, u16 sg_cnt, u64 sgl_addr, u64 sns_addr_pa,
+ u32 sense_len)
+{
+ snic_io_hdr_enc(&req->hdr, SNIC_REQ_ICMND, 0, cmnd_id, host_id, sg_cnt,
+ ctx);
+
+ req->u.icmnd.flags = flags;
+ req->u.icmnd.tgt_id = tgt_id;
+ memcpy(&req->u.icmnd.lun_id, lun, LUN_ADDR_LEN);
+ req->u.icmnd.cdb_len = cdb_len;
+ memset(req->u.icmnd.cdb, 0, SNIC_CDB_LEN);
+ memcpy(req->u.icmnd.cdb, scsi_cdb, cdb_len);
+ req->u.icmnd.data_len = data_len;
+ req->u.icmnd.sg_addr = sgl_addr;
+ req->u.icmnd.sense_len = sense_len;
+ req->u.icmnd.sense_addr = sns_addr_pa;
+}
+
+static inline void
+snic_itmf_init(struct snic_host_req *req, u32 cmnd_id, u32 host_id, u64 ctx,
+ u16 flags, u32 req_id, u64 tgt_id, u8 *lun, u8 tm_type)
+{
+ snic_io_hdr_enc(&req->hdr, SNIC_REQ_ITMF, 0, cmnd_id, host_id, 0, ctx);
+
+ req->u.itmf.tm_type = tm_type;
+ req->u.itmf.flags = flags;
+ req->u.itmf.req_id = req_id; /* valid only in abort, clear task */
+ req->u.itmf.tgt_id = tgt_id;
+ memcpy(&req->u.itmf.lun_id, lun, LUN_ADDR_LEN);
+}
+
+static inline void
+snic_queue_wq_eth_desc(struct vnic_wq *wq,
+ void *os_buf,
+ dma_addr_t dma_addr,
+ unsigned int len,
+ int vlan_tag_insert,
+ unsigned int vlan_tag,
+ int cq_entry)
+{
+ struct wq_enet_desc *desc = vnic_wq_next_desc(wq);
+
+ wq_enet_desc_enc(desc,
+ (u64)dma_addr | VNIC_PADDR_TARGET,
+ (u16)len,
+ 0, /* mss_or_csum_offset */
+ 0, /* fc_eof */
+ 0, /* offload mode */
+ 1, /* eop */
+ (u8)cq_entry,
+ 0, /* fcoe_encap */
+ (u8)vlan_tag_insert,
+ (u16)vlan_tag,
+ 0 /* loopback */);
+
+ vnic_wq_post(wq, os_buf, dma_addr, len, 1, 1);
+}
+
+struct snic;
+
+int snic_get_vnic_config(struct snic *);
+int snic_alloc_vnic_res(struct snic *);
+void snic_free_vnic_res(struct snic *);
+void snic_get_res_counts(struct snic *);
+void snic_log_q_error(struct snic *);
+int snic_get_vnic_resources_size(struct snic *);
+#endif /* __SNIC_RES_H */
--
1.8.5.4
^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH v4 3/9] snic:Add meta request, handling of meta requests.
2015-04-09 11:49 [PATCH v4 0/9] snic:initial submission of snic driver for Cisco SCSI HBA Narsimhulu Musini
2015-04-09 11:49 ` [PATCH v4 1/9] snic: snic module infrastructure Narsimhulu Musini
2015-04-09 11:49 ` [PATCH v4 2/9] snic:Add interrupt, resource firmware interfaces Narsimhulu Musini
@ 2015-04-09 11:49 ` Narsimhulu Musini
2015-04-09 12:50 ` Hannes Reinecke
2015-04-09 11:49 ` [PATCH v4 4/9] snic:Add snic target discovery Narsimhulu Musini
` (5 subsequent siblings)
8 siblings, 1 reply; 24+ messages in thread
From: Narsimhulu Musini @ 2015-04-09 11:49 UTC (permalink / raw)
To: JBottomley, linux-scsi, hare, hch; +Cc: Narsimhulu Musini, Sesidhar Baddela
snic_io.h contains meta request structure definition
meta request contains high level information about firmware requests.
such as request information, size, SGLs.
snic_io.c contains interfaces to handle meta request, firmware acknowledgment,
and high level generic queueing interface.
Signed-off-by: Narsimhulu Musini <nmusini@cisco.com>
Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
---
* v3
- Removed request alignment functionality.
drivers/scsi/snic/snic_io.c | 519 ++++++++++++++++++++++++++++++++++++++++++++
drivers/scsi/snic/snic_io.h | 118 ++++++++++
2 files changed, 637 insertions(+)
create mode 100644 drivers/scsi/snic/snic_io.c
create mode 100644 drivers/scsi/snic/snic_io.h
diff --git a/drivers/scsi/snic/snic_io.c b/drivers/scsi/snic/snic_io.c
new file mode 100644
index 0000000..75235ea
--- /dev/null
+++ b/drivers/scsi/snic/snic_io.c
@@ -0,0 +1,519 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/errno.h>
+#include <linux/pci.h>
+#include <linux/slab.h>
+
+#include <linux/interrupt.h>
+#include <linux/workqueue.h>
+#include <linux/spinlock.h>
+#include <linux/mempool.h>
+#include <scsi/scsi_tcq.h>
+
+#include "snic_io.h"
+#include "snic.h"
+#include "cq_enet_desc.h"
+#include "snic_fwint.h"
+
+static void
+snic_wq_cmpl_frame_send(struct vnic_wq *wq,
+ struct cq_desc *cq_desc,
+ struct vnic_wq_buf *buf,
+ void *opaque)
+{
+ struct snic *snic = vnic_dev_priv(wq->vdev);
+
+ SNIC_BUG_ON(buf->os_buf == NULL);
+
+ if (snic_log_level & SNIC_DESC_LOGGING)
+ SNIC_HOST_INFO(snic->shost,
+ "Ack received for snic_host_req %p.\n",
+ buf->os_buf);
+
+ SNIC_TRC(snic->shost->host_no, 0, 0,
+ ((u64)(buf->os_buf) - sizeof(struct snic_req_info)), 0, 0, 0);
+ pci_unmap_single(snic->pdev, buf->dma_addr, buf->len, PCI_DMA_TODEVICE);
+ buf->os_buf = NULL;
+}
+
+static int
+snic_wq_cmpl_handler_cont(struct vnic_dev *vdev,
+ struct cq_desc *cq_desc,
+ u8 type,
+ u16 q_num,
+ u16 cmpl_idx,
+ void *opaque)
+{
+ struct snic *snic = vnic_dev_priv(vdev);
+ unsigned long flags;
+
+ SNIC_BUG_ON(q_num != 0);
+
+ spin_lock_irqsave(&snic->wq_lock[q_num], flags);
+ vnic_wq_service(&snic->wq[q_num],
+ cq_desc,
+ cmpl_idx,
+ snic_wq_cmpl_frame_send,
+ NULL);
+ spin_unlock_irqrestore(&snic->wq_lock[q_num], flags);
+
+ return 0;
+} /* end of snic_cmpl_handler_cont */
+
+int
+snic_wq_cmpl_handler(struct snic *snic, int work_to_do)
+{
+ unsigned int work_done = 0;
+ unsigned int i;
+
+ snic->s_stats.misc.last_ack_time = jiffies;
+ for (i = 0; i < snic->wq_count; i++) {
+ work_done += vnic_cq_service(&snic->cq[i],
+ work_to_do,
+ snic_wq_cmpl_handler_cont,
+ NULL);
+ }
+
+ return work_done;
+} /* end of snic_wq_cmpl_handler */
+
+void
+snic_free_wq_buf(struct vnic_wq *wq, struct vnic_wq_buf *buf)
+{
+
+ struct snic_host_req *req = buf->os_buf;
+ struct snic *snic = vnic_dev_priv(wq->vdev);
+ struct snic_req_info *rqi = NULL;
+ unsigned long flags;
+
+ pci_unmap_single(snic->pdev, buf->dma_addr, buf->len, PCI_DMA_TODEVICE);
+
+ rqi = req_to_rqi(req);
+ spin_lock_irqsave(&snic->spl_cmd_lock, flags);
+ if (list_empty(&rqi->list)) {
+ spin_unlock_irqrestore(&snic->spl_cmd_lock, flags);
+ goto end;
+ }
+
+ SNIC_BUG_ON(rqi->list.next == NULL); /* if not added to spl_cmd_list */
+ list_del_init(&rqi->list);
+ spin_unlock_irqrestore(&snic->spl_cmd_lock, flags);
+
+ if (rqi->sge_va) {
+ snic_pci_unmap_rsp_buf(snic, rqi);
+ kfree((void *)rqi->sge_va);
+ rqi->sge_va = 0;
+ }
+ snic_req_free(snic, rqi);
+ SNIC_HOST_INFO(snic->shost, "snic_free_wq_buf .. freed.\n");
+
+end:
+ return;
+}
+
+/* Criteria to select work queue in multi queue mode */
+static int
+snic_select_wq(struct snic *snic)
+{
+ /* No multi queue support for now */
+ BUILD_BUG_ON(SNIC_WQ_MAX > 1);
+
+ return 0;
+}
+
+int
+snic_queue_wq_desc(struct snic *snic, void *os_buf, u16 len)
+{
+ dma_addr_t pa = 0;
+ unsigned long flags;
+ struct snic_fw_stats *fwstats = &snic->s_stats.fw;
+ long act_reqs;
+ int q_num = 0;
+
+ snic_print_desc(__func__, os_buf, len);
+
+ /* Map request buffer */
+ pa = pci_map_single(snic->pdev, os_buf, len, PCI_DMA_TODEVICE);
+ if (pci_dma_mapping_error(snic->pdev, pa)) {
+ SNIC_HOST_ERR(snic->shost, "qdesc: PCI DMA Mapping Fail.\n");
+
+ return -ENOMEM;
+ }
+
+ q_num = snic_select_wq(snic);
+
+ spin_lock_irqsave(&snic->wq_lock[q_num], flags);
+ if (!vnic_wq_desc_avail(snic->wq)) {
+ pci_unmap_single(snic->pdev, pa, len, PCI_DMA_TODEVICE);
+ spin_unlock_irqrestore(&snic->wq_lock[q_num], flags);
+ atomic64_inc(&snic->s_stats.misc.wq_alloc_fail);
+ SNIC_DBG("host = %d, WQ is Full\n", snic->shost->host_no);
+
+ return -ENOMEM;
+ }
+
+ snic_queue_wq_eth_desc(&snic->wq[q_num], os_buf, pa, len, 0, 0, 1);
+ spin_unlock_irqrestore(&snic->wq_lock[q_num], flags);
+
+ /* Update stats */
+ act_reqs = atomic64_inc_return(&fwstats->actv_reqs);
+ if (act_reqs > atomic64_read(&fwstats->max_actv_reqs))
+ atomic64_set(&fwstats->max_actv_reqs, act_reqs);
+
+ return 0;
+} /* end of snic_queue_wq_desc() */
+
+/*
+ * snic_handle_untagged_req: Adds snic specific requests to spl_cmd_list.
+ * Purpose : Used during driver unload to clean up the requests.
+ */
+void
+snic_handle_untagged_req(struct snic *snic, struct snic_req_info *rqi)
+{
+ unsigned long flags;
+
+ INIT_LIST_HEAD(&rqi->list);
+
+ spin_lock_irqsave(&snic->spl_cmd_lock, flags);
+ list_add_tail(&rqi->list, &snic->spl_cmd_list);
+ spin_unlock_irqrestore(&snic->spl_cmd_lock, flags);
+}
+
+/*
+ * snic_req_init:
+ * Allocates snic_req_info + snic_host_req + sgl data, and initializes.
+ */
+struct snic_req_info *
+snic_req_init(struct snic *snic, int sg_cnt)
+{
+ u8 typ;
+ struct snic_req_info *rqi = NULL;
+
+ typ = (sg_cnt <= SNIC_REQ_CACHE_DFLT_SGL) ?
+ SNIC_REQ_CACHE_DFLT_SGL : SNIC_REQ_CACHE_MAX_SGL;
+
+ rqi = mempool_alloc(snic->req_pool[typ], GFP_ATOMIC);
+ if (!rqi) {
+ atomic64_inc(&snic->s_stats.io.alloc_fail);
+ SNIC_HOST_ERR(snic->shost,
+ "Failed to allocate memory from snic req pool id = %d\n",
+ typ);
+ return rqi;
+ }
+
+ memset(rqi, 0, sizeof(*rqi));
+ rqi->rq_pool_type = typ;
+ rqi->start_time = jiffies;
+ rqi->req = (struct snic_host_req *) (rqi + 1);
+ rqi->req_len = sizeof(struct snic_host_req);
+ rqi->snic = snic;
+
+ BUILD_BUG_ON((sizeof(struct snic_req_info) % SNIC_SG_DESC_ALIGN) != 0);
+
+ rqi->req = (struct snic_host_req *)(rqi + 1);
+
+ if (sg_cnt == 0)
+ goto end;
+
+ rqi->req_len += (sg_cnt * sizeof(struct snic_sg_desc));
+
+ if (sg_cnt > atomic64_read(&snic->s_stats.io.max_sgl))
+ atomic64_set(&snic->s_stats.io.max_sgl, sg_cnt);
+
+ SNIC_BUG_ON(sg_cnt > SNIC_MAX_SG_DESC_CNT);
+ atomic64_inc(&snic->s_stats.io.sgl_cnt[sg_cnt - 1]);
+
+end:
+ memset(rqi->req, 0, rqi->req_len);
+
+ /* pre initialization of init_ctx to support req_to_rqi */
+ rqi->req->hdr.init_ctx = (u64) rqi;
+
+ SNIC_SCSI_DBG(snic->shost, "Req_alloc:rqi = %p allocatd.\n", rqi);
+
+ return rqi;
+} /* end of snic_req_init */
+
+/*
+ * snic_abort_req_init : Inits abort request.
+ */
+struct snic_host_req *
+snic_abort_req_init(struct snic *snic, struct snic_req_info *rqi)
+{
+ struct snic_host_req *req = NULL;
+
+ SNIC_BUG_ON(!rqi);
+
+ /* If abort to be issued second time, then reuse */
+ if (rqi->abort_req)
+ return rqi->abort_req;
+
+
+ req = mempool_alloc(snic->req_pool[SNIC_REQ_TM_CACHE], GFP_ATOMIC);
+ if (!req) {
+ SNIC_HOST_ERR(snic->shost, "abts:Failed to alloc tm req.\n");
+ WARN_ON_ONCE(1);
+
+ return NULL;
+ }
+
+ rqi->abort_req = req;
+ memset(req, 0, sizeof(struct snic_host_req));
+ /* pre initialization of init_ctx to support req_to_rqi */
+ req->hdr.init_ctx = (u64) rqi;
+
+ return req;
+} /* end of snic_abort_req_init */
+
+/*
+ * snic_dr_req_init : Inits device reset req
+ */
+struct snic_host_req *
+snic_dr_req_init(struct snic *snic, struct snic_req_info *rqi)
+{
+ struct snic_host_req *req = NULL;
+
+ SNIC_BUG_ON(!rqi);
+
+ req = mempool_alloc(snic->req_pool[SNIC_REQ_TM_CACHE], GFP_ATOMIC);
+ if (!req) {
+ SNIC_HOST_ERR(snic->shost, "dr:Failed to alloc tm req.\n");
+ WARN_ON_ONCE(1);
+
+ return NULL;
+ }
+
+ SNIC_BUG_ON(rqi->dr_req != NULL);
+ rqi->dr_req = req;
+ memset(req, 0, sizeof(struct snic_host_req));
+ /* pre initialization of init_ctx to support req_to_rqi */
+ req->hdr.init_ctx = (u64) rqi;
+
+ return req;
+} /* end of snic_dr_req_init */
+
+/* frees snic_req_info and snic_host_req */
+void
+snic_req_free(struct snic *snic, struct snic_req_info *rqi)
+{
+ SNIC_BUG_ON(rqi->req == rqi->abort_req);
+ SNIC_BUG_ON(rqi->req == rqi->dr_req);
+ SNIC_BUG_ON(rqi->sge_va != 0);
+
+ SNIC_SCSI_DBG(snic->shost,
+ "Req_free:rqi %p:ioreq %p:abt %p:dr %p\n",
+ rqi, rqi->req, rqi->abort_req, rqi->dr_req);
+
+ if (rqi->abort_req)
+ mempool_free(rqi->abort_req, snic->req_pool[SNIC_REQ_TM_CACHE]);
+
+ if (rqi->dr_req)
+ mempool_free(rqi->dr_req, snic->req_pool[SNIC_REQ_TM_CACHE]);
+
+ mempool_free(rqi, snic->req_pool[rqi->rq_pool_type]);
+}
+
+void
+snic_pci_unmap_rsp_buf(struct snic *snic, struct snic_req_info *rqi)
+{
+ struct snic_sg_desc *sgd;
+
+ sgd = req_to_sgl(rqi_to_req(rqi));
+ SNIC_BUG_ON(sgd[0].addr == 0);
+ pci_unmap_single(snic->pdev,
+ sgd[0].addr,
+ sgd[0].len,
+ PCI_DMA_FROMDEVICE);
+}
+
+/*
+ * snic_free_all_untagged_reqs: Walks through untagged reqs and frees them.
+ */
+void
+snic_free_all_untagged_reqs(struct snic *snic)
+{
+ struct snic_req_info *rqi;
+ struct list_head *cur, *nxt;
+ unsigned long flags;
+
+ spin_lock_irqsave(&snic->spl_cmd_lock, flags);
+ list_for_each_safe(cur, nxt, &snic->spl_cmd_list) {
+ rqi = list_entry(cur, struct snic_req_info, list);
+ list_del_init(&rqi->list);
+ if (rqi->sge_va) {
+ snic_pci_unmap_rsp_buf(snic, rqi);
+ kfree((void *)rqi->sge_va);
+ rqi->sge_va = 0;
+ }
+
+ snic_req_free(snic, rqi);
+ }
+ spin_unlock_irqrestore(&snic->spl_cmd_lock, flags);
+}
+
+/*
+ * snic_release_untagged_req : Unlinks the untagged req and frees it.
+ */
+void
+snic_release_untagged_req(struct snic *snic, struct snic_req_info *rqi)
+{
+ unsigned long flags;
+
+ spin_lock_irqsave(&snic->snic_lock, flags);
+ if (snic->in_remove) {
+ spin_unlock_irqrestore(&snic->snic_lock, flags);
+ goto end;
+ }
+ spin_unlock_irqrestore(&snic->snic_lock, flags);
+
+ spin_lock_irqsave(&snic->spl_cmd_lock, flags);
+ if (list_empty(&rqi->list)) {
+ spin_unlock_irqrestore(&snic->spl_cmd_lock, flags);
+ goto end;
+ }
+ list_del_init(&rqi->list);
+ spin_unlock_irqrestore(&snic->spl_cmd_lock, flags);
+ snic_req_free(snic, rqi);
+
+end:
+ return;
+}
+
+/* dump buf in hex fmt */
+void
+snic_hex_dump(char *pfx, char *data, int len)
+{
+ SNIC_INFO("%s Dumping Data of Len = %d\n", pfx, len);
+ print_hex_dump_bytes(pfx, DUMP_PREFIX_NONE, data, len);
+}
+
+#define LINE_BUFSZ 128 /* for snic_print_desc fn */
+static void
+snic_dump_desc(const char *fn, char *os_buf, int len)
+{
+ struct snic_host_req *req = (struct snic_host_req *) os_buf;
+ struct snic_fw_req *fwreq = (struct snic_fw_req *) os_buf;
+ struct snic_req_info *rqi = NULL;
+ char line[LINE_BUFSZ] = { '\0' };
+ char *cmd_str = NULL;
+
+ if (req->hdr.type >= SNIC_RSP_REPORT_TGTS_CMPL)
+ rqi = (struct snic_req_info *) fwreq->hdr.init_ctx;
+ else
+ rqi = (struct snic_req_info *) req->hdr.init_ctx;
+
+ SNIC_BUG_ON(rqi == NULL || rqi->req == NULL);
+ switch (req->hdr.type) {
+ case SNIC_REQ_REPORT_TGTS:
+ cmd_str = "report-tgt : ";
+ snprintf(line, LINE_BUFSZ, "SNIC_REQ_REPORT_TGTS :");
+ break;
+
+ case SNIC_REQ_ICMND:
+ cmd_str = "icmnd : ";
+ snprintf(line, LINE_BUFSZ, "SNIC_REQ_ICMND : 0x%x :",
+ req->u.icmnd.cdb[0]);
+ break;
+
+ case SNIC_REQ_ITMF:
+ cmd_str = "itmf : ";
+ snprintf(line, LINE_BUFSZ, "SNIC_REQ_ITMF :");
+ break;
+
+ case SNIC_REQ_HBA_RESET:
+ cmd_str = "hba reset :";
+ snprintf(line, LINE_BUFSZ, "SNIC_REQ_HBA_RESET :");
+ break;
+
+ case SNIC_REQ_EXCH_VER:
+ cmd_str = "exch ver : ";
+ snprintf(line, LINE_BUFSZ, "SNIC_REQ_EXCH_VER :");
+ break;
+
+ case SNIC_REQ_TGT_INFO:
+ cmd_str = "tgt info : ";
+ break;
+
+ case SNIC_RSP_REPORT_TGTS_CMPL:
+ cmd_str = "report tgt cmpl : ";
+ snprintf(line, LINE_BUFSZ, "SNIC_RSP_REPORT_TGTS_CMPL :");
+ break;
+
+ case SNIC_RSP_ICMND_CMPL:
+ cmd_str = "icmnd_cmpl : ";
+ snprintf(line, LINE_BUFSZ, "SNIC_RSP_ICMND_CMPL : 0x%x :",
+ rqi->req->u.icmnd.cdb[0]);
+ break;
+
+ case SNIC_RSP_ITMF_CMPL:
+ cmd_str = "itmf_cmpl : ";
+ snprintf(line, LINE_BUFSZ, "SNIC_RSP_ITMF_CMPL :");
+ break;
+
+ case SNIC_RSP_HBA_RESET_CMPL:
+ cmd_str = "hba_reset_cmpl : ";
+ snprintf(line, LINE_BUFSZ, "SNIC_RSP_HBA_RESET_CMPL :");
+ break;
+
+ case SNIC_RSP_EXCH_VER_CMPL:
+ cmd_str = "exch_ver_cmpl : ";
+ snprintf(line, LINE_BUFSZ, "SNIC_RSP_EXCH_VER_CMPL :");
+ break;
+
+ case SNIC_MSG_ACK:
+ cmd_str = "msg ack : ";
+ snprintf(line, LINE_BUFSZ, "SNIC_MSG_ACK :");
+ break;
+
+ case SNIC_MSG_ASYNC_EVNOTIFY:
+ cmd_str = "async notify : ";
+ snprintf(line, LINE_BUFSZ, "SNIC_MSG_ASYNC_EVNOTIFY :");
+ break;
+
+ default:
+ cmd_str = "unknown : ";
+ SNIC_BUG_ON(1);
+ break;
+ }
+
+ SNIC_INFO("%s:%s >>cmndid=%x:sg_cnt = %x:status = %x:ctx = %llx.\n",
+ fn, line, req->hdr.cmnd_id, req->hdr.sg_cnt, req->hdr.status,
+ req->hdr.init_ctx);
+
+ /* Enable it, to dump byte stream */
+ if (snic_log_level & 0x20)
+ snic_hex_dump(cmd_str, os_buf, len);
+} /* end of __snic_print_desc */
+
+void
+snic_print_desc(const char *fn, char *os_buf, int len)
+{
+ if (snic_log_level & SNIC_DESC_LOGGING)
+ snic_dump_desc(fn, os_buf, len);
+}
+
+void
+snic_calc_io_process_time(struct snic *snic, struct snic_req_info *rqi)
+{
+ u64 duration;
+
+ duration = jiffies - rqi->start_time;
+
+ if (duration > atomic64_read(&snic->s_stats.io.max_time))
+ atomic64_set(&snic->s_stats.io.max_time, duration);
+}
diff --git a/drivers/scsi/snic/snic_io.h b/drivers/scsi/snic/snic_io.h
new file mode 100644
index 0000000..eca16e4
--- /dev/null
+++ b/drivers/scsi/snic/snic_io.h
@@ -0,0 +1,118 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef _SNIC_IO_H
+#define _SNIC_IO_H
+
+#define SNIC_DFLT_SG_DESC_CNT 32 /* Default descriptors for sgl */
+#define SNIC_MAX_SG_DESC_CNT 60 /* Max descriptor for sgl */
+#define SNIC_SG_DESC_ALIGN 16 /* Descriptor address alignment */
+
+/* SG descriptor for snic */
+struct snic_sg_desc {
+ __le64 addr;
+ __le32 len;
+ u32 _resvd;
+};
+
+struct snic_dflt_sgl {
+ struct snic_sg_desc sg_desc[SNIC_DFLT_SG_DESC_CNT];
+};
+
+struct snic_max_sgl {
+ struct snic_sg_desc sg_desc[SNIC_MAX_SG_DESC_CNT];
+};
+
+enum snic_req_cache_type {
+ SNIC_REQ_CACHE_DFLT_SGL = 0, /* cache with default size sgl */
+ SNIC_REQ_CACHE_MAX_SGL, /* cache with max size sgl */
+ SNIC_REQ_TM_CACHE, /* cache for task mgmt reqs contains
+ snic_host_req objects only*/
+ SNIC_REQ_MAX_CACHES /* number of sgl caches */
+};
+
+/* Per IO internal state */
+struct snic_internal_io_state {
+ char *rqi;
+ u64 flags;
+ u32 state;
+ u32 abts_status; /* Abort completion status */
+ u32 lr_status; /* device reset completion status */
+};
+
+/* IO state machine */
+enum snic_ioreq_state {
+ SNIC_IOREQ_NOT_INITED = 0,
+ SNIC_IOREQ_PENDING,
+ SNIC_IOREQ_ABTS_PENDING,
+ SNIC_IOREQ_ABTS_COMPLETE,
+ SNIC_IOREQ_LR_PENDING,
+ SNIC_IOREQ_LR_COMPLETE,
+ SNIC_IOREQ_COMPLETE,
+};
+
+struct snic;
+struct snic_host_req;
+
+/*
+ * snic_req_info : Contains info about IO, one per scsi command.
+ * Notes: Make sure that the structure is aligned to 16 B
+ * this helps in easy access to snic_req_info from snic_host_req
+ */
+struct snic_req_info {
+ struct list_head list;
+ struct snic_host_req *req;
+ u64 start_time; /* start time in jiffies */
+ u16 rq_pool_type; /* noticion of request pool type */
+ u16 req_len; /* buf len passing to fw (req + sgl)*/
+ u32 tgt_id;
+
+ u32 tm_tag;
+ u8 io_cmpl:1; /* sets to 1 when fw completes IO */
+ u8 _resvd1[3];
+ struct scsi_cmnd *sc; /* Associated scsi cmd */
+ struct snic *snic; /* Associated snic */
+ u64 sge_va; /* Pointer to Resp Buffer */
+ u64 snsbuf_va;
+
+ struct snic_host_req *abort_req;
+ struct completion *abts_done;
+
+ struct snic_host_req *dr_req;
+ struct completion *dr_done;
+};
+
+
+#define rqi_to_req(rqi) \
+ ((struct snic_host_req *) (((struct snic_req_info *)rqi)->req))
+
+#define req_to_rqi(req) \
+ ((struct snic_req_info *) (((struct snic_host_req *)req)->hdr.init_ctx))
+
+#define req_to_sgl(req) \
+ ((struct snic_sg_desc *) (((struct snic_host_req *)req)+1))
+
+struct snic_req_info *
+snic_req_init(struct snic *, int sg_cnt);
+void snic_req_free(struct snic *, struct snic_req_info *);
+void snic_calc_io_process_time(struct snic *, struct snic_req_info *);
+void snic_pci_unmap_rsp_buf(struct snic *, struct snic_req_info *);
+struct snic_host_req *
+snic_abort_req_init(struct snic *, struct snic_req_info *);
+struct snic_host_req *
+snic_dr_req_init(struct snic *, struct snic_req_info *);
+#endif /* _SNIC_IO_H */
--
1.8.5.4
^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH v4 4/9] snic:Add snic target discovery
2015-04-09 11:49 [PATCH v4 0/9] snic:initial submission of snic driver for Cisco SCSI HBA Narsimhulu Musini
` (2 preceding siblings ...)
2015-04-09 11:49 ` [PATCH v4 3/9] snic:Add meta request, handling of meta requests Narsimhulu Musini
@ 2015-04-09 11:49 ` Narsimhulu Musini
2015-04-09 12:59 ` Hannes Reinecke
2015-04-09 11:49 ` [PATCH v4 5/9] snic:add SCSI handling, AEN, and fwreset handling Narsimhulu Musini
` (4 subsequent siblings)
8 siblings, 1 reply; 24+ messages in thread
From: Narsimhulu Musini @ 2015-04-09 11:49 UTC (permalink / raw)
To: JBottomley, linux-scsi, hare, hch; +Cc: Narsimhulu Musini, Sesidhar Baddela
snic_disc.h contains snic target structure definition.
snic_disc.c contains target discovery, setup, lookup, and cleanup
snic_ctl.c contains retrieval of snic capabilities includes
max ios, size, SGs per request, and max concurrent requests.
Signed-off-by: Narsimhulu Musini <nmusini@cisco.com>
Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
---
* v3
- Cleaned up redundant comment.
drivers/scsi/snic/snic_ctl.c | 276 +++++++++++++++++++
drivers/scsi/snic/snic_disc.c | 602 ++++++++++++++++++++++++++++++++++++++++++
drivers/scsi/snic/snic_disc.h | 124 +++++++++
3 files changed, 1002 insertions(+)
create mode 100644 drivers/scsi/snic/snic_ctl.c
create mode 100644 drivers/scsi/snic/snic_disc.c
create mode 100644 drivers/scsi/snic/snic_disc.h
diff --git a/drivers/scsi/snic/snic_ctl.c b/drivers/scsi/snic/snic_ctl.c
new file mode 100644
index 0000000..4bc55be
--- /dev/null
+++ b/drivers/scsi/snic/snic_ctl.c
@@ -0,0 +1,276 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/errno.h>
+#include <linux/pci.h>
+#include <linux/slab.h>
+
+#include <linux/interrupt.h>
+#include <linux/workqueue.h>
+#include <linux/spinlock.h>
+#include <linux/mempool.h>
+#include <scsi/scsi_tcq.h>
+#include <linux/ctype.h>
+
+#include "snic_io.h"
+#include "snic.h"
+#include "cq_enet_desc.h"
+#include "snic_fwint.h"
+
+/*
+ * snic_handle_link : Handles link flaps.
+ */
+void
+snic_handle_link(struct work_struct *work)
+{
+ struct snic *snic = container_of(work, struct snic, link_work);
+
+ if (snic->config.xpt_type != SNIC_DAS) {
+ SNIC_HOST_INFO(snic->shost, "Link Event Received.\n");
+ SNIC_ASSERT_NOT_IMPL(1);
+
+ return;
+ }
+
+ snic->link_status = vnic_dev_link_status(snic->vdev);
+ snic->link_down_cnt = vnic_dev_link_down_cnt(snic->vdev);
+ SNIC_HOST_INFO(snic->shost, "Link Event: Link %s.\n",
+ ((snic->link_status) ? "Up" : "Down"));
+}
+
+
+/*
+ * snic_ver_enc : Encodes version str to int
+ * version string is similar to netmask string
+ */
+static int
+snic_ver_enc(const char *s)
+{
+ int v[4] = {0};
+ int i = 0, x = 0;
+ char c;
+ const char *p = s;
+
+ /* validate version string */
+ if ((strlen(s) > 15) || (strlen(s) < 7))
+ goto end;
+
+ while ((c = *p++)) {
+ if (c == '.') {
+ i++;
+ continue;
+ }
+
+ if (i > 4 || !isdigit(c))
+ goto end;
+
+ v[i] = v[i] * 10 + (c - '0');
+ }
+
+ /* validate sub version numbers */
+ for (i = 3; i >= 0; i--)
+ if (v[i] > 0xff)
+ goto end;
+
+ x |= (v[0] << 24) | v[1] << 16 | v[2] << 8 | v[3];
+
+end:
+ if (x == 0) {
+ SNIC_ERR("Invalid version string [%s].\n", s);
+
+ return -1;
+ }
+
+ return x;
+} /* end of snic_ver_enc */
+
+/*
+ * snic_qeueue_exch_ver_req :
+ *
+ * Queues Exchange Version Request, to communicate host information
+ * in return, it gets firmware version details
+ */
+int
+snic_queue_exch_ver_req(struct snic *snic)
+{
+ struct snic_req_info *rqi = NULL;
+ struct snic_host_req *req = NULL;
+ int ret = 0;
+
+ SNIC_HOST_INFO(snic->shost, "Exch Ver Req Preparing...\n");
+
+ rqi = snic_req_init(snic, 0);
+ if (!rqi) {
+ SNIC_HOST_ERR(snic->shost,
+ "Queuing Exch Ver Req failed, err = %d\n",
+ ret);
+
+ ret = -ENOMEM;
+ goto error;
+ }
+
+ req = rqi_to_req(rqi);
+
+ /* Initialize snic_host_req */
+ snic_io_hdr_enc(&req->hdr, SNIC_REQ_EXCH_VER, 0, SCSI_NO_TAG,
+ snic->config.hid, 0, (u64)rqi);
+ req->u.exch_ver.drvr_ver = snic_ver_enc(SNIC_DRV_VERSION);
+ req->u.exch_ver.os_type = SNIC_OS_LINUX;
+
+ snic_handle_untagged_req(snic, rqi);
+
+ ret = snic_queue_wq_desc(snic, req, sizeof(*req));
+ if (ret) {
+ snic_release_untagged_req(snic, rqi);
+ SNIC_HOST_ERR(snic->shost,
+ "Queuing Exch Ver Req failed, err = %d\n",
+ ret);
+ goto error;
+ }
+
+ SNIC_HOST_INFO(snic->shost, "Exch Ver Req is issued. ret = %d\n", ret);
+
+error:
+ return ret;
+} /* end of snic_queue_exch_ver_req */
+
+/*
+ * snic_io_exch_ver_cmpl_handler
+ */
+int
+snic_io_exch_ver_cmpl_handler(struct snic *snic, struct snic_fw_req *fwreq)
+{
+ struct snic_req_info *rqi = NULL;
+ u8 typ, hdr_stat;
+ u32 cmnd_id, hid, max_sgs;
+ u64 ctx = 0;
+ unsigned long flags;
+ int ret = 0;
+
+ SNIC_HOST_INFO(snic->shost, "Exch Ver Compl Received.\n");
+ snic_io_hdr_dec(&fwreq->hdr, &typ, &hdr_stat, &cmnd_id, &hid, &ctx);
+ SNIC_BUG_ON(snic->config.hid != hid);
+ rqi = (struct snic_req_info *) ctx;
+
+ if (hdr_stat) {
+ SNIC_HOST_ERR(snic->shost,
+ "Exch Ver Completed w/ err status %d\n",
+ hdr_stat);
+
+ goto exch_cmpl_end;
+ }
+
+ spin_lock_irqsave(&snic->snic_lock, flags);
+ snic->fwinfo.fw_ver = fwreq->u.exch_ver_cmpl.version;
+ snic->fwinfo.hid = fwreq->u.exch_ver_cmpl.hid;
+ snic->fwinfo.max_concur_ios = fwreq->u.exch_ver_cmpl.max_concur_ios;
+ snic->fwinfo.max_sgs_per_cmd = fwreq->u.exch_ver_cmpl.max_sgs_per_cmd;
+ snic->fwinfo.max_io_sz = fwreq->u.exch_ver_cmpl.max_io_sz;
+ snic->fwinfo.max_tgts = fwreq->u.exch_ver_cmpl.max_tgts;
+ snic->fwinfo.io_tmo = fwreq->u.exch_ver_cmpl.io_timeout;
+
+ SNIC_HOST_INFO(snic->shost,
+ "vers %u hid %u max_concur_ios %u max_sgs_per_cmd %u max_io_sz %u max_tgts %u fw tmo %u\n",
+ fwreq->u.exch_ver_cmpl.version,
+ fwreq->u.exch_ver_cmpl.hid,
+ fwreq->u.exch_ver_cmpl.max_concur_ios,
+ fwreq->u.exch_ver_cmpl.max_sgs_per_cmd,
+ fwreq->u.exch_ver_cmpl.max_io_sz,
+ fwreq->u.exch_ver_cmpl.max_tgts,
+ fwreq->u.exch_ver_cmpl.io_timeout);
+
+ SNIC_HOST_INFO(snic->shost,
+ "HBA Capabilities = 0x%x\n",
+ fwreq->u.exch_ver_cmpl.hba_cap);
+
+ /* Updating SGList size */
+ max_sgs = snic->fwinfo.max_sgs_per_cmd;
+ if (max_sgs && max_sgs < SNIC_MAX_SG_DESC_CNT) {
+ snic->shost->sg_tablesize = max_sgs;
+ SNIC_HOST_INFO(snic->shost, "Max SGs set to %d\n",
+ snic->shost->sg_tablesize);
+ } else if (max_sgs > snic->shost->sg_tablesize) {
+ SNIC_HOST_INFO(snic->shost,
+ "Target type %d Supports Larger Max SGList %d than driver's Max SG List %d.\n",
+ snic->config.xpt_type, max_sgs,
+ snic->shost->sg_tablesize);
+ }
+
+ if (snic->shost->can_queue > snic->fwinfo.max_concur_ios)
+ snic->shost->can_queue = snic->fwinfo.max_concur_ios;
+
+ snic->shost->max_sectors = snic->fwinfo.max_io_sz >> 9;
+ if (snic->fwinfo.wait)
+ complete(snic->fwinfo.wait);
+
+ spin_unlock_irqrestore(&snic->snic_lock, flags);
+
+exch_cmpl_end:
+ snic_release_untagged_req(snic, rqi);
+
+ SNIC_HOST_INFO(snic->shost, "Exch_cmpl Done, hdr_stat %d.\n", hdr_stat);
+
+ return ret;
+} /* end of snic_io_exch_ver_cmpl_handler */
+
+/*
+ * snic_get_conf
+ *
+ * Synchronous call, and Retrieves snic params.
+ */
+int
+snic_get_conf(struct snic *snic)
+{
+ DECLARE_COMPLETION_ONSTACK(wait);
+ unsigned long flags;
+ int ret;
+ int nr_retries = 3;
+
+ SNIC_HOST_INFO(snic->shost, "Retrieving snic params.\n");
+ spin_lock_irqsave(&snic->snic_lock, flags);
+ memset(&snic->fwinfo, 0, sizeof(snic->fwinfo));
+ snic->fwinfo.wait = &wait;
+ spin_unlock_irqrestore(&snic->snic_lock, flags);
+
+ /* Additional delay to handle HW Resource initialization. */
+ msleep(50);
+
+ /*
+ * Exch ver req can be ignored by FW, if HW Resource initialization
+ * is in progress, Hence retry.
+ */
+ do {
+ ret = snic_queue_exch_ver_req(snic);
+ if (ret)
+ return ret;
+
+ wait_for_completion_timeout(&wait, msecs_to_jiffies(2000));
+ spin_lock_irqsave(&snic->snic_lock, flags);
+ ret = (snic->fwinfo.fw_ver != 0) ? 0 : -ETIMEDOUT;
+ if (ret)
+ SNIC_HOST_ERR(snic->shost,
+ "Failed to retrieve snic params,\n");
+
+ /* Unset fwinfo.wait, on success or on last retry */
+ if (ret == 0 || nr_retries == 1)
+ snic->fwinfo.wait = NULL;
+
+ spin_unlock_irqrestore(&snic->snic_lock, flags);
+ } while (ret && --nr_retries);
+
+ return ret;
+} /* end of snic_get_info */
diff --git a/drivers/scsi/snic/snic_disc.c b/drivers/scsi/snic/snic_disc.c
new file mode 100644
index 0000000..b6b6cca
--- /dev/null
+++ b/drivers/scsi/snic/snic_disc.c
@@ -0,0 +1,602 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/errno.h>
+#include <linux/mempool.h>
+
+#include <scsi/scsi_tcq.h>
+
+#include "snic_disc.h"
+#include "snic.h"
+#include "snic_io.h"
+
+
+/* snic target types */
+static const char * const snic_tgt_type_str[] = {
+ [SNIC_TGT_DAS] = "DAS",
+ [SNIC_TGT_SAN] = "SAN",
+};
+
+static inline const char *
+snic_tgt_type_to_str(int typ)
+{
+ return ((typ > SNIC_TGT_NONE && typ <= SNIC_TGT_SAN) ?
+ snic_tgt_type_str[typ] : "Unknown");
+}
+
+static const char * const snic_tgt_state_str[] = {
+ [SNIC_TGT_STAT_INIT] = "INIT",
+ [SNIC_TGT_STAT_ONLINE] = "ONLINE",
+ [SNIC_TGT_STAT_OFFLINE] = "OFFLINE",
+ [SNIC_TGT_STAT_DEL] = "DELETION IN PROGRESS",
+};
+
+const char *
+snic_tgt_state_to_str(int state)
+{
+ return ((state >= SNIC_TGT_STAT_INIT && state <= SNIC_TGT_STAT_DEL) ?
+ snic_tgt_state_str[state] : "UNKNOWN");
+}
+
+/*
+ * Initiate report_tgt req desc
+ */
+static void
+snic_report_tgt_init(struct snic_host_req *req, u32 hid, u8 *buf, u32 len,
+ u64 rsp_buf_pa, u64 ctx)
+{
+ struct snic_sg_desc *sgd = NULL;
+
+
+ snic_io_hdr_enc(&req->hdr, SNIC_REQ_REPORT_TGTS, 0, SCSI_NO_TAG, hid,
+ 1, ctx);
+
+ req->u.rpt_tgts.sg_cnt = 1;
+ sgd = req_to_sgl(req);
+ sgd[0].addr = cpu_to_le64(rsp_buf_pa);
+ sgd[0].len = cpu_to_le32(len);
+ sgd[0]._resvd = 0;
+ req->u.rpt_tgts.sg_addr = (u64) sgd;
+}
+
+/*
+ * snic_queue_report_tgt_req: Queues report target request.
+ */
+static int
+snic_queue_report_tgt_req(struct snic *snic)
+{
+ struct snic_req_info *rqi = NULL;
+ u32 ntgts, buf_len = 0;
+ u8 *buf = NULL;
+ dma_addr_t pa = 0;
+ int ret = 0;
+
+ rqi = snic_req_init(snic, 1);
+ if (!rqi) {
+ ret = -ENOMEM;
+ goto error;
+ }
+
+ if (snic->fwinfo.max_tgts)
+ ntgts = min_t(u32, snic->fwinfo.max_tgts, snic->shost->max_id);
+ else
+ ntgts = snic->shost->max_id;
+
+ /* Allocate Response Buffer */
+ SNIC_BUG_ON(ntgts == 0);
+ buf_len = ntgts * sizeof(struct snic_tgt_id) + SNIC_SG_DESC_ALIGN;
+
+ buf = kzalloc(buf_len, GFP_KERNEL|GFP_DMA);
+ if (!buf) {
+ snic_req_free(snic, rqi);
+ SNIC_HOST_ERR(snic->shost, "Resp Buf Alloc Failed.\n");
+
+ ret = -ENOMEM;
+ goto error;
+ }
+
+ SNIC_BUG_ON((((unsigned long)buf) % SNIC_SG_DESC_ALIGN) != 0);
+
+ pa = pci_map_single(snic->pdev, buf, buf_len, PCI_DMA_FROMDEVICE);
+ if (pci_dma_mapping_error(snic->pdev, pa)) {
+ kfree(buf);
+ snic_req_free(snic, rqi);
+ SNIC_HOST_ERR(snic->shost,
+ "Rpt-tgt rspbuf %p: PCI DMA Mapping Failed\n",
+ buf);
+ ret = -EINVAL;
+
+ goto error;
+ }
+
+
+ SNIC_BUG_ON(pa == 0);
+ rqi->sge_va = (u64) buf;
+
+ snic_report_tgt_init(rqi->req,
+ snic->config.hid,
+ buf,
+ buf_len,
+ (u64)pa,
+ (u64)rqi);
+
+ snic_handle_untagged_req(snic, rqi);
+
+ ret = snic_queue_wq_desc(snic, rqi->req, rqi->req_len);
+ if (ret) {
+ pci_unmap_single(snic->pdev, pa, buf_len, PCI_DMA_FROMDEVICE);
+ kfree(buf);
+ rqi->sge_va = 0;
+ snic_release_untagged_req(snic, rqi);
+ SNIC_HOST_ERR(snic->shost, "Queuing Report Tgts Failed.\n");
+
+ goto error;
+ }
+
+ SNIC_DISC_DBG(snic->shost, "Report Targets Issued.\n");
+
+ return ret;
+
+error:
+ SNIC_HOST_ERR(snic->shost,
+ "Queuing Report Targets Failed, err = %d\n",
+ ret);
+ return ret;
+} /* end of snic_queue_report_tgt_req */
+
+/* call into SML */
+static void
+snic_scsi_scan_tgt(struct work_struct *work)
+{
+ struct snic_tgt *tgt = container_of(work, struct snic_tgt, scan_work);
+ struct Scsi_Host *shost = dev_to_shost(&tgt->dev);
+ unsigned long flags;
+
+ SNIC_HOST_INFO(shost, "Scanning Target id 0x%x\n", tgt->id);
+ scsi_scan_target(&tgt->dev,
+ tgt->channel,
+ tgt->scsi_tgt_id,
+ SCAN_WILD_CARD,
+ 1);
+
+ spin_lock_irqsave(shost->host_lock, flags);
+ tgt->flags &= ~SNIC_TGT_SCAN_PENDING;
+ spin_unlock_irqrestore(shost->host_lock, flags);
+} /* end of snic_scsi_scan_tgt */
+
+/*
+ * snic_tgt_lookup :
+ */
+static struct snic_tgt *
+snic_tgt_lookup(struct snic *snic, struct snic_tgt_id *tgtid)
+{
+ struct list_head *cur, *nxt;
+ struct snic_tgt *tgt = NULL;
+
+ list_for_each_safe(cur, nxt, &snic->disc.tgt_list) {
+ tgt = list_entry(cur, struct snic_tgt, list);
+ if (tgt->id == tgtid->tgt_id)
+ return tgt;
+ tgt = NULL;
+ }
+
+ return tgt;
+} /* end of snic_tgt_lookup */
+
+/*
+ * snic_tgt_dev_release : Called on dropping last ref for snic_tgt object
+ */
+void
+snic_tgt_dev_release(struct device *dev)
+{
+ struct snic_tgt *tgt = dev_to_tgt(dev);
+
+ SNIC_HOST_INFO(snic_tgt_to_shost(tgt),
+ "Target Device ID %d (%s) Permanently Deleted.\n",
+ tgt->id,
+ dev_name(dev));
+
+ SNIC_BUG_ON(!list_empty(&tgt->list));
+ kfree(tgt);
+}
+
+/*
+ * snic_tgt_del : work function to delete snic_tgt
+ */
+static void
+snic_tgt_del(struct work_struct *work)
+{
+ struct snic_tgt *tgt = container_of(work, struct snic_tgt, del_work);
+ struct Scsi_Host *shost = snic_tgt_to_shost(tgt);
+
+ if (tgt->flags & SNIC_TGT_SCAN_PENDING)
+ scsi_flush_work(shost);
+
+ /* Block IOs on child devices, stops new IOs */
+ scsi_target_block(&tgt->dev);
+
+ /* Cleanup IOs */
+ snic_tgt_scsi_abort_io(tgt);
+
+ /* Unblock IOs now, to flush if there are any. */
+ scsi_target_unblock(&tgt->dev, SDEV_TRANSPORT_OFFLINE);
+
+ /* Delete SCSI Target and sdevs */
+ scsi_remove_target(&tgt->dev); /* ?? */
+ device_del(&tgt->dev);
+ put_device(&tgt->dev);
+} /* end of snic_tgt_del */
+
+/* snic_tgt_create: checks for existence of snic_tgt, if it doesn't
+ * it creates one.
+ */
+static struct snic_tgt *
+snic_tgt_create(struct snic *snic, struct snic_tgt_id *tgtid)
+{
+ struct snic_tgt *tgt = NULL;
+ unsigned long flags;
+ int ret;
+
+ tgt = snic_tgt_lookup(snic, tgtid);
+ if (tgt) {
+ /* update the information if required */
+ return tgt;
+ }
+
+ tgt = kzalloc(sizeof(*tgt), GFP_KERNEL);
+ if (!tgt) {
+ SNIC_HOST_ERR(snic->shost, "Failure to allocate snic_tgt.\n");
+ ret = -ENOMEM;
+
+ return tgt;
+ }
+
+ INIT_LIST_HEAD(&tgt->list);
+ tgt->id = tgtid->tgt_id;
+ tgt->channel = 0;
+
+ SNIC_BUG_ON(tgtid->tgt_type > SNIC_TGT_SAN);
+ tgt->tdata.typ = tgtid->tgt_type;
+
+ /*
+ * Plugging into SML Device Tree
+ */
+ tgt->tdata.disc_id = 0;
+ tgt->state = SNIC_TGT_STAT_INIT;
+ device_initialize(&tgt->dev);
+ tgt->dev.parent = get_device(&snic->shost->shost_gendev);
+ tgt->dev.release = snic_tgt_dev_release;
+ INIT_WORK(&tgt->scan_work, snic_scsi_scan_tgt);
+ INIT_WORK(&tgt->del_work, snic_tgt_del);
+ switch (tgtid->tgt_type) {
+ case SNIC_TGT_DAS:
+ dev_set_name(&tgt->dev, "snic_das_tgt:%d:%d-%d",
+ snic->shost->host_no, tgt->channel, tgt->id);
+ break;
+
+ case SNIC_TGT_SAN:
+ dev_set_name(&tgt->dev, "snic_san_tgt:%d:%d-%d",
+ snic->shost->host_no, tgt->channel, tgt->id);
+ break;
+
+ default:
+ SNIC_HOST_INFO(snic->shost, "Target type Unknown Detected.\n");
+ dev_set_name(&tgt->dev, "snic_das_tgt:%d:%d-%d",
+ snic->shost->host_no, tgt->channel, tgt->id);
+ break;
+ }
+
+ spin_lock_irqsave(snic->shost->host_lock, flags);
+ list_add_tail(&tgt->list, &snic->disc.tgt_list);
+ tgt->scsi_tgt_id = snic->disc.nxt_tgt_id++;
+ tgt->state = SNIC_TGT_STAT_ONLINE;
+ spin_unlock_irqrestore(snic->shost->host_lock, flags);
+
+ SNIC_HOST_INFO(snic->shost,
+ "Tgt %d, type = %s detected. Adding..\n",
+ tgt->id, snic_tgt_type_to_str(tgt->tdata.typ));
+
+ ret = device_add(&tgt->dev);
+ if (ret) {
+ SNIC_HOST_ERR(snic->shost,
+ "Snic Tgt: device_add, with err = %d\n",
+ ret);
+
+ put_device(&snic->shost->shost_gendev);
+ kfree(tgt);
+ tgt = NULL;
+
+ return tgt;
+ }
+
+ SNIC_HOST_INFO(snic->shost, "Scanning %s.\n", dev_name(&tgt->dev));
+
+ scsi_queue_work(snic->shost, &tgt->scan_work);
+
+ return tgt;
+} /* end of snic_tgt_create */
+
+/* Handler for discovery */
+void
+snic_handle_tgt_disc(struct work_struct *work)
+{
+ struct snic *snic = container_of(work, struct snic, tgt_work);
+ struct snic_tgt_id *tgtid = NULL;
+ struct snic_tgt *tgt = NULL;
+ unsigned long flags;
+ int i;
+
+ spin_lock_irqsave(&snic->snic_lock, flags);
+ if (snic->in_remove) {
+ spin_unlock_irqrestore(&snic->snic_lock, flags);
+ kfree(snic->disc.rtgt_info);
+
+ return;
+ }
+ spin_unlock_irqrestore(&snic->snic_lock, flags);
+
+ mutex_lock(&snic->disc.mutex);
+ /* Discover triggered during disc in progress */
+ if (snic->disc.req_cnt) {
+ snic->disc.state = SNIC_DISC_DONE;
+ snic->disc.req_cnt = 0;
+ mutex_unlock(&snic->disc.mutex);
+ kfree(snic->disc.rtgt_info);
+ snic->disc.rtgt_info = NULL;
+
+ SNIC_HOST_INFO(snic->shost, "tgt_disc: Discovery restart.\n");
+ /* Start Discovery Again */
+ snic_disc_start(snic);
+
+ return;
+ }
+
+ tgtid = (struct snic_tgt_id *)snic->disc.rtgt_info;
+
+ SNIC_BUG_ON(snic->disc.rtgt_cnt == 0 || tgtid == NULL);
+
+ for (i = 0; i < snic->disc.rtgt_cnt; i++) {
+ tgt = snic_tgt_create(snic, &tgtid[i]);
+ if (!tgt) {
+ int buf_sz = snic->disc.rtgt_cnt * sizeof(*tgtid);
+
+ SNIC_HOST_ERR(snic->shost, "Failed to create tgt.\n");
+ snic_hex_dump("rpt_tgt_rsp", (char *)tgtid, buf_sz);
+ break;
+ }
+ }
+
+ snic->disc.rtgt_info = NULL;
+ snic->disc.state = SNIC_DISC_DONE;
+ mutex_unlock(&snic->disc.mutex);
+
+ SNIC_HOST_INFO(snic->shost, "Discovery Completed.\n");
+
+ kfree(tgtid);
+} /* end of snic_handle_tgt_disc */
+
+
+int
+snic_report_tgt_cmpl_handler(struct snic *snic, struct snic_fw_req *fwreq)
+{
+
+ u8 typ, cmpl_stat;
+ u32 cmnd_id, hid;
+ u64 ctx;
+ struct snic_req_info *rqi = NULL;
+ struct snic_tgt_id *tgtid;
+ int i, ret = 0;
+
+ snic_io_hdr_dec(&fwreq->hdr, &typ, &cmpl_stat, &cmnd_id, &hid, &ctx);
+ rqi = (struct snic_req_info *)ctx;
+ tgtid = (struct snic_tgt_id *) (rqi->sge_va);
+
+ if (fwreq->u.rpt_tgts_cmpl.tgt_cnt == 0) {
+ SNIC_HOST_ERR(snic->shost, "No Targets Found on this host.\n");
+ ret = 1;
+
+ goto end;
+ }
+
+ /* printing list of targets here */
+ SNIC_HOST_INFO(snic->shost,
+ "Target Count = %d\n",
+ fwreq->u.rpt_tgts_cmpl.tgt_cnt);
+
+ SNIC_BUG_ON(fwreq->u.rpt_tgts_cmpl.tgt_cnt > snic->fwinfo.max_tgts);
+
+ for (i = 0; i < fwreq->u.rpt_tgts_cmpl.tgt_cnt; i++)
+ SNIC_HOST_INFO(snic->shost, "Tgt id = 0x%x\n", tgtid[i].tgt_id);
+
+ /*
+ * Queue work for further processing,
+ * Response Buffer Memory is freed after creating targets
+ */
+ snic->disc.rtgt_cnt = fwreq->u.rpt_tgts_cmpl.tgt_cnt;
+ snic->disc.rtgt_info = (u8 *)tgtid;
+ queue_work(snic_glob->event_q, &snic->tgt_work);
+ ret = 0;
+
+end:
+ /* Unmap Response Buffer */
+ snic_pci_unmap_rsp_buf(snic, rqi);
+ if (ret)
+ kfree(tgtid);
+
+ rqi->sge_va = 0;
+ snic_release_untagged_req(snic, rqi);
+
+ return ret;
+} /* end of snic_report_tgt_cmpl_handler */
+
+/* Discovery init fn */
+void
+snic_disc_init(struct snic_disc *disc)
+{
+ INIT_LIST_HEAD(&disc->tgt_list);
+ mutex_init(&disc->mutex);
+ disc->disc_id = 0;
+ disc->nxt_tgt_id = 0;
+ disc->state = SNIC_DISC_INIT;
+ disc->req_cnt = 0;
+ disc->rtgt_cnt = 0;
+ disc->rtgt_info = NULL;
+ disc->cb = NULL;
+} /* end of snic_disc_init */
+
+/* Discovery, uninit fn */
+void
+snic_disc_term(struct snic *snic)
+{
+ struct snic_disc *disc = &snic->disc;
+
+ mutex_lock(&disc->mutex);
+ if (disc->req_cnt) {
+ disc->req_cnt = 0;
+ SNIC_SCSI_DBG(snic->shost, "Terminating Discovery.\n");
+ }
+ mutex_unlock(&disc->mutex);
+}
+
+/*
+ * snic_disc_start: Discovery Start ...
+ */
+int
+snic_disc_start(struct snic *snic)
+{
+ struct snic_disc *disc = &snic->disc;
+ int ret = 0;
+
+ SNIC_SCSI_DBG(snic->shost, "Discovery Start.\n");
+
+ mutex_lock(&disc->mutex);
+ if (disc->state == SNIC_DISC_PENDING) {
+ disc->req_cnt++;
+ mutex_unlock(&disc->mutex);
+
+ return ret;
+ }
+ disc->state = SNIC_DISC_PENDING;
+ mutex_unlock(&disc->mutex);
+
+ ret = snic_queue_report_tgt_req(snic);
+ if (ret)
+ SNIC_HOST_INFO(snic->shost, "Discovery Failed, err=%d.\n", ret);
+
+ return ret;
+} /* end of snic_disc_start */
+
+/*
+ * snic_disc_work :
+ */
+void
+snic_handle_disc(struct work_struct *work)
+{
+ struct snic *snic = container_of(work, struct snic, disc_work);
+ int ret = 0;
+
+ SNIC_HOST_INFO(snic->shost, "disc_work: Discovery\n");
+
+ ret = snic_disc_start(snic);
+ if (ret)
+ goto disc_err;
+
+disc_err:
+ SNIC_HOST_ERR(snic->shost,
+ "disc_work: Discovery Failed w/ err = %d\n",
+ ret);
+} /* end of snic_disc_work */
+
+/*
+ * snic_tgt_del_all : cleanup all snic targets
+ * Called on unbinding the interface
+ */
+void
+snic_tgt_del_all(struct snic *snic)
+{
+ struct snic_tgt *tgt = NULL;
+ struct list_head *cur, *nxt;
+ unsigned long flags;
+
+ mutex_lock(&snic->disc.mutex);
+ spin_lock_irqsave(snic->shost->host_lock, flags);
+
+ list_for_each_safe(cur, nxt, &snic->disc.tgt_list) {
+ tgt = list_entry(cur, struct snic_tgt, list);
+ tgt->state = SNIC_TGT_STAT_DEL;
+ list_del_init(&tgt->list);
+ SNIC_HOST_INFO(snic->shost, "Tgt %d q'ing for del\n", tgt->id);
+ queue_work(snic_glob->event_q, &tgt->del_work);
+ tgt = NULL;
+ }
+ spin_unlock_irqrestore(snic->shost->host_lock, flags);
+
+ scsi_flush_work(snic->shost);
+ mutex_unlock(&snic->disc.mutex);
+} /* end of snic_tgt_del_all */
+#ifdef SNIC_TGTINFO_ENABLE
+
+/*
+ * snic_queue_get_tgtinfo_req: Queues get target info request.
+ */
+int
+snic_queue_get_tgtinfo_req(struct snic *snic, u32 tgt_id)
+{
+ struct snic_req_info *rqi = NULL;
+ struct snic_host_req *req = NULL;
+ int ret = 0;
+
+ rqi = snic_req_init(snic, 0);
+ if (!rqi) {
+ ret = -ENOMEM;
+ goto error;
+ }
+
+ req = rqi->req;
+
+ snic_handle_untagged_req(snic, rqi);
+
+ snic_io_hdr_enc(&req->hdr,
+ SNIC_REQ_TGT_INFO,
+ 0 /* Status */,
+ SCSI_NO_TAG,
+ snic->config.hid,
+ 0 /* SG Count */,
+ (u64)rqi /* Initiator Context */);
+
+ req->u.tgt_info.tgt_id = tgt_id;
+ ret = snic_queue_wq_desc(snic, rqi->req, rqi->req_len);
+ if (ret) {
+ snic_release_untagged_req(snic, rqi);
+ SNIC_HOST_ERR(snic->shost, "Queuing Get Tgt Info Failed.\n");
+
+ goto error;
+ }
+
+ SNIC_DISC_DBG(snic->shost, "Get Tgt Info Issued for Tgt Id %d.\n",
+ tgt_id);
+
+ return ret;
+
+error:
+ SNIC_HOST_ERR(snic->shost,
+ "Queuing Get Target Info Failed for Tgt %d, err = %d\n",
+ tgt_id,
+ ret);
+ return ret;
+} /* end of snic_queue_tgtinfo_req */
+#endif
diff --git a/drivers/scsi/snic/snic_disc.h b/drivers/scsi/snic/snic_disc.h
new file mode 100644
index 0000000..97fa3f5
--- /dev/null
+++ b/drivers/scsi/snic/snic_disc.h
@@ -0,0 +1,124 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef __SNIC_DISC_H
+#define __SNIC_DISC_H
+
+#include "snic_fwint.h"
+
+enum snic_disc_state {
+ SNIC_DISC_NONE,
+ SNIC_DISC_INIT,
+ SNIC_DISC_PENDING,
+ SNIC_DISC_DONE
+};
+
+struct snic;
+struct snic_disc {
+ struct list_head tgt_list;
+ enum snic_disc_state state;
+ struct mutex mutex;
+ u16 disc_id;
+ u8 req_cnt;
+ u32 nxt_tgt_id;
+ u32 rtgt_cnt;
+ u8 *rtgt_info;
+ struct delayed_work disc_timeout;
+ void (*cb)(struct snic *);
+};
+
+#define SNIC_TGT_NAM_LEN 16
+
+enum snic_tgt_state {
+ SNIC_TGT_STAT_NONE,
+ SNIC_TGT_STAT_INIT,
+ SNIC_TGT_STAT_ONLINE, /* Target is Online */
+ SNIC_TGT_STAT_OFFLINE, /* Target is Offline */
+ SNIC_TGT_STAT_DEL,
+};
+
+struct snic_tgt_priv {
+ struct list_head list;
+ enum snic_tgt_type typ;
+ u16 disc_id;
+ char *name[SNIC_TGT_NAM_LEN];
+
+ union {
+ /*DAS Target specific info */
+ /*SAN Target specific info */
+ u8 dummmy;
+ } u;
+};
+
+/* snic tgt flags */
+#define SNIC_TGT_SCAN_PENDING 0x01
+
+struct snic_tgt {
+ struct list_head list;
+ u16 id;
+ u16 channel;
+ u32 flags;
+ u32 scsi_tgt_id;
+ enum snic_tgt_state state;
+ struct device dev;
+ struct work_struct scan_work;
+ struct work_struct del_work;
+ struct snic_tgt_priv tdata;
+};
+
+
+struct snic_fw_req;
+
+void snic_disc_init(struct snic_disc *);
+int snic_disc_start(struct snic *);
+void snic_disc_term(struct snic *);
+int snic_report_tgt_cmpl_handler(struct snic *, struct snic_fw_req *);
+int snic_tgtinfo_cmpl_handler(struct snic *snic, struct snic_fw_req *fwreq);
+void snic_process_report_tgts_rsp(struct work_struct *);
+void snic_handle_tgt_disc(struct work_struct *);
+void snic_handle_disc(struct work_struct *);
+void snic_tgt_dev_release(struct device *);
+void snic_tgt_del_all(struct snic *);
+
+#define dev_to_tgt(d) \
+ container_of(d, struct snic_tgt, dev)
+
+static inline int
+is_snic_target(struct device *dev)
+{
+ return dev->release == snic_tgt_dev_release;
+}
+
+#define starget_to_tgt(st) \
+ (is_snic_target(((struct scsi_target *) st)->dev.parent) ? \
+ dev_to_tgt(st->dev.parent) : NULL)
+
+#define snic_tgt_to_shost(t) \
+ dev_to_shost(t->dev.parent)
+
+static inline int
+snic_tgt_chkready(struct snic_tgt *tgt)
+{
+ if (tgt->state == SNIC_TGT_STAT_ONLINE)
+ return 0;
+ else
+ return DID_NO_CONNECT << 16;
+}
+
+const char *snic_tgt_state_to_str(int);
+int snic_tgt_scsi_abort_io(struct snic_tgt *);
+#endif /* end of __SNIC_DISC_H */
--
1.8.5.4
^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH v4 5/9] snic:add SCSI handling, AEN, and fwreset handling
2015-04-09 11:49 [PATCH v4 0/9] snic:initial submission of snic driver for Cisco SCSI HBA Narsimhulu Musini
` (3 preceding siblings ...)
2015-04-09 11:49 ` [PATCH v4 4/9] snic:Add snic target discovery Narsimhulu Musini
@ 2015-04-09 11:49 ` Narsimhulu Musini
2015-04-09 13:16 ` Hannes Reinecke
2015-04-09 11:49 ` [PATCH v4 6/9] snic:Add low level queuing interfaces Narsimhulu Musini
` (3 subsequent siblings)
8 siblings, 1 reply; 24+ messages in thread
From: Narsimhulu Musini @ 2015-04-09 11:49 UTC (permalink / raw)
To: JBottomley, linux-scsi, hare, hch; +Cc: Narsimhulu Musini, Sesidhar Baddela
snic_scsi.c contains scsi handling, includes queuing io, abort, lun reset,
and host reset. Also it handles asynchronous event notifications from FW.
Signed-off-by: Narsimhulu Musini <nmusini@cisco.com>
Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
---
* v2
- Changed queuecommand to lock-free version.
- Converted custom error codes to standard error codes.
drivers/scsi/snic/snic_scsi.c | 2638 +++++++++++++++++++++++++++++++++++++++++
1 file changed, 2638 insertions(+)
create mode 100644 drivers/scsi/snic/snic_scsi.c
diff --git a/drivers/scsi/snic/snic_scsi.c b/drivers/scsi/snic/snic_scsi.c
new file mode 100644
index 0000000..5d877f0
--- /dev/null
+++ b/drivers/scsi/snic/snic_scsi.c
@@ -0,0 +1,2638 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/mempool.h>
+#include <linux/errno.h>
+#include <linux/init.h>
+#include <linux/workqueue.h>
+#include <linux/pci.h>
+#include <linux/spinlock.h>
+#include <linux/delay.h>
+#include <linux/gfp.h>
+#include <scsi/scsi.h>
+#include <scsi/scsi_host.h>
+#include <scsi/scsi_device.h>
+#include <scsi/scsi_cmnd.h>
+#include <scsi/scsi_tcq.h>
+#include <scsi/scsi_dbg.h>
+
+#include "snic_io.h"
+#include "snic.h"
+
+#define snic_cmd_tag(sc) (((struct scsi_cmnd *) sc)->request->tag)
+
+const char *snic_state_str[] = {
+ [SNIC_INIT] = "SNIC_INIT",
+ [SNIC_ERROR] = "SNIC_ERROR",
+ [SNIC_ONLINE] = "SNIC_ONLINE",
+ [SNIC_OFFLINE] = "SNIC_OFFLINE",
+ [SNIC_FWRESET] = "SNIC_FWRESET",
+};
+
+static const char * const snic_req_state_str[] = {
+ [SNIC_IOREQ_NOT_INITED] = "SNIC_IOREQ_NOT_INITED",
+ [SNIC_IOREQ_PENDING] = "SNIC_IOREQ_PENDING",
+ [SNIC_IOREQ_ABTS_PENDING] = "SNIC_IOREQ_ABTS_PENDING",
+ [SNIC_IOREQ_ABTS_COMPLETE] = "SNIC_IOREQ_ABTS_COMPELTE",
+ [SNIC_IOREQ_LR_PENDING] = "SNIC_IOREQ_LR_PENDING",
+ [SNIC_IOREQ_LR_COMPLETE] = "SNIC_IOREQ_LR_COMPELTE",
+ [SNIC_IOREQ_COMPLETE] = "SNIC_IOREQ_CMD_COMPELTE",
+};
+
+/* snic cmd status strings */
+static const char * const snic_io_status_str[] = {
+ [SNIC_STAT_IO_SUCCESS] = "SNIC_STAT_IO_SUCCESS", /* 0x0 */
+ [SNIC_STAT_INVALID_HDR] = "SNIC_STAT_INVALID_HDR",
+ [SNIC_STAT_OUT_OF_RES] = "SNIC_STAT_OUT_OF_RES",
+ [SNIC_STAT_INVALID_PARM] = "SNIC_STAT_INVALID_PARM",
+ [SNIC_STAT_REQ_NOT_SUP] = "SNIC_STAT_REQ_NOT_SUP",
+ [SNIC_STAT_IO_NOT_FOUND] = "SNIC_STAT_IO_NOT_FOUND",
+ [SNIC_STAT_ABORTED] = "SNIC_STAT_ABORTED",
+ [SNIC_STAT_TIMEOUT] = "SNIC_STAT_TIMEOUT",
+ [SNIC_STAT_SGL_INVALID] = "SNIC_STAT_SGL_INVALID",
+ [SNIC_STAT_DATA_CNT_MISMATCH] = "SNIC_STAT_DATA_CNT_MISMATCH",
+ [SNIC_STAT_FW_ERR] = "SNIC_STAT_FW_ERR",
+ [SNIC_STAT_ITMF_REJECT] = "SNIC_STAT_ITMF_REJECT",
+ [SNIC_STAT_ITMF_FAIL] = "SNIC_STAT_ITMF_FAIL",
+ [SNIC_STAT_ITMF_INCORRECT_LUN] = "SNIC_STAT_ITMF_INCORRECT_LUN",
+ [SNIC_STAT_CMND_REJECT] = "SNIC_STAT_CMND_REJECT",
+ [SNIC_STAT_DEV_OFFLINE] = "SNIC_STAT_DEV_OFFLINE",
+ [SNIC_STAT_NO_BOOTLUN] = "SNIC_STAT_NO_BOOTLUN",
+ [SNIC_STAT_SCSI_ERR] = "SNIC_STAT_SCSI_ERR",
+ [SNIC_STAT_NOT_READY] = "SNIC_STAT_NOT_READY",
+ [SNIC_STAT_FATAL_ERROR] = "SNIC_STAT_FATAL_ERROR",
+};
+
+static void snic_scsi_cleanup(struct snic *, int);
+
+const char *
+snic_state_to_str(unsigned int state)
+{
+ if (state >= ARRAY_SIZE(snic_state_str) || !snic_state_str[state])
+ return "Unknown";
+
+ return snic_state_str[state];
+}
+
+static const char *
+snic_io_status_to_str(unsigned int state)
+{
+ if ((state >= ARRAY_SIZE(snic_io_status_str)) ||
+ (!snic_io_status_str[state]))
+ return "Unknown";
+
+ return snic_io_status_str[state];
+}
+
+static const char *
+snic_ioreq_state_to_str(unsigned int state)
+{
+ if (state >= ARRAY_SIZE(snic_req_state_str) ||
+ !snic_req_state_str[state])
+ return "Unknown";
+
+ return snic_req_state_str[state];
+}
+
+static inline spinlock_t *
+snic_io_lock_hash(struct snic *snic, struct scsi_cmnd *sc)
+{
+ u32 hash = snic_cmd_tag(sc) & (SNIC_IO_LOCKS - 1);
+
+ return &snic->io_req_lock[hash];
+}
+
+static inline spinlock_t *
+snic_io_lock_tag(struct snic *snic, int tag)
+{
+ return &snic->io_req_lock[tag & (SNIC_IO_LOCKS - 1)];
+}
+
+/* snic_release_req_buf : Releases snic_req_info */
+static void
+snic_release_req_buf(struct snic *snic,
+ struct snic_req_info *rqi,
+ struct scsi_cmnd *sc)
+{
+ struct snic_host_req *req = rqi_to_req(rqi);
+
+ /* Freeing cmd without marking completion, not okay */
+ SNIC_BUG_ON(!((CMD_STATE(sc) == SNIC_IOREQ_COMPLETE) ||
+ (CMD_STATE(sc) == SNIC_IOREQ_ABTS_COMPLETE) ||
+ (CMD_FLAGS(sc) & SNIC_DEV_RST_NOTSUP) ||
+ (CMD_FLAGS(sc) & SNIC_IO_INTERNAL_TERM_ISSUED) ||
+ (CMD_FLAGS(sc) & SNIC_DEV_RST_TERM_ISSUED) ||
+ (CMD_FLAGS(sc) & SNIC_SCSI_CLEANUP) ||
+ (CMD_STATE(sc) == SNIC_IOREQ_LR_COMPLETE)));
+
+ SNIC_SCSI_DBG(snic->shost,
+ "Rel_req:sc %p:tag %x:rqi %p:ioreq %p:abt %p:dr %p: state %s:flags 0x%llx\n",
+ sc, snic_cmd_tag(sc), rqi, rqi->req, rqi->abort_req,
+ rqi->dr_req, snic_ioreq_state_to_str(CMD_STATE(sc)),
+ CMD_FLAGS(sc));
+
+ if (req->u.icmnd.sense_addr)
+ pci_unmap_single(snic->pdev,
+ req->u.icmnd.sense_addr,
+ SCSI_SENSE_BUFFERSIZE,
+ PCI_DMA_FROMDEVICE);
+
+ scsi_dma_unmap(sc);
+
+ snic_req_free(snic, rqi);
+} /* end of snic_release_req_buf */
+
+/*
+ * snic_queue_icmnd_req : Queues snic_icmnd request
+ */
+static int
+snic_queue_icmnd_req(struct snic *snic,
+ struct snic_req_info *rqi,
+ struct scsi_cmnd *sc,
+ int sg_cnt)
+{
+ struct scatterlist *sg;
+ struct snic_sg_desc *sgd;
+ dma_addr_t pa = 0;
+ struct scsi_lun lun;
+ int flags = 0;
+ int ret = 0;
+ unsigned int i;
+
+ if (sg_cnt) {
+ flags = SNIC_ICMND_ESGL;
+ sgd = (struct snic_sg_desc *) req_to_sgl(rqi->req);
+
+ for_each_sg(scsi_sglist(sc), sg, sg_cnt, i) {
+ sgd->addr = cpu_to_le64(sg_dma_address(sg));
+ sgd->len = cpu_to_le32(sg_dma_len(sg));
+ sgd->_resvd = 0;
+ sgd++;
+ }
+ }
+
+ pa = pci_map_single(snic->pdev,
+ sc->sense_buffer,
+ SCSI_SENSE_BUFFERSIZE,
+ PCI_DMA_FROMDEVICE);
+
+ if (pci_dma_mapping_error(snic->pdev, pa)) {
+ SNIC_HOST_ERR(snic->shost,
+ "QIcmnd:PCI Map Failed for sns buf %p tag %x\n",
+ sc->sense_buffer, snic_cmd_tag(sc));
+ ret = -ENOMEM;
+
+ return ret;
+ }
+
+ int_to_scsilun(sc->device->lun, &lun);
+ if (sc->sc_data_direction == DMA_FROM_DEVICE)
+ flags |= SNIC_ICMND_RD;
+ if (sc->sc_data_direction == DMA_TO_DEVICE)
+ flags |= SNIC_ICMND_WR;
+
+ /* Initialize icmnd */
+ snic_icmnd_init(rqi->req,
+ snic_cmd_tag(sc),
+ snic->config.hid, /* hid */
+ (u64)rqi,
+ flags, /* command flags */
+ rqi->tgt_id,
+ lun.scsi_lun,
+ sc->cmnd,
+ sc->cmd_len,
+ scsi_bufflen(sc),
+ sg_cnt,
+ (u64)req_to_sgl(rqi->req),
+ pa, /* sense buffer pa */
+ SCSI_SENSE_BUFFERSIZE);
+
+ ret = snic_queue_wq_desc(snic, rqi->req, rqi->req_len);
+ if (ret)
+ SNIC_HOST_ERR(snic->shost,
+ "QIcmnd: Queuing Icmnd Failed. ret = %d\n",
+ ret);
+
+ return ret;
+} /* end of snic_queue_icmnd_req */
+
+/*
+ * snic_issue_scsi_req : Prepares IO request and Issues to FW.
+ */
+static int
+snic_issue_scsi_req(struct snic *snic,
+ struct snic_tgt *tgt,
+ struct scsi_cmnd *sc)
+{
+ struct snic_req_info *rqi = NULL;
+ int sg_cnt = 0;
+ int ret = 0;
+ u32 tag = snic_cmd_tag(sc);
+ u64 cmd_trc = 0, cmd_st_flags = 0;
+ spinlock_t *io_lock = NULL;
+ unsigned long flags;
+
+ CMD_STATE(sc) = SNIC_IOREQ_NOT_INITED;
+ CMD_FLAGS(sc) = SNIC_NO_FLAGS;
+ sg_cnt = scsi_dma_map(sc);
+ if (sg_cnt < 0) {
+ SNIC_TRC((u16)snic->shost->host_no, tag, (u64)sc, 0,
+ sc->cmnd[0], sg_cnt, CMD_STATE(sc));
+
+ SNIC_HOST_ERR(snic->shost, "issue_sc:Failed to map SG List.\n");
+ ret = -ENOMEM;
+
+ goto issue_sc_end;
+ }
+
+ rqi = snic_req_init(snic, sg_cnt);
+ if (!rqi) {
+ scsi_dma_unmap(sc);
+ ret = -ENOMEM;
+
+ goto issue_sc_end;
+ }
+
+ rqi->tgt_id = tgt->id;
+ rqi->sc = sc;
+
+ CMD_STATE(sc) = SNIC_IOREQ_PENDING;
+ CMD_SP(sc) = (char *) rqi;
+ cmd_trc = SNIC_TRC_CMD(sc);
+ CMD_FLAGS(sc) |= (SNIC_IO_INITIALIZED | SNIC_IO_ISSUED);
+ cmd_st_flags = SNIC_TRC_CMD_STATE_FLAGS(sc);
+ io_lock = snic_io_lock_hash(snic, sc);
+
+ /* create wq desc and enqueue it */
+ ret = snic_queue_icmnd_req(snic, rqi, sc, sg_cnt);
+ if (ret) {
+ SNIC_HOST_ERR(snic->shost,
+ "issue_sc: icmnd qing Failed for sc %p, err %d\n",
+ sc, ret);
+
+ spin_lock_irqsave(io_lock, flags);
+ rqi = (struct snic_req_info *) CMD_SP(sc);
+ CMD_SP(sc) = NULL;
+ CMD_STATE(sc) = SNIC_IOREQ_COMPLETE;
+ CMD_FLAGS(sc) &= ~SNIC_IO_ISSUED; /* turn off the flag */
+ spin_unlock_irqrestore(io_lock, flags);
+
+ if (rqi)
+ snic_release_req_buf(snic, rqi, sc);
+
+ SNIC_TRC(snic->shost->host_no, tag, (u64)sc, 0, 0, 0,
+ SNIC_TRC_CMD_STATE_FLAGS(sc));
+ } else {
+ u32 io_sz = scsi_bufflen(sc) >> 9;
+ u32 qtime = jiffies - rqi->start_time;
+ struct snic_io_stats *iostats = &snic->s_stats.io;
+
+ if (io_sz > atomic64_read(&iostats->max_io_sz))
+ atomic64_set(&iostats->max_io_sz, io_sz);
+
+ if (qtime > atomic64_read(&iostats->max_qtime))
+ atomic64_set(&iostats->max_qtime, qtime);
+
+ SNIC_SCSI_DBG(snic->shost,
+ "issue_sc:sc %p, tag %d queued to WQ.\n",
+ sc, tag);
+
+ SNIC_TRC(snic->shost->host_no, tag, (u64)sc, (u64)rqi, sg_cnt,
+ cmd_trc, cmd_st_flags);
+ }
+
+issue_sc_end:
+
+ return ret;
+} /* end of snic_issue_scsi_req */
+
+
+/*
+ * snic_queuecommand
+ * Routine to send a scsi cdb to LLD
+ * Called with host_lock held and interrupts disabled
+ */
+int
+snic_queuecommand(struct Scsi_Host *shost, struct scsi_cmnd *sc)
+{
+ struct snic_tgt *tgt = NULL;
+ struct snic *snic = shost_priv(shost);
+ int ret;
+
+ tgt = starget_to_tgt(scsi_target(sc->device));
+ ret = snic_tgt_chkready(tgt);
+ if (ret) {
+ SNIC_HOST_ERR(shost, "Tgt %p id %d Not Ready.\n", tgt, tgt->id);
+ atomic64_inc(&snic->s_stats.misc.tgt_not_rdy);
+ sc->result = ret;
+ sc->scsi_done(sc);
+
+ return 0;
+ }
+
+ if (snic_get_state(snic) != SNIC_ONLINE) {
+ SNIC_HOST_ERR(shost, "snic state is %s\n",
+ snic_state_str[snic_get_state(snic)]);
+
+ return SCSI_MLQUEUE_HOST_BUSY;
+ }
+ atomic_inc(&snic->ios_inflight);
+
+ SNIC_SCSI_DBG(shost, "sc %p Tag %d (sc %0x) lun %lld in snic_qcmd\n",
+ sc, snic_cmd_tag(sc), sc->cmnd[0], sc->device->lun);
+
+ memset(scsi_cmd_priv(sc), 0, sizeof(struct snic_internal_io_state));
+
+#ifdef SNIC_DEBUG
+ scsi_print_command(sc);
+#endif
+
+ ret = snic_issue_scsi_req(snic, tgt, sc);
+ if (ret) {
+ SNIC_HOST_ERR(shost, "Failed to Q, Scsi Req w/ err %d.\n", ret);
+ ret = SCSI_MLQUEUE_HOST_BUSY;
+ } else
+ snic_stats_update_active_ios(&snic->s_stats);
+
+ atomic_dec(&snic->ios_inflight);
+
+ return ret;
+} /* end of snic_queuecommand */
+
+/*
+ * snic_process_abts_pending_state:
+ * caller should hold IO lock
+ */
+static void
+snic_proc_tmreq_pending_state(struct snic *snic,
+ struct scsi_cmnd *sc,
+ u8 cmpl_status)
+{
+ int state = CMD_STATE(sc);
+
+ if (state == SNIC_IOREQ_ABTS_PENDING)
+ CMD_FLAGS(sc) |= SNIC_IO_ABTS_PENDING;
+ else if (state == SNIC_IOREQ_LR_PENDING)
+ CMD_FLAGS(sc) |= SNIC_DEV_RST_PENDING;
+ else
+ SNIC_BUG_ON(1);
+
+ switch (cmpl_status) {
+ case SNIC_STAT_IO_SUCCESS:
+ CMD_FLAGS(sc) |= SNIC_IO_DONE;
+ break;
+
+ case SNIC_STAT_ABORTED:
+ CMD_FLAGS(sc) |= SNIC_IO_ABORTED;
+ break;
+
+ default:
+ SNIC_BUG_ON(1);
+ }
+}
+
+/*
+ * snic_process_io_failed_state:
+ * Processes IO's error states
+ */
+static void
+snic_process_io_failed_state(struct snic *snic,
+ struct snic_icmnd_cmpl *icmnd_cmpl,
+ struct scsi_cmnd *sc,
+ u8 cmpl_stat)
+{
+ int res = 0;
+
+ switch (cmpl_stat) {
+ case SNIC_STAT_TIMEOUT: /* Req was timedout */
+ atomic64_inc(&snic->s_stats.misc.io_tmo);
+ res = DID_TIME_OUT;
+ break;
+
+ case SNIC_STAT_ABORTED: /* Req was aborted */
+ atomic64_inc(&snic->s_stats.misc.io_aborted);
+ res = DID_ERROR;
+ break;
+
+ case SNIC_STAT_DATA_CNT_MISMATCH:/* Recv/Sent more/less data than exp */
+ atomic64_inc(&snic->s_stats.misc.data_cnt_mismat);
+ scsi_set_resid(sc, icmnd_cmpl->resid);
+ res = DID_ERROR;
+ break;
+
+ case SNIC_STAT_OUT_OF_RES: /* Out of resources to complete request */
+ atomic64_inc(&snic->s_stats.fw.out_of_res);
+ res = DID_REQUEUE;
+ break;
+
+ case SNIC_STAT_IO_NOT_FOUND: /* Requested I/O was not found */
+ atomic64_inc(&snic->s_stats.io.io_not_found);
+ res = DID_ERROR;
+ break;
+
+ case SNIC_STAT_SGL_INVALID: /* Req was aborted to due to sgl error*/
+ atomic64_inc(&snic->s_stats.misc.sgl_inval);
+ res = DID_ERROR;
+ break;
+
+ case SNIC_STAT_FW_ERR: /* Req terminated due to FW Error */
+ atomic64_inc(&snic->s_stats.fw.io_errs);
+ res = DID_ERROR;
+ break;
+
+ case SNIC_STAT_SCSI_ERR: /* FW hits SCSI Error */
+ atomic64_inc(&snic->s_stats.fw.scsi_errs);
+ res = DID_ERROR;
+ break;
+
+ case SNIC_STAT_INVALID_HDR: /* Hdr contains invalid data */
+ case SNIC_STAT_INVALID_PARM: /* Some param in req is invalid */
+ case SNIC_STAT_REQ_NOT_SUP: /* Req type is not supported */
+ case SNIC_STAT_CMND_REJECT: /* Req rejected */
+ case SNIC_STAT_DEV_OFFLINE: /* Device offline */
+ case SNIC_STAT_NOT_READY: /* XPT yet to initialize */
+ case SNIC_STAT_FATAL_ERROR: /* XPT Error */
+ default:
+ SNIC_SCSI_DBG(snic->shost,
+ "Invalid Hdr/Param or Req Not Supported or Cmnd Rejected or Device Offline. or Unknown\n");
+ res = DID_ERROR;
+ break;
+ }
+
+ SNIC_HOST_ERR(snic->shost, "fw returns failed status %s flags 0x%llx\n",
+ snic_io_status_to_str(cmpl_stat), CMD_FLAGS(sc));
+
+ /* Set sc->result */
+ sc->result = (res << 16) | icmnd_cmpl->scsi_status;
+} /* end of snic_process_io_failed_state */
+
+/*
+ * snic_tmreq_pending : is task management in progress.
+ */
+static int
+snic_tmreq_pending(struct scsi_cmnd *sc)
+{
+ int state = CMD_STATE(sc);
+
+ return ((state == SNIC_IOREQ_ABTS_PENDING) ||
+ (state == SNIC_IOREQ_LR_PENDING));
+}
+
+/*
+ * snic_process_icmnd_cmpl_status:
+ * Caller should hold io_lock
+ */
+static int
+snic_process_icmnd_cmpl_status(struct snic *snic,
+ struct snic_icmnd_cmpl *icmnd_cmpl,
+ u8 cmpl_stat,
+ struct scsi_cmnd *sc)
+{
+ u8 scsi_stat = icmnd_cmpl->scsi_status;
+ u64 xfer_len = 0;
+ int ret = 0;
+
+ /* Mark the IO as complete */
+ CMD_STATE(sc) = SNIC_IOREQ_COMPLETE;
+
+ if (likely(cmpl_stat == SNIC_STAT_IO_SUCCESS)) {
+ sc->result = (DID_OK << 16) | scsi_stat;
+
+ xfer_len = scsi_bufflen(sc);
+
+ /* Update SCSI Cmd with resid value */
+ scsi_set_resid(sc, icmnd_cmpl->resid);
+
+ if (icmnd_cmpl->flags & SNIC_ICMND_CMPL_UNDR_RUN) {
+ xfer_len -= icmnd_cmpl->resid;
+ atomic64_inc(&snic->s_stats.misc.io_under_run);
+ }
+
+ if (icmnd_cmpl->scsi_status == SAM_STAT_TASK_SET_FULL)
+ atomic64_inc(&snic->s_stats.misc.qfull);
+
+ ret = 0;
+ } else {
+ snic_process_io_failed_state(snic, icmnd_cmpl, sc, cmpl_stat);
+ atomic64_inc(&snic->s_stats.io.fail);
+ SNIC_HOST_ERR(snic->shost,
+ "icmnd_cmpl: IO Failed : Hdr Status %s flags 0x%llx\n",
+ snic_io_status_to_str(cmpl_stat), CMD_FLAGS(sc));
+ ret = 1;
+ }
+
+ return ret;
+} /* end of snic_process_icmnd_cmpl_status */
+
+
+/*
+ * snic_icmnd_cmpl_handler
+ * Routine to handle icmnd completions
+ */
+static void
+snic_icmnd_cmpl_handler(struct snic *snic, struct snic_fw_req *fwreq)
+{
+ u8 typ, hdr_stat;
+ u32 cmnd_id, hid;
+ u64 ctx;
+ struct scsi_cmnd *sc = NULL;
+ struct snic_icmnd_cmpl *icmnd_cmpl = NULL;
+ struct snic_host_req *req = NULL;
+ struct snic_req_info *rqi = NULL;
+ unsigned long flags, start_time;
+ spinlock_t *io_lock;
+ u8 sc_stat = 0;
+
+ snic_io_hdr_dec(&fwreq->hdr, &typ, &hdr_stat, &cmnd_id, &hid, &ctx);
+ icmnd_cmpl = &fwreq->u.icmnd_cmpl;
+ sc_stat = icmnd_cmpl->scsi_status;
+
+ SNIC_SCSI_DBG(snic->shost,
+ "Icmnd_cmpl: type = %x, hdr_stat = %x, cmnd_id = %x, hid = %x,i ctx = %llx\n",
+ typ, hdr_stat, cmnd_id, hid, ctx);
+
+ if (cmnd_id >= snic->max_tag_id) {
+ SNIC_HOST_ERR(snic->shost,
+ "Icmnd_cmpl:Tag Error:Out of Range Tag %d, hdr status = %s\n",
+ cmnd_id, snic_io_status_to_str(hdr_stat));
+ return;
+ }
+
+ sc = scsi_host_find_tag(snic->shost, cmnd_id);
+ WARN_ON_ONCE(!sc);
+
+ if (!sc) {
+ atomic64_inc(&snic->s_stats.io.sc_null);
+ SNIC_HOST_ERR(snic->shost,
+ "Icmnd_cmpl: Scsi Cmnd Not found, sc = NULL Hdr Status = %s tag = 0x%x fwreq = 0x%p\n",
+ snic_io_status_to_str(hdr_stat),
+ cmnd_id,
+ fwreq);
+
+ SNIC_TRC(snic->shost->host_no, cmnd_id, 0,
+ ((u64)hdr_stat << 16 |
+ (u64)sc_stat << 8 | (u64)icmnd_cmpl->flags),
+ fwreq, icmnd_cmpl->resid, ctx);
+
+ return;
+ }
+
+
+#ifdef SNIC_DEBUG
+ scsi_print_result(sc);
+#endif
+
+ io_lock = snic_io_lock_hash(snic, sc);
+
+ spin_lock_irqsave(io_lock, flags);
+ rqi = (struct snic_req_info *) CMD_SP(sc);
+ SNIC_SCSI_DBG(snic->shost,
+ "Icmnd_cmpl:lun %lld sc %p cmd %xtag %d flags 0x%llx rqi %p\n",
+ sc->device->lun, sc, sc->cmnd[0], snic_cmd_tag(sc),
+ CMD_FLAGS(sc), rqi);
+
+ SNIC_BUG_ON(rqi != (struct snic_req_info *)ctx);
+ WARN_ON_ONCE(req);
+ if (!rqi) {
+ atomic64_inc(&snic->s_stats.io.req_null);
+ CMD_FLAGS(sc) |= SNIC_IO_REQ_NULL;
+ spin_unlock_irqrestore(io_lock, flags);
+
+ SNIC_HOST_ERR(snic->shost,
+ "Icmnd_cmpl:Host Req Not Found(null), Hdr Status %s, Tag 0x%x, sc 0x%p flags 0x%llx\n",
+ snic_io_status_to_str(hdr_stat),
+ cmnd_id, sc, CMD_FLAGS(sc));
+ return;
+ }
+
+ rqi = (struct snic_req_info *) ctx;
+ start_time = rqi->start_time;
+
+ /* firmware completed the io */
+ rqi->io_cmpl = 1;
+
+ /*
+ * if SCSI-ML has already issued abort on this command,
+ * ignore completion of the IO. The abts path will clean it up
+ */
+ if (unlikely(snic_tmreq_pending(sc))) {
+ snic_proc_tmreq_pending_state(snic, sc, hdr_stat);
+ spin_unlock_irqrestore(io_lock, flags);
+
+ snic_stats_update_io_cmpl(&snic->s_stats);
+
+ /* Expected value is SNIC_STAT_ABORTED */
+ if (likely(hdr_stat == SNIC_STAT_ABORTED))
+ return;
+
+ SNIC_SCSI_DBG(snic->shost,
+ "icmnd_cmpl:TM Req Pending(%s), Hdr Status %s sc 0x%p scsi status %x resid %d flags 0x%llx\n",
+ snic_ioreq_state_to_str(CMD_STATE(sc)),
+ snic_io_status_to_str(hdr_stat),
+ sc, sc_stat, icmnd_cmpl->resid, CMD_FLAGS(sc));
+
+ SNIC_TRC(snic->shost->host_no, cmnd_id, sc,
+ jiffies_to_msecs(jiffies - start_time), fwreq,
+ SNIC_TRC_CMD(sc), SNIC_TRC_CMD_STATE_FLAGS(sc));
+
+ return;
+ }
+
+ if (snic_process_icmnd_cmpl_status(snic, icmnd_cmpl, hdr_stat, sc)) {
+#ifdef SNIC_DEBUG
+ scsi_print_command(sc);
+#endif
+ SNIC_HOST_ERR(snic->shost,
+ "icmnd_cmpl:IO Failed, sc 0x%p Tag %d Cmd %x Hdr Status %s flags 0x%llx\n",
+ sc, sc->cmnd[0], cmnd_id,
+ snic_io_status_to_str(hdr_stat), CMD_FLAGS(sc));
+ }
+
+ /* Break link with the SCSI Command */
+ CMD_SP(sc) = NULL;
+ CMD_FLAGS(sc) |= SNIC_IO_DONE;
+
+ spin_unlock_irqrestore(io_lock, flags);
+
+ /* For now, consider only successful IO. */
+ snic_calc_io_process_time(snic, rqi);
+
+ snic_release_req_buf(snic, rqi, sc);
+
+ SNIC_TRC(snic->shost->host_no, cmnd_id, (u64)sc,
+ jiffies_to_msecs(jiffies - start_time), (u64)fwreq,
+ SNIC_TRC_CMD(sc), SNIC_TRC_CMD_STATE_FLAGS(sc));
+
+
+ if (sc->scsi_done)
+ sc->scsi_done(sc);
+
+ snic_stats_update_io_cmpl(&snic->s_stats);
+} /* end of snic_icmnd_cmpl_handler */
+
+static void
+snic_proc_dr_cmpl_locked(struct snic *snic,
+ struct snic_fw_req *fwreq,
+ u8 cmpl_stat,
+ u32 cmnd_id,
+ struct scsi_cmnd *sc)
+{
+ struct snic_req_info *rqi = (struct snic_req_info *) CMD_SP(sc);
+ u32 start_time = rqi->start_time;
+
+ CMD_LR_STATUS(sc) = cmpl_stat;
+
+ SNIC_SCSI_DBG(snic->shost, "itmf_cmpl: Cmd State = %s\n",
+ snic_ioreq_state_to_str(CMD_STATE(sc)));
+
+ if (CMD_STATE(sc) == SNIC_IOREQ_ABTS_PENDING) {
+ CMD_FLAGS(sc) |= SNIC_DEV_RST_ABTS_PENDING;
+
+ SNIC_TRC(snic->shost->host_no, cmnd_id, sc,
+ jiffies_to_msecs(jiffies - start_time),
+ fwreq, 0, SNIC_TRC_CMD_STATE_FLAGS(sc));
+
+ SNIC_SCSI_DBG(snic->shost,
+ "itmf_cmpl: Terminate Pending Dev Reset Cmpl Recvd.id %x, status %s flags 0x%llx\n",
+ (int)(cmnd_id & SNIC_TAG_MASK),
+ snic_io_status_to_str(cmpl_stat),
+ CMD_FLAGS(sc));
+
+ return;
+ }
+
+
+ if (CMD_FLAGS(sc) & SNIC_DEV_RST_TIMEDOUT) {
+ SNIC_TRC(snic->shost->host_no, cmnd_id, sc,
+ jiffies_to_msecs(jiffies - start_time),
+ fwreq, 0, SNIC_TRC_CMD_STATE_FLAGS(sc));
+
+ SNIC_SCSI_DBG(snic->shost,
+ "itmf_cmpl:Dev Reset Completion Received after timeout. id %d cmpl status %s flags 0x%llx\n",
+ (int)(cmnd_id & SNIC_TAG_MASK),
+ snic_io_status_to_str(cmpl_stat),
+ CMD_FLAGS(sc));
+
+ return;
+ }
+
+ CMD_STATE(sc) = SNIC_IOREQ_LR_COMPLETE;
+ CMD_FLAGS(sc) |= SNIC_DEV_RST_DONE;
+
+ SNIC_SCSI_DBG(snic->shost,
+ "itmf_cmpl:Dev Reset Cmpl Recvd id %d cmpl status %s flags 0x%llx\n",
+ (int)(cmnd_id & SNIC_TAG_MASK),
+ snic_io_status_to_str(cmpl_stat),
+ CMD_FLAGS(sc));
+
+ if (rqi->dr_done)
+ complete(rqi->dr_done);
+} /* end of snic_proc_dr_cmpl_locked */
+
+/*
+ * snic_update_abort_stats : Updates abort stats based on completion status.
+ */
+static void
+snic_update_abort_stats(struct snic *snic, u8 cmpl_stat)
+{
+ struct snic_abort_stats *abt_stats = &snic->s_stats.abts;
+
+ SNIC_SCSI_DBG(snic->shost, "Updating Abort stats.\n");
+
+ switch (cmpl_stat) {
+ case SNIC_STAT_IO_SUCCESS:
+ break;
+
+ case SNIC_STAT_TIMEOUT:
+ atomic64_inc(&abt_stats->fw_tmo);
+ break;
+
+ case SNIC_STAT_IO_NOT_FOUND:
+ atomic64_inc(&abt_stats->io_not_found);
+ break;
+
+ default:
+ atomic64_inc(&abt_stats->fail);
+ break;
+ }
+}
+
+static int
+snic_process_itmf_cmpl(struct snic *snic,
+ struct snic_fw_req *fwreq,
+ u32 cmnd_id,
+ u8 cmpl_stat,
+ struct scsi_cmnd *sc)
+{
+ struct snic_req_info *rqi = NULL;
+ u32 tm_tags = 0;
+ spinlock_t *io_lock = NULL;
+ unsigned long flags;
+ u32 start_time = 0;
+ int ret = 0;
+
+ io_lock = snic_io_lock_hash(snic, sc);
+ spin_lock_irqsave(io_lock, flags);
+ rqi = (struct snic_req_info *) CMD_SP(sc);
+ WARN_ON_ONCE(!rqi);
+
+ if (!rqi) {
+ atomic64_inc(&snic->s_stats.io.req_null);
+ spin_unlock_irqrestore(io_lock, flags);
+ CMD_FLAGS(sc) |= SNIC_IO_ABTS_TERM_REQ_NULL;
+ SNIC_HOST_ERR(snic->shost,
+ "itmf_cmpl: rqi is null,Hdr stat = %s Tag = 0x%x sc = 0x%p flags 0x%llx\n",
+ snic_io_status_to_str(cmpl_stat), cmnd_id, sc,
+ CMD_FLAGS(sc));
+
+ return ret;
+ }
+
+ /* Extract task management flags */
+ tm_tags = cmnd_id & ~(SNIC_TAG_MASK);
+
+ start_time = rqi->start_time;
+ cmnd_id &= (SNIC_TAG_MASK);
+
+ switch (tm_tags) {
+ case SNIC_TAG_ABORT:
+ /* Abort only issued on cmd */
+ snic_update_abort_stats(snic, cmpl_stat);
+
+ if (CMD_STATE(sc) != SNIC_IOREQ_ABTS_PENDING) {
+ /* This is a late completion. Ignore it. */
+ ret = -1;
+ spin_unlock_irqrestore(io_lock, flags);
+ break;
+ }
+
+ CMD_STATE(sc) = SNIC_IOREQ_ABTS_COMPLETE;
+ CMD_ABTS_STATUS(sc) = cmpl_stat;
+ CMD_FLAGS(sc) |= SNIC_IO_ABTS_TERM_DONE;
+
+ SNIC_SCSI_DBG(snic->shost,
+ "itmf_cmpl:Abort Cmpl Recvd.Tag 0x%x Status %s flags 0x%llx\n",
+ cmnd_id,
+ snic_io_status_to_str(cmpl_stat),
+ CMD_FLAGS(sc));
+
+ /*
+ * If scsi_eh thread is blocked waiting for abts complete,
+ * signal completion to it. IO will be cleaned in the thread,
+ * else clean it in this context.
+ */
+ if (rqi->abts_done) {
+ complete(rqi->abts_done);
+ spin_unlock_irqrestore(io_lock, flags);
+
+ break; /* jump out */
+ }
+
+ CMD_SP(sc) = NULL;
+ sc->result = (DID_ERROR << 16);
+ SNIC_SCSI_DBG(snic->shost,
+ "itmf_cmpl: Completing IO. sc %p flags 0x%llx\n",
+ sc, CMD_FLAGS(sc));
+
+ spin_unlock_irqrestore(io_lock, flags);
+
+ snic_release_req_buf(snic, rqi, sc);
+
+ if (sc->scsi_done) {
+ SNIC_TRC(snic->shost->host_no, cmnd_id, sc,
+ jiffies_to_msecs(jiffies - start_time),
+ fwreq,
+ SNIC_TRC_CMD(sc),
+ SNIC_TRC_CMD_STATE_FLAGS(sc));
+
+ sc->scsi_done(sc);
+ }
+
+ break;
+
+ case SNIC_TAG_DEV_RST:
+ case SNIC_TAG_DEV_RST | SNIC_TAG_IOCTL_DEV_RST:
+ snic_proc_dr_cmpl_locked(snic, fwreq, cmpl_stat, cmnd_id, sc);
+ spin_unlock_irqrestore(io_lock, flags);
+ ret = 0;
+
+ break;
+
+ case SNIC_TAG_ABORT | SNIC_TAG_DEV_RST:
+ /* Abort and terminate completion of device reset req */
+
+ CMD_STATE(sc) = SNIC_IOREQ_ABTS_COMPLETE;
+ CMD_ABTS_STATUS(sc) = cmpl_stat;
+ CMD_FLAGS(sc) |= SNIC_DEV_RST_DONE;
+
+ SNIC_SCSI_DBG(snic->shost,
+ "itmf_cmpl:dev reset abts cmpl recvd. id %d status %s flags 0x%llx\n",
+ cmnd_id, snic_io_status_to_str(cmpl_stat),
+ CMD_FLAGS(sc));
+
+ if (rqi->abts_done)
+ complete(rqi->abts_done);
+
+ spin_unlock_irqrestore(io_lock, flags);
+
+ break;
+
+ default:
+ spin_unlock_irqrestore(io_lock, flags);
+ SNIC_HOST_ERR(snic->shost,
+ "itmf_cmpl: Unknown TM tag bit 0x%x\n", tm_tags);
+
+ SNIC_HOST_ERR(snic->shost,
+ "itmf_cmpl:Unexpected itmf io stat %s Tag = 0x%x flags 0x%llx\n",
+ snic_ioreq_state_to_str(CMD_STATE(sc)),
+ cmnd_id,
+ CMD_FLAGS(sc));
+ ret = -1;
+ SNIC_BUG_ON(1);
+
+ break;
+ }
+
+ return ret;
+} /* end of snic_process_itmf_cmpl_status */
+
+/*
+ * snic_itmf_cmpl_handler.
+ * Routine to handle itmf completions.
+ */
+static void
+snic_itmf_cmpl_handler(struct snic *snic, struct snic_fw_req *fwreq)
+{
+ struct scsi_cmnd *sc = NULL;
+ struct snic_req_info *rqi = NULL;
+ struct snic_itmf_cmpl *itmf_cmpl = NULL;
+ u64 ctx;
+ u32 cmnd_id;
+ u32 hid;
+ u8 typ;
+ u8 hdr_stat;
+
+ snic_io_hdr_dec(&fwreq->hdr, &typ, &hdr_stat, &cmnd_id, &hid, &ctx);
+ SNIC_SCSI_DBG(snic->shost,
+ "Itmf_cmpl: %s: type = %x, hdr_stat = %x, cmnd_id = %x, hid = %x,ctx = %llx\n",
+ __func__, typ, hdr_stat, cmnd_id, hid, ctx);
+
+ itmf_cmpl = &fwreq->u.itmf_cmpl;
+ SNIC_SCSI_DBG(snic->shost,
+ "Itmf_cmpl: nterm %u , flags 0x%x\n",
+ itmf_cmpl->nterminated, itmf_cmpl->flags);
+
+ /* spl case, dev reset issued through ioctl */
+ if (cmnd_id & SNIC_TAG_IOCTL_DEV_RST) {
+ rqi = (struct snic_req_info *) ctx;
+ sc = rqi->sc;
+
+ goto ioctl_dev_rst;
+ }
+
+ if ((cmnd_id & SNIC_TAG_MASK) >= snic->max_tag_id) {
+ SNIC_HOST_ERR(snic->shost,
+ "Itmf_cmpl: Tag 0x%x out of Range,HdrStat %s\n",
+ cmnd_id, snic_io_status_to_str(hdr_stat));
+ SNIC_BUG_ON(1);
+
+ return;
+ }
+
+ sc = scsi_host_find_tag(snic->shost, cmnd_id & SNIC_TAG_MASK);
+ WARN_ON_ONCE(!sc);
+
+ioctl_dev_rst:
+ if (!sc) {
+ atomic64_inc(&snic->s_stats.io.sc_null);
+ SNIC_HOST_ERR(snic->shost,
+ "Itmf_cmpl: sc is NULL - Hdr Stat %s Tag 0x%x\n",
+ snic_io_status_to_str(hdr_stat), cmnd_id);
+
+ return;
+ }
+
+ snic_process_itmf_cmpl(snic, fwreq, cmnd_id, hdr_stat, sc);
+} /* end of snic_itmf_cmpl_handler */
+
+
+
+static void
+snic_hba_reset_scsi_cleanup(struct snic *snic, struct scsi_cmnd *sc)
+{
+ struct snic_stats *st = &snic->s_stats;
+ long act_ios = 0, act_fwreqs = 0;
+
+ SNIC_SCSI_DBG(snic->shost, "HBA Reset scsi cleanup.\n");
+ snic_scsi_cleanup(snic, snic_cmd_tag(sc));
+
+ /* Update stats on pending IOs */
+ act_ios = atomic64_read(&st->io.active);
+ atomic64_add(act_ios, &st->io.compl);
+ atomic64_sub(act_ios, &st->io.active);
+
+ act_fwreqs = atomic64_read(&st->fw.actv_reqs);
+ atomic64_sub(act_fwreqs, &st->fw.actv_reqs);
+}
+
+/*
+ * snic_hba_reset_cmpl_handler :
+ *
+ * Notes :
+ * 1. Cleanup all the scsi cmds, release all snic specific cmds
+ * 2. Issue Report Targets in case of SAN targets
+ */
+static int
+snic_hba_reset_cmpl_handler(struct snic *snic, struct snic_fw_req *fwreq)
+{
+ u64 ctx;
+ u32 cmnd_id;
+ u32 hid;
+ u8 typ;
+ u8 hdr_stat;
+ struct scsi_cmnd *sc = NULL;
+ struct snic_req_info *rqi = NULL;
+ spinlock_t *io_lock = NULL;
+ unsigned long flags, gflags;
+ int ret = 0;
+
+ SNIC_HOST_INFO(snic->shost,
+ "reset_cmpl:HBA Reset Completion received.\n");
+
+ snic_io_hdr_dec(&fwreq->hdr, &typ, &hdr_stat, &cmnd_id, &hid, &ctx);
+ SNIC_SCSI_DBG(snic->shost,
+ "reset_cmpl: type = %x, hdr_stat = %x, cmnd_id = %x, hid = %x, ctx = %llx\n",
+ typ, hdr_stat, cmnd_id, hid, ctx);
+
+ /* spl case, host reset issued through ioctl */
+ if (cmnd_id == SCSI_NO_TAG) {
+ rqi = (struct snic_req_info *) ctx;
+ sc = rqi->sc;
+
+ goto ioctl_hba_rst;
+ }
+
+ if (cmnd_id >= snic->max_tag_id) {
+ SNIC_HOST_ERR(snic->shost,
+ "reset_cmpl: Tag 0x%x out of Range,HdrStat %s\n",
+ cmnd_id, snic_io_status_to_str(hdr_stat));
+ SNIC_BUG_ON(1);
+
+ return 1;
+ }
+
+ sc = scsi_host_find_tag(snic->shost, cmnd_id);
+ioctl_hba_rst:
+ if (!sc) {
+ atomic64_inc(&snic->s_stats.io.sc_null);
+ SNIC_HOST_ERR(snic->shost,
+ "reset_cmpl: sc is NULL - Hdr Stat %s Tag 0x%x\n",
+ snic_io_status_to_str(hdr_stat), cmnd_id);
+ ret = 1;
+
+ return ret;
+ }
+
+ io_lock = snic_io_lock_hash(snic, sc);
+ spin_lock_irqsave(io_lock, flags);
+
+ if (!snic->remove_wait) {
+ spin_unlock_irqrestore(io_lock, flags);
+ SNIC_HOST_ERR(snic->shost,
+ "reset_cmpl:host reset completed after timout\n");
+ ret = 1;
+
+ return ret;
+ }
+
+ rqi = (struct snic_req_info *) CMD_SP(sc);
+ WARN_ON_ONCE(!rqi);
+
+ if (!rqi) {
+ atomic64_inc(&snic->s_stats.io.req_null);
+ spin_unlock_irqrestore(io_lock, flags);
+ CMD_FLAGS(sc) |= SNIC_IO_ABTS_TERM_REQ_NULL;
+ SNIC_HOST_ERR(snic->shost,
+ "reset_cmpl: rqi is null,Hdr stat %s Tag 0x%x sc 0x%p flags 0x%llx\n",
+ snic_io_status_to_str(hdr_stat), cmnd_id, sc,
+ CMD_FLAGS(sc));
+
+ ret = 1;
+
+ return ret;
+ }
+ /* stats */
+ spin_unlock_irqrestore(io_lock, flags);
+
+ /* scsi cleanup */
+ snic_hba_reset_scsi_cleanup(snic, sc);
+
+ SNIC_BUG_ON(snic_get_state(snic) != SNIC_OFFLINE &&
+ snic_get_state(snic) != SNIC_FWRESET);
+
+ /* Careful locking between snic_lock and io lock */
+ spin_lock_irqsave(io_lock, flags);
+ spin_lock_irqsave(&snic->snic_lock, gflags);
+ if (snic_get_state(snic) == SNIC_FWRESET)
+ snic_set_state(snic, SNIC_ONLINE);
+ spin_unlock_irqrestore(&snic->snic_lock, gflags);
+
+ if (snic->remove_wait)
+ complete(snic->remove_wait);
+
+ spin_unlock_irqrestore(io_lock, flags);
+ atomic64_inc(&snic->s_stats.reset.hba_reset_cmpl);
+
+ ret = 0;
+ /* Rediscovery is for SAN */
+ if (snic->config.xpt_type == SNIC_DAS)
+ return ret;
+
+ SNIC_SCSI_DBG(snic->shost, "reset_cmpl: Queuing discovery work.\n");
+ queue_work(snic_glob->event_q, &snic->disc_work);
+
+ return ret;
+}
+
+static void
+snic_msg_ack_handler(struct snic *snic, struct snic_fw_req *fwreq)
+{
+ SNIC_HOST_INFO(snic->shost, "Message Ack Received.\n");
+
+ SNIC_ASSERT_NOT_IMPL(1);
+}
+
+static void
+snic_aen_handler(struct snic *snic, struct snic_fw_req *fwreq)
+{
+ u8 typ, hdr_stat;
+ u32 cmnd_id, hid;
+ u64 ctx;
+ struct snic_async_evnotify *aen = &fwreq->u.async_ev;
+
+ snic_io_hdr_dec(&fwreq->hdr, &typ, &hdr_stat, &cmnd_id, &hid, &ctx);
+ SNIC_SCSI_DBG(snic->shost,
+ "aen: type = %x, hdr_stat = %x, cmnd_id = %x, hid = %x, ctx = %llx\n",
+ typ, hdr_stat, cmnd_id, hid, ctx);
+
+ switch (aen->ev_id) {
+ case SNIC_EV_TGT_OFFLINE:
+ SNIC_HOST_INFO(snic->shost, "aen:TGT_OFFLINE Event Recvd.\n");
+ break;
+
+ case SNIC_EV_TGT_ONLINE:
+ SNIC_HOST_INFO(snic->shost, "aen:TGT_ONLINE Event Recvd.\n");
+ break;
+
+ case SNIC_EV_LUN_OFFLINE:
+ SNIC_HOST_INFO(snic->shost, "aen:LUN_OFFLINE Event Recvd.\n");
+ break;
+
+ case SNIC_EV_LUN_ONLINE:
+ SNIC_HOST_INFO(snic->shost, "aen:LUN_ONLINE Event Recvd.\n");
+ break;
+
+ case SNIC_EV_CONF_CHG:
+ SNIC_HOST_INFO(snic->shost, "aen:Config Change Event Recvd.\n");
+ break;
+
+ case SNIC_EV_TGT_ADDED:
+ SNIC_HOST_INFO(snic->shost, "aen:TGT_ADD Event Recvd.\n");
+ break;
+
+ case SNIC_EV_TGT_DELTD:
+ SNIC_HOST_INFO(snic->shost, "aen:TGT_DEL Event Recvd.\n");
+ break;
+
+ case SNIC_EV_LUN_ADDED:
+ SNIC_HOST_INFO(snic->shost, "aen:LUN_ADD Event Recvd.\n");
+ break;
+
+ case SNIC_EV_LUN_DELTD:
+ SNIC_HOST_INFO(snic->shost, "aen:LUN_DEL Event Recvd.\n");
+ break;
+
+ case SNIC_EV_DISC_CMPL:
+ SNIC_HOST_INFO(snic->shost, "aen:DISC_CMPL Event Recvd.\n");
+ break;
+
+ default:
+ SNIC_HOST_INFO(snic->shost, "aen:Unknown Event Recvd.\n");
+ SNIC_BUG_ON(1);
+ break;
+ }
+
+ SNIC_ASSERT_NOT_IMPL(1);
+} /* end of snic_aen_handler */
+
+/*
+ * snic_io_cmpl_handler
+ * Routine to process CQ entries(IO Completions) posted by fw.
+ */
+static int
+snic_io_cmpl_handler(struct vnic_dev *vdev,
+ unsigned int cq_idx,
+ struct snic_fw_req *fwreq)
+{
+ struct snic *snic = vnic_dev_priv(vdev);
+ u64 start = jiffies, cmpl_time;
+
+ snic_print_desc(__func__, (char *)fwreq, sizeof(*fwreq));
+
+ /* Update FW Stats */
+ if ((fwreq->hdr.type >= SNIC_RSP_REPORT_TGTS_CMPL) &&
+ (fwreq->hdr.type <= SNIC_RSP_BOOT_LUNS_CMPL))
+ atomic64_dec(&snic->s_stats.fw.actv_reqs);
+
+ SNIC_BUG_ON((fwreq->hdr.type > SNIC_RSP_BOOT_LUNS_CMPL) &&
+ (fwreq->hdr.type < SNIC_MSG_ASYNC_EVNOTIFY));
+
+ /* Check for snic subsys errors */
+ switch (fwreq->hdr.status) {
+ case SNIC_STAT_NOT_READY: /* XPT yet to initialize */
+ SNIC_HOST_ERR(snic->shost,
+ "sNIC SubSystem is NOT Ready.\n");
+ break;
+
+ case SNIC_STAT_FATAL_ERROR: /* XPT Error */
+ SNIC_HOST_ERR(snic->shost,
+ "sNIC SubSystem in Unrecoverable State.\n");
+ break;
+ }
+
+ switch (fwreq->hdr.type) {
+ case SNIC_RSP_EXCH_VER_CMPL:
+ snic_io_exch_ver_cmpl_handler(snic, fwreq);
+ break;
+
+ case SNIC_RSP_REPORT_TGTS_CMPL:
+ snic_report_tgt_cmpl_handler(snic, fwreq);
+ break;
+
+ case SNIC_RSP_ICMND_CMPL:
+ snic_icmnd_cmpl_handler(snic, fwreq);
+ break;
+
+ case SNIC_RSP_ITMF_CMPL:
+ snic_itmf_cmpl_handler(snic, fwreq);
+ break;
+
+ case SNIC_RSP_HBA_RESET_CMPL:
+ snic_hba_reset_cmpl_handler(snic, fwreq);
+ break;
+
+ case SNIC_MSG_ACK:
+ snic_msg_ack_handler(snic, fwreq);
+ break;
+
+ case SNIC_MSG_ASYNC_EVNOTIFY:
+ snic_aen_handler(snic, fwreq);
+ break;
+
+ default:
+ SNIC_BUG_ON(1);
+ SNIC_SCSI_DBG(snic->shost,
+ "Unknown Firmwqre completion request type %d\n",
+ fwreq->hdr.type);
+ break;
+ }
+
+ /* Update Stats */
+ cmpl_time = jiffies - start;
+ if (cmpl_time > atomic64_read(&snic->s_stats.io.max_cmpl_time))
+ atomic64_set(&snic->s_stats.io.max_cmpl_time, cmpl_time);
+
+ return 0;
+} /* end of snic_io_cmpl_handler */
+
+/*
+ * snic_fwcq_cmpl_handler
+ * Routine to process fwCQ
+ * This CQ is independent, and not associated with wq/rq/wq_copy queues
+ */
+int
+snic_fwcq_cmpl_handler(struct snic *snic, int io_cmpl_work)
+{
+ unsigned int num_ent = 0; /* number cq entries processed */
+ unsigned int cq_idx;
+ unsigned int nent_per_cq;
+ struct snic_misc_stats *misc_stats = &snic->s_stats.misc;
+
+ for (cq_idx = snic->wq_count; cq_idx < snic->cq_count; cq_idx++) {
+ nent_per_cq = vnic_cq_fw_service(&snic->cq[cq_idx],
+ snic_io_cmpl_handler,
+ io_cmpl_work);
+ num_ent += nent_per_cq;
+
+ if (nent_per_cq > atomic64_read(&misc_stats->max_cq_ents))
+ atomic64_set(&misc_stats->max_cq_ents, nent_per_cq);
+ }
+
+ return num_ent;
+} /* end of snic_fwcq_cmpl_handler */
+
+/*
+ * snic_queue_itmf_req: Common API to queue Task Management requests.
+ * Use rqi->tm_tag for passing special tags.
+ * @req_id : aborted request's tag, -1 for lun reset.
+ */
+static int
+snic_queue_itmf_req(struct snic *snic,
+ struct snic_host_req *tmreq,
+ struct scsi_cmnd *sc,
+ u32 tmf,
+ u32 req_id)
+{
+ struct snic_req_info *rqi = req_to_rqi(tmreq);
+ struct scsi_lun lun;
+ int tm_tag = snic_cmd_tag(sc) | rqi->tm_tag;
+ int ret = 0;
+
+ SNIC_BUG_ON(!rqi);
+ SNIC_BUG_ON(!rqi->tm_tag);
+
+ /* fill in lun info */
+ int_to_scsilun(sc->device->lun, &lun);
+
+ /* Initialize snic_host_req: itmf */
+ snic_itmf_init(tmreq,
+ tm_tag,
+ snic->config.hid,
+ (u64) rqi,
+ 0 /* flags */,
+ req_id, /* Command to be aborted. */
+ rqi->tgt_id,
+ lun.scsi_lun,
+ tmf);
+
+ /*
+ * In case of multiple aborts on same cmd,
+ * use try_wait_for_completion and completion_done() to check
+ * whether it queues aborts even after completion of abort issued
+ * prior.SNIC_BUG_ON(completion_done(&rqi->done));
+ */
+
+ ret = snic_queue_wq_desc(snic, tmreq, sizeof(*tmreq));
+ if (ret)
+ SNIC_HOST_ERR(snic->shost,
+ "qitmf:Queuing ITMF(%d) Req sc %p, rqi %p, req_id %d tag %d Failed, ret = %d\n",
+ tmf, sc, rqi, req_id, snic_cmd_tag(sc), ret);
+ else
+ SNIC_SCSI_DBG(snic->shost,
+ "qitmf:Queuing ITMF(%d) Req sc %p, rqi %p, req_id %d, tag %d (req_id)- Success.",
+ tmf, sc, rqi, req_id, snic_cmd_tag(sc));
+
+ return ret;
+} /* end of snic_queue_itmf_req */
+
+static int
+snic_issue_tm_req(struct snic *snic,
+ struct snic_req_info *rqi,
+ struct scsi_cmnd *sc,
+ int tmf)
+{
+ struct snic_host_req *tmreq = NULL;
+ int req_id = 0, tag = snic_cmd_tag(sc);
+ int ret = 0;
+
+ if (snic_get_state(snic) == SNIC_FWRESET)
+ return -EBUSY;
+
+ atomic_inc(&snic->ios_inflight);
+
+ SNIC_SCSI_DBG(snic->shost,
+ "issu_tmreq: Task mgmt req %d. rqi %p w/ tag %x\n",
+ tmf, rqi, tag);
+
+
+ if (tmf == SNIC_ITMF_LUN_RESET) {
+ tmreq = snic_dr_req_init(snic, rqi);
+ req_id = SCSI_NO_TAG;
+ } else {
+ tmreq = snic_abort_req_init(snic, rqi);
+ req_id = tag;
+ }
+
+ if (!tmreq) {
+ ret = -ENOMEM;
+
+ goto tmreq_err;
+ }
+
+ ret = snic_queue_itmf_req(snic, tmreq, sc, tmf, req_id);
+ if (ret)
+ goto tmreq_err;
+
+ ret = 0;
+
+tmreq_err:
+ if (ret) {
+ SNIC_HOST_ERR(snic->shost,
+ "issu_tmreq: Queing ITMF(%d) Req, sc %p rqi %p req_id %d tag %x fails err = %d\n",
+ tmf, sc, rqi, req_id, tag, ret);
+ } else {
+ SNIC_SCSI_DBG(snic->shost,
+ "issu_tmreq: Queuing ITMF(%d) Req, sc %p, rqi %p, req_id %d tag %x - Success.\n",
+ tmf, sc, rqi, req_id, tag);
+ }
+
+ atomic_dec(&snic->ios_inflight);
+
+ return ret;
+}
+
+/*
+ * snic_queue_abort_req : Queues abort req to WQ
+ */
+static int
+snic_queue_abort_req(struct snic *snic,
+ struct snic_req_info *rqi,
+ struct scsi_cmnd *sc,
+ int tmf)
+{
+ SNIC_SCSI_DBG(snic->shost, "q_abtreq: sc %p, rqi %p, tag %x, tmf %d\n",
+ sc, rqi, snic_cmd_tag(sc), tmf);
+
+ /* Add special tag for abort */
+ rqi->tm_tag |= SNIC_TAG_ABORT;
+
+ return snic_issue_tm_req(snic, rqi, sc, tmf);
+}
+
+/*
+ * snic_abort_finish : called by snic_abort_cmd on queuing abort successfully.
+ */
+static int
+snic_abort_finish(struct snic *snic, struct scsi_cmnd *sc)
+{
+ struct snic_req_info *rqi = NULL;
+ spinlock_t *io_lock = NULL;
+ unsigned long flags;
+ int ret = 0, tag = snic_cmd_tag(sc);
+
+ io_lock = snic_io_lock_hash(snic, sc);
+ spin_lock_irqsave(io_lock, flags);
+ rqi = (struct snic_req_info *) CMD_SP(sc);
+ if (!rqi) {
+ atomic64_inc(&snic->s_stats.io.req_null);
+ CMD_FLAGS(sc) |= SNIC_IO_ABTS_TERM_REQ_NULL;
+
+ SNIC_SCSI_DBG(snic->shost,
+ "abt_fini:req info is null tag 0x%x, sc 0x%p flags 0x%llx\n",
+ tag, sc, CMD_FLAGS(sc));
+ ret = FAILED;
+
+ goto abort_fail;
+ }
+
+ rqi->abts_done = NULL;
+
+ ret = FAILED;
+
+ /* Check the abort status. */
+ switch (CMD_ABTS_STATUS(sc)) {
+ case SNIC_INVALID_CODE:
+ /* Firmware didn't complete abort req, timedout */
+ CMD_FLAGS(sc) |= SNIC_IO_ABTS_TIMEDOUT;
+ atomic64_inc(&snic->s_stats.abts.drv_tmo);
+ SNIC_SCSI_DBG(snic->shost,
+ "abt_fini:sc %p Tag %x Driver Timeout.flags 0x%llx\n",
+ sc, snic_cmd_tag(sc), CMD_FLAGS(sc));
+ /* do not release snic request in timedout case */
+ rqi = NULL;
+
+ goto abort_fail;
+
+ case SNIC_STAT_IO_SUCCESS:
+ case SNIC_STAT_IO_NOT_FOUND:
+ ret = SUCCESS;
+ break;
+
+ default:
+ /* Firmware completed abort with error */
+ ret = FAILED;
+ break;
+ }
+
+ CMD_SP(sc) = NULL;
+ SNIC_HOST_INFO(snic->shost,
+ "abt_fini: Tag %x, Cmpl Status %s flags 0x%llx\n",
+ tag, snic_io_status_to_str(CMD_ABTS_STATUS(sc)),
+ CMD_FLAGS(sc));
+
+abort_fail:
+ spin_unlock_irqrestore(io_lock, flags);
+ if (rqi)
+ snic_release_req_buf(snic, rqi, sc);
+
+ return ret;
+} /* end of snic_abort_finish */
+
+/*
+ * snic_send_abort_and_wait : Issues Abort, and Waits
+ */
+static int
+snic_send_abort_and_wait(struct snic *snic, struct scsi_cmnd *sc)
+{
+ struct snic_req_info *rqi = NULL;
+ enum snic_ioreq_state sv_state;
+ struct snic_tgt *tgt = NULL;
+ spinlock_t *io_lock = NULL;
+ DECLARE_COMPLETION_ONSTACK(tm_done);
+ unsigned long flags;
+ int ret = 0, tmf = 0, tag = snic_cmd_tag(sc);
+
+ tgt = starget_to_tgt(scsi_target(sc->device));
+ if ((snic_tgt_chkready(tgt) != 0) && (tgt->tdata.typ == SNIC_TGT_SAN))
+ tmf = SNIC_ITMF_ABTS_TASK_TERM;
+ else
+ tmf = SNIC_ITMF_ABTS_TASK;
+
+ /* stats */
+
+ io_lock = snic_io_lock_hash(snic, sc);
+
+ /*
+ * Avoid a race between SCSI issuing the abort and the device
+ * completing the command.
+ *
+ * If the command is already completed by fw_cmpl code,
+ * we just return SUCCESS from here. This means that the abort
+ * succeeded. In the SCSI ML, since the timeout for command has
+ * happend, the completion wont actually complete the command
+ * and it will be considered as an aborted command
+ *
+ * The CMD_SP will not be cleared except while holding io_lock
+ */
+ spin_lock_irqsave(io_lock, flags);
+ rqi = (struct snic_req_info *) CMD_SP(sc);
+ if (!rqi) {
+ spin_unlock_irqrestore(io_lock, flags);
+
+ SNIC_HOST_ERR(snic->shost,
+ "abt_cmd: rqi is null. Tag %d flags 0x%llx\n",
+ tag, CMD_FLAGS(sc));
+
+ ret = SUCCESS;
+
+ goto send_abts_end;
+ }
+
+ rqi->abts_done = &tm_done;
+ if (CMD_STATE(sc) == SNIC_IOREQ_ABTS_PENDING) {
+ spin_unlock_irqrestore(io_lock, flags);
+
+ ret = 0;
+ goto abts_pending;
+ }
+ SNIC_BUG_ON(!rqi->abts_done);
+
+ /* Save Command State, should be restored on failed to Queue. */
+ sv_state = CMD_STATE(sc);
+
+ /*
+ * Command is still pending, need to abort it
+ * If the fw completes the command after this point,
+ * the completion won't be done till mid-layer, since abot
+ * has already started.
+ */
+ CMD_STATE(sc) = SNIC_IOREQ_ABTS_PENDING;
+ CMD_ABTS_STATUS(sc) = SNIC_INVALID_CODE;
+
+ SNIC_SCSI_DBG(snic->shost, "send_abt_cmd: TAG 0x%x\n", tag);
+
+ spin_unlock_irqrestore(io_lock, flags);
+
+ /* Now Queue the abort command to firmware */
+ ret = snic_queue_abort_req(snic, rqi, sc, tmf);
+ if (ret) {
+ SNIC_HOST_ERR(snic->shost,
+ "send_abt_cmd: IO w/ Tag 0x%x fail w/ err %d flags 0x%llx\n",
+ tag, ret, CMD_FLAGS(sc));
+
+ spin_lock_irqsave(io_lock, flags);
+ /* Restore Command's previous state */
+ CMD_STATE(sc) = sv_state;
+ rqi = (struct snic_req_info *) CMD_SP(sc);
+ if (rqi)
+ rqi->abts_done = NULL;
+ spin_unlock_irqrestore(io_lock, flags);
+ ret = FAILED;
+
+ goto send_abts_end;
+ }
+
+ spin_lock_irqsave(io_lock, flags);
+ if (tmf == SNIC_ITMF_ABTS_TASK) {
+ CMD_FLAGS(sc) |= SNIC_IO_ABTS_ISSUED;
+ atomic64_inc(&snic->s_stats.abts.num);
+ } else {
+ /* term stats */
+ CMD_FLAGS(sc) |= SNIC_IO_TERM_ISSUED;
+ }
+ spin_unlock_irqrestore(io_lock, flags);
+
+ SNIC_SCSI_DBG(snic->shost,
+ "send_abt_cmd: sc %p Tag %x flags 0x%llx\n",
+ sc, tag, CMD_FLAGS(sc));
+
+
+ ret = 0;
+
+abts_pending:
+ /*
+ * Queued an abort IO, wait for its completion.
+ * Once the fw completes the abort command, it will
+ * wakeup this thread.
+ */
+ wait_for_completion_timeout(&tm_done, SNIC_ABTS_TIMEOUT);
+
+send_abts_end:
+ return ret;
+} /* end of snic_send_abort_and_wait */
+
+/*
+ * This function is exported to SCSI for sending abort cmnds.
+ * A SCSI IO is represent by snic_ioreq in the driver.
+ * The snic_ioreq is linked to the SCSI Cmd, thus a link with the ULP'S IO
+ */
+int
+snic_abort_cmd(struct scsi_cmnd *sc)
+{
+ struct snic *snic = shost_priv(sc->device->host);
+ int ret = SUCCESS, tag = snic_cmd_tag(sc);
+ u32 start_time = jiffies;
+
+ SNIC_SCSI_DBG(snic->shost, "abt_cmd:sc %p :0x%x :req = %p :tag = %d\n",
+ sc, sc->cmnd[0], sc->request, tag);
+
+ if (unlikely(snic_get_state(snic) != SNIC_ONLINE)) {
+ SNIC_HOST_ERR(snic->shost,
+ "abt_cmd: tag %x Parent Devs are not rdy\n",
+ tag);
+ ret = FAST_IO_FAIL;
+
+ goto abort_end;
+ }
+
+
+ ret = snic_send_abort_and_wait(snic, sc);
+ if (ret)
+ goto abort_end;
+
+ ret = snic_abort_finish(snic, sc);
+
+abort_end:
+ SNIC_TRC(snic->shost->host_no, tag, sc,
+ jiffies_to_msecs(jiffies - start_time), 0,
+ SNIC_TRC_CMD(sc), SNIC_TRC_CMD_STATE_FLAGS(sc));
+
+ SNIC_SCSI_DBG(snic->shost,
+ "abts: Abort Req Status = %s\n",
+ (ret == SUCCESS) ? "SUCCESS" :
+ ((ret == FAST_IO_FAIL) ? "FAST_IO_FAIL" : "FAILED"));
+
+ return ret;
+}
+
+
+
+static int
+snic_is_abts_pending(struct snic *snic, struct scsi_cmnd *lr_sc)
+{
+ struct snic_req_info *rqi = NULL;
+ struct scsi_cmnd *sc = NULL;
+ struct scsi_device *lr_sdev = NULL;
+ spinlock_t *io_lock = NULL;
+ u32 tag;
+ unsigned long flags;
+
+ if (lr_sc)
+ lr_sdev = lr_sc->device;
+
+ /* walk through the tag map, an dcheck if IOs are still pending in fw*/
+ for (tag = 0; tag < snic->max_tag_id; tag++) {
+ io_lock = snic_io_lock_tag(snic, tag);
+
+ spin_lock_irqsave(io_lock, flags);
+ sc = scsi_host_find_tag(snic->shost, tag);
+
+ if (!sc || (lr_sc && (sc->device != lr_sdev || sc == lr_sc))) {
+ spin_unlock_irqrestore(io_lock, flags);
+
+ continue;
+ }
+
+ rqi = (struct snic_req_info *) CMD_SP(sc);
+ if (!rqi) {
+ spin_unlock_irqrestore(io_lock, flags);
+
+ continue;
+ }
+
+ /*
+ * Found IO that is still pending w/ firmware and belongs to
+ * the LUN that is under reset, if lr_sc != NULL
+ */
+ SNIC_SCSI_DBG(snic->shost, "Found IO in %s on LUN\n",
+ snic_ioreq_state_to_str(CMD_STATE(sc)));
+
+ if (CMD_STATE(sc) == SNIC_IOREQ_ABTS_PENDING) {
+ spin_unlock_irqrestore(io_lock, flags);
+
+ return 1;
+ }
+
+ spin_unlock_irqrestore(io_lock, flags);
+ }
+
+ return 0;
+} /* end of snic_is_abts_pending */
+
+static int
+snic_dr_clean_single_req(struct snic *snic,
+ u32 tag,
+ struct scsi_device *lr_sdev)
+{
+ struct snic_req_info *rqi = NULL;
+ struct snic_tgt *tgt = NULL;
+ struct scsi_cmnd *sc = NULL;
+ spinlock_t *io_lock = NULL;
+ u32 sv_state = 0, tmf = 0;
+ DECLARE_COMPLETION_ONSTACK(tm_done);
+ unsigned long flags;
+ int ret = 0;
+
+ io_lock = snic_io_lock_tag(snic, tag);
+ spin_lock_irqsave(io_lock, flags);
+ sc = scsi_host_find_tag(snic->shost, tag);
+
+ /* Ignore Cmd that don't belong to Lun Reset device */
+ if (!sc || sc->device != lr_sdev)
+ goto skip_clean;
+
+ rqi = (struct snic_req_info *) CMD_SP(sc);
+
+ if (!rqi)
+ goto skip_clean;
+
+
+ if (CMD_STATE(sc) == SNIC_IOREQ_ABTS_PENDING)
+ goto skip_clean;
+
+
+ if ((CMD_FLAGS(sc) & SNIC_DEVICE_RESET) &&
+ (!(CMD_FLAGS(sc) & SNIC_DEV_RST_ISSUED))) {
+
+ SNIC_SCSI_DBG(snic->shost,
+ "clean_single_req: devrst is not pending sc 0x%p\n",
+ sc);
+
+ goto skip_clean;
+ }
+
+ SNIC_SCSI_DBG(snic->shost,
+ "clean_single_req: Found IO in %s on lun\n",
+ snic_ioreq_state_to_str(CMD_STATE(sc)));
+
+ /* Save Command State */
+ sv_state = CMD_STATE(sc);
+
+ /*
+ * Any pending IO issued prior to reset is expected to be
+ * in abts pending state, if not we need to set SNIC_IOREQ_ABTS_PENDING
+ * to indicate the IO is abort pending.
+ * When IO is completed, the IO will be handed over and handled
+ * in this function.
+ */
+
+ CMD_STATE(sc) = SNIC_IOREQ_ABTS_PENDING;
+ SNIC_BUG_ON(rqi->abts_done);
+
+ if (CMD_FLAGS(sc) & SNIC_DEVICE_RESET) {
+ rqi->tm_tag = SNIC_TAG_DEV_RST;
+
+ SNIC_SCSI_DBG(snic->shost,
+ "clean_single_req:devrst sc 0x%p\n", sc);
+ }
+
+ CMD_ABTS_STATUS(sc) = SNIC_INVALID_CODE;
+ rqi->abts_done = &tm_done;
+ spin_unlock_irqrestore(io_lock, flags);
+
+ tgt = starget_to_tgt(scsi_target(sc->device));
+ if ((snic_tgt_chkready(tgt) != 0) && (tgt->tdata.typ == SNIC_TGT_SAN))
+ tmf = SNIC_ITMF_ABTS_TASK_TERM;
+ else
+ tmf = SNIC_ITMF_ABTS_TASK;
+
+ /* Now queue the abort command to firmware */
+ ret = snic_queue_abort_req(snic, rqi, sc, tmf);
+ if (ret) {
+ SNIC_HOST_ERR(snic->shost,
+ "clean_single_req_err:sc %p, tag %d abt failed. tm_tag %d flags 0x%llx\n",
+ sc, tag, rqi->tm_tag, CMD_FLAGS(sc));
+
+ spin_lock_irqsave(io_lock, flags);
+ rqi = (struct snic_req_info *) CMD_SP(sc);
+ if (rqi)
+ rqi->abts_done = NULL;
+
+ /* Restore Command State */
+ if (CMD_STATE(sc) == SNIC_IOREQ_ABTS_PENDING)
+ CMD_STATE(sc) = sv_state;
+
+ ret = 1;
+ goto skip_clean;
+ }
+
+ spin_lock_irqsave(io_lock, flags);
+ if (CMD_FLAGS(sc) & SNIC_DEVICE_RESET)
+ CMD_FLAGS(sc) |= SNIC_DEV_RST_TERM_ISSUED;
+
+ CMD_FLAGS(sc) |= SNIC_IO_INTERNAL_TERM_ISSUED;
+ spin_unlock_irqrestore(io_lock, flags);
+
+ wait_for_completion_timeout(&tm_done, SNIC_ABTS_TIMEOUT);
+
+ /* Recheck cmd state to check if it now aborted. */
+ spin_lock_irqsave(io_lock, flags);
+ rqi = (struct snic_req_info *) CMD_SP(sc);
+ if (!rqi) {
+ CMD_FLAGS(sc) |= SNIC_IO_ABTS_TERM_REQ_NULL;
+ goto skip_clean;
+ }
+ rqi->abts_done = NULL;
+
+ /* if abort is still pending w/ fw, fail */
+ if (CMD_ABTS_STATUS(sc) == SNIC_INVALID_CODE) {
+ SNIC_HOST_ERR(snic->shost,
+ "clean_single_req_err:sc %p tag %d abt still pending w/ fw, tm_tag %d flags 0x%llx\n",
+ sc, tag, rqi->tm_tag, CMD_FLAGS(sc));
+
+ CMD_FLAGS(sc) |= SNIC_IO_ABTS_TERM_DONE;
+ ret = 1;
+
+ goto skip_clean;
+ }
+
+ CMD_STATE(sc) = SNIC_IOREQ_ABTS_COMPLETE;
+ CMD_SP(sc) = NULL;
+ spin_unlock_irqrestore(io_lock, flags);
+
+ snic_release_req_buf(snic, rqi, sc);
+
+ ret = 0;
+
+ return ret;
+
+skip_clean:
+ spin_unlock_irqrestore(io_lock, flags);
+
+ return ret;
+} /* end of snic_dr_clean_single_req */
+
+static int
+snic_dr_clean_pending_req(struct snic *snic, struct scsi_cmnd *lr_sc)
+{
+ struct scsi_device *lr_sdev = lr_sc->device;
+ u32 tag = 0;
+ int ret = FAILED;
+
+ for (tag = 0; tag < snic->max_tag_id; tag++) {
+ if (tag == snic_cmd_tag(lr_sc))
+ continue;
+
+ ret = snic_dr_clean_single_req(snic, tag, lr_sdev);
+ if (ret) {
+ SNIC_HOST_ERR(snic->shost, "clean_err:tag = %d\n", tag);
+
+ goto clean_err;
+ }
+ }
+
+ schedule_timeout(msecs_to_jiffies(100));
+
+ /* Walk through all the cmds and check abts status. */
+ if (snic_is_abts_pending(snic, lr_sc)) {
+ ret = FAILED;
+
+ goto clean_err;
+ }
+
+ ret = 0;
+ SNIC_SCSI_DBG(snic->shost, "clean_pending_req: Success.\n");
+
+ return ret;
+
+clean_err:
+ ret = FAILED;
+ SNIC_HOST_ERR(snic->shost,
+ "Failed to Clean Pending IOs on %s device.\n",
+ dev_name(&lr_sdev->sdev_gendev));
+
+ return ret;
+
+} /* end of snic_dr_clean_pending_req */
+
+/*
+ * snic_dr_finish : Called by snic_device_reset
+ */
+static int
+snic_dr_finish(struct snic *snic, struct scsi_cmnd *sc)
+{
+ struct snic_req_info *rqi = NULL;
+ spinlock_t *io_lock = NULL;
+ unsigned long flags;
+ int lr_res = 0;
+ int ret = FAILED;
+
+ io_lock = snic_io_lock_hash(snic, sc);
+ spin_lock_irqsave(io_lock, flags);
+ rqi = (struct snic_req_info *) CMD_SP(sc);
+ if (!rqi) {
+ spin_unlock_irqrestore(io_lock, flags);
+ SNIC_SCSI_DBG(snic->shost,
+ "dr_fini: rqi is null tag 0x%x sc 0x%p flags 0x%llx\n",
+ snic_cmd_tag(sc), sc, CMD_FLAGS(sc));
+
+ ret = FAILED;
+ goto dr_fini_end;
+ }
+
+ rqi->dr_done = NULL;
+
+ lr_res = CMD_LR_STATUS(sc);
+
+ switch (lr_res) {
+ case SNIC_INVALID_CODE:
+ /* stats */
+ SNIC_SCSI_DBG(snic->shost,
+ "dr_fini: Tag %x Dev Reset Timedout. flags 0x%llx\n",
+ snic_cmd_tag(sc), CMD_FLAGS(sc));
+
+ CMD_FLAGS(sc) |= SNIC_DEV_RST_TIMEDOUT;
+ ret = FAILED;
+
+ goto dr_failed;
+
+ case SNIC_STAT_IO_SUCCESS:
+ SNIC_SCSI_DBG(snic->shost,
+ "dr_fini: Tag %x Dev Reset cmpl\n",
+ snic_cmd_tag(sc));
+ ret = 0;
+ break;
+
+ default:
+ SNIC_HOST_ERR(snic->shost,
+ "dr_fini:Device Reset completed& failed.Tag = %x lr_status %s flags 0x%llx\n",
+ snic_cmd_tag(sc),
+ snic_io_status_to_str(lr_res), CMD_FLAGS(sc));
+ ret = FAILED;
+ goto dr_failed;
+ }
+ spin_unlock_irqrestore(io_lock, flags);
+
+ /*
+ * Cleanup any IOs on this LUN that have still not completed.
+ * If any of these fail, then LUN Reset fails.
+ * Cleanup cleans all commands on this LUN except
+ * the lun reset command. If all cmds get cleaned, the LUN Reset
+ * succeeds.
+ */
+
+ ret = snic_dr_clean_pending_req(snic, sc);
+ if (ret) {
+ spin_lock_irqsave(io_lock, flags);
+ SNIC_SCSI_DBG(snic->shost,
+ "dr_fini: Device Reset Failed since could not abort all IOs. Tag = %x.\n",
+ snic_cmd_tag(sc));
+ rqi = (struct snic_req_info *) CMD_SP(sc);
+
+ goto dr_failed;
+ } else {
+ /* Cleanup LUN Reset Command */
+ spin_lock_irqsave(io_lock, flags);
+ rqi = (struct snic_req_info *) CMD_SP(sc);
+ if (rqi)
+ ret = SUCCESS; /* Completed Successfully */
+ else
+ ret = FAILED;
+ }
+
+dr_failed:
+ SNIC_BUG_ON(!spin_is_locked(io_lock));
+ if (rqi)
+ CMD_SP(sc) = NULL;
+ spin_unlock_irqrestore(io_lock, flags);
+
+ if (rqi)
+ snic_release_req_buf(snic, rqi, sc);
+
+dr_fini_end:
+ return ret;
+} /* end of snic_dr_finish */
+
+static int
+snic_queue_dr_req(struct snic *snic,
+ struct snic_req_info *rqi,
+ struct scsi_cmnd *sc)
+{
+ /* Add special tag for device reset */
+ rqi->tm_tag |= SNIC_TAG_DEV_RST;
+
+ return snic_issue_tm_req(snic, rqi, sc, SNIC_ITMF_LUN_RESET);
+}
+
+static int
+snic_send_dr_and_wait(struct snic *snic, struct scsi_cmnd *sc)
+{
+ struct snic_req_info *rqi = NULL;
+ enum snic_ioreq_state sv_state;
+ spinlock_t *io_lock = NULL;
+ unsigned long flags;
+ DECLARE_COMPLETION_ONSTACK(tm_done);
+ int ret = FAILED, tag = snic_cmd_tag(sc);
+
+ io_lock = snic_io_lock_hash(snic, sc);
+ spin_lock_irqsave(io_lock, flags);
+ CMD_FLAGS(sc) |= SNIC_DEVICE_RESET;
+ rqi = (struct snic_req_info *) CMD_SP(sc);
+ if (!rqi) {
+ SNIC_HOST_ERR(snic->shost,
+ "send_dr: rqi is null, Tag 0x%x flags 0x%llx\n",
+ tag, CMD_FLAGS(sc));
+ spin_unlock_irqrestore(io_lock, flags);
+
+ ret = FAILED;
+ goto send_dr_end;
+ }
+
+ /* Save Command state to restore in case Queuing failed. */
+ sv_state = CMD_STATE(sc);
+
+ CMD_STATE(sc) = SNIC_IOREQ_LR_PENDING;
+ CMD_LR_STATUS(sc) = SNIC_INVALID_CODE;
+
+ SNIC_SCSI_DBG(snic->shost, "dr: TAG = %x\n", tag);
+
+ rqi->dr_done = &tm_done;
+ SNIC_BUG_ON(!rqi->dr_done);
+
+ spin_unlock_irqrestore(io_lock, flags);
+ /*
+ * The Command state is changed to IOREQ_PENDING,
+ * in this case, if the command is completed, the icmnd_cmpl will
+ * mark the cmd as completed.
+ * This logic still makes LUN Reset is inevitable.
+ */
+
+ ret = snic_queue_dr_req(snic, rqi, sc);
+ if (ret) {
+ SNIC_HOST_ERR(snic->shost,
+ "send_dr: IO w/ Tag 0x%x Failed err = %d. flags 0x%llx\n",
+ tag, ret, CMD_FLAGS(sc));
+
+ spin_lock_irqsave(io_lock, flags);
+ /* Restore State */
+ CMD_STATE(sc) = sv_state;
+ rqi = (struct snic_req_info *) CMD_SP(sc);
+ if (rqi)
+ rqi->dr_done = NULL;
+ /* rqi is freed in caller. */
+ spin_unlock_irqrestore(io_lock, flags);
+ ret = FAILED;
+
+ goto send_dr_end;
+ }
+
+ spin_lock_irqsave(io_lock, flags);
+ CMD_FLAGS(sc) |= SNIC_DEV_RST_ISSUED;
+ spin_unlock_irqrestore(io_lock, flags);
+
+ ret = 0;
+
+ wait_for_completion_timeout(&tm_done, SNIC_LUN_RESET_TIMEOUT);
+
+send_dr_end:
+ return ret;
+}
+
+/*
+ * auxillary funciton to check lun reset op is supported or not
+ * Not supported if returns 0
+ */
+static int
+snic_dev_reset_supported(struct scsi_device *sdev)
+{
+ struct snic_tgt *tgt = starget_to_tgt(scsi_target(sdev));
+
+ if (tgt->tdata.typ == SNIC_TGT_DAS)
+ return 0;
+
+ return 1;
+}
+
+static void
+snic_unlink_and_release_req(struct snic *snic, struct scsi_cmnd *sc, int flag)
+{
+ struct snic_req_info *rqi = NULL;
+ spinlock_t *io_lock = NULL;
+ unsigned long flags;
+ u32 start_time = jiffies;
+
+ io_lock = snic_io_lock_hash(snic, sc);
+ spin_lock_irqsave(io_lock, flags);
+ rqi = (struct snic_req_info *) CMD_SP(sc);
+ if (rqi) {
+ start_time = rqi->start_time;
+ CMD_SP(sc) = NULL;
+ }
+
+ CMD_FLAGS(sc) |= flag;
+ spin_unlock_irqrestore(io_lock, flags);
+
+ if (rqi)
+ snic_release_req_buf(snic, rqi, sc);
+
+ SNIC_TRC(snic->shost->host_no, snic_cmd_tag(sc), sc,
+ jiffies_to_msecs(jiffies - start_time),
+ rqi, SNIC_TRC_CMD(sc), SNIC_TRC_CMD_STATE_FLAGS(sc));
+}
+
+/*
+ * SCSI Eh thread issues a LUN Reset when one or more commands on a LUN
+ * fail to get aborted. It calls driver's eh_device_reset with a SCSI
+ * command on the LUN.
+ */
+int
+snic_device_reset(struct scsi_cmnd *sc)
+{
+ struct Scsi_Host *shost = sc->device->host;
+ struct snic *snic = shost_priv(shost);
+ struct snic_req_info *rqi = NULL;
+ int tag = snic_cmd_tag(sc);
+ int start_time = jiffies;
+ int ret = FAILED;
+ int dr_supp = 0;
+
+ SNIC_SCSI_DBG(shost, "dev_reset:sc %p :0x%x :req = %p :tag = %d\n",
+ sc, sc->cmnd[0], sc->request,
+ snic_cmd_tag(sc));
+ dr_supp = snic_dev_reset_supported(sc->device);
+ if (!dr_supp) {
+ /* device reset op is not supported */
+ SNIC_HOST_INFO(shost, "LUN Reset Op not supported.\n");
+ snic_unlink_and_release_req(snic, sc, SNIC_DEV_RST_NOTSUP);
+
+ goto dev_rst_end;
+ }
+
+ if (unlikely(snic_get_state(snic) != SNIC_ONLINE)) {
+ snic_unlink_and_release_req(snic, sc, 0);
+ SNIC_HOST_ERR(shost, "Devrst: Parent Devs are not online.\n");
+
+ goto dev_rst_end;
+ }
+
+ /* There is no tag when lun reset is issue through ioctl. */
+ if (unlikely(tag <= SNIC_NO_TAG)) {
+ SNIC_HOST_INFO(snic->shost,
+ "Devrst: LUN Reset Recvd thru IOCTL.\n");
+
+ rqi = snic_req_init(snic, 0);
+ if (!rqi)
+ goto dev_rst_end;
+
+ memset(scsi_cmd_priv(sc), 0,
+ sizeof(struct snic_internal_io_state));
+ CMD_SP(sc) = (char *)rqi;
+ CMD_FLAGS(sc) = SNIC_NO_FLAGS;
+
+ /* Add special tag for dr coming from user spc */
+ rqi->tm_tag = SNIC_TAG_IOCTL_DEV_RST;
+ rqi->sc = sc;
+ }
+
+ ret = snic_send_dr_and_wait(snic, sc);
+ if (ret) {
+ SNIC_HOST_ERR(snic->shost,
+ "Devrst: IO w/ Tag %x Failed w/ err = %d\n",
+ tag, ret);
+
+ snic_unlink_and_release_req(snic, sc, 0);
+
+ goto dev_rst_end;
+ }
+
+ ret = snic_dr_finish(snic, sc);
+
+dev_rst_end:
+ SNIC_TRC(snic->shost->host_no, tag, sc,
+ jiffies_to_msecs(jiffies - start_time),
+ 0, SNIC_TRC_CMD(sc), SNIC_TRC_CMD_STATE_FLAGS(sc));
+
+ SNIC_SCSI_DBG(snic->shost,
+ "Devrst: Returning from Device Reset : %s\n",
+ (ret == SUCCESS) ? "SUCCESS" : "FAILED");
+
+ return ret;
+} /* end of snic_device_reset */
+
+/*
+ * SCSI Error handling calls driver's eh_host_reset if all prior
+ * error handling levels return FAILED.
+ *
+ * Host Reset is the highest level of error recovery. If this fails, then
+ * host is offlined by SCSI.
+ */
+/*
+ * snic_issue_hba_reset : Queues FW Reset Request.
+ */
+static int
+snic_issue_hba_reset(struct snic *snic, struct scsi_cmnd *sc)
+{
+ struct snic_req_info *rqi = NULL;
+ struct snic_host_req *req = NULL;
+ spinlock_t *io_lock = NULL;
+ DECLARE_COMPLETION_ONSTACK(wait);
+ unsigned long flags;
+ int ret = -ENOMEM;
+
+ rqi = snic_req_init(snic, 0);
+ if (!rqi) {
+ ret = -ENOMEM;
+
+ goto hba_rst_end;
+ }
+
+ if (snic_cmd_tag(sc) == SCSI_NO_TAG) {
+ memset(scsi_cmd_priv(sc), 0,
+ sizeof(struct snic_internal_io_state));
+ SNIC_HOST_INFO(snic->shost, "issu_hr:Host reset thru ioctl.\n");
+ rqi->sc = sc;
+ }
+
+ req = rqi_to_req(rqi);
+
+ io_lock = snic_io_lock_hash(snic, sc);
+ spin_lock_irqsave(io_lock, flags);
+ SNIC_BUG_ON(CMD_SP(sc) != NULL);
+ CMD_STATE(sc) = SNIC_IOREQ_PENDING;
+ CMD_SP(sc) = (char *) rqi;
+ CMD_FLAGS(sc) |= SNIC_IO_INITIALIZED;
+ snic->remove_wait = &wait;
+ spin_unlock_irqrestore(io_lock, flags);
+
+ /* Initialize Request */
+ snic_io_hdr_enc(&req->hdr, SNIC_REQ_HBA_RESET, 0, snic_cmd_tag(sc),
+ snic->config.hid, 0, (u64)rqi);
+
+ req->u.reset.flags = 0;
+
+ ret = snic_queue_wq_desc(snic, req, sizeof(*req));
+ if (ret) {
+ SNIC_HOST_ERR(snic->shost,
+ "issu_hr:Queuing HBA Reset Failed. w err %d\n",
+ ret);
+
+ goto hba_rst_err;
+ }
+
+ spin_lock_irqsave(io_lock, flags);
+ CMD_FLAGS(sc) |= SNIC_HOST_RESET_ISSUED;
+ spin_unlock_irqrestore(io_lock, flags);
+ atomic64_inc(&snic->s_stats.reset.hba_resets);
+ SNIC_HOST_INFO(snic->shost, "Queued HBA Reset Successfully.\n");
+
+ wait_for_completion_timeout(snic->remove_wait,
+ SNIC_HOST_RESET_TIMEOUT);
+
+ if (snic_get_state(snic) == SNIC_FWRESET) {
+ SNIC_HOST_ERR(snic->shost, "reset_cmpl: Reset Timedout.\n");
+ ret = -ETIMEDOUT;
+
+ goto hba_rst_err;
+ }
+
+ spin_lock_irqsave(io_lock, flags);
+ snic->remove_wait = NULL;
+ rqi = (struct snic_req_info *) CMD_SP(sc);
+ CMD_SP(sc) = NULL;
+ spin_unlock_irqrestore(io_lock, flags);
+
+ if (rqi)
+ snic_req_free(snic, rqi);
+
+ ret = 0;
+
+ return ret;
+
+hba_rst_err:
+ spin_lock_irqsave(io_lock, flags);
+ snic->remove_wait = NULL;
+ rqi = (struct snic_req_info *) CMD_SP(sc);
+ CMD_SP(sc) = NULL;
+ spin_unlock_irqrestore(io_lock, flags);
+
+ if (rqi)
+ snic_req_free(snic, rqi);
+
+hba_rst_end:
+ SNIC_HOST_ERR(snic->shost,
+ "reset:HBA Reset Failed w/ err = %d.\n",
+ ret);
+
+ return ret;
+} /* end of snic_issue_hba_reset */
+
+int
+snic_reset(struct Scsi_Host *shost, struct scsi_cmnd *sc)
+{
+ struct snic *snic = shost_priv(shost);
+ enum snic_state sv_state;
+ unsigned long flags;
+ int ret = FAILED;
+
+ /* Set snic state as SNIC_FWRESET*/
+ sv_state = snic_get_state(snic);
+
+ spin_lock_irqsave(&snic->snic_lock, flags);
+ if (snic_get_state(snic) == SNIC_FWRESET) {
+ spin_unlock_irqrestore(&snic->snic_lock, flags);
+ SNIC_HOST_INFO(shost, "reset:prev reset is in progres\n");
+
+ msleep(SNIC_HOST_RESET_TIMEOUT);
+ ret = SUCCESS;
+
+ goto reset_end;
+ }
+
+ snic_set_state(snic, SNIC_FWRESET);
+ spin_unlock_irqrestore(&snic->snic_lock, flags);
+
+
+ /* Wait for all the IOs that are entered in Qcmd */
+ while (atomic_read(&snic->ios_inflight))
+ schedule_timeout(msecs_to_jiffies(1));
+
+ ret = snic_issue_hba_reset(snic, sc);
+ if (ret) {
+ SNIC_HOST_ERR(shost,
+ "reset:Host Reset Failed w/ err %d.\n",
+ ret);
+ spin_lock_irqsave(&snic->snic_lock, flags);
+ snic_set_state(snic, sv_state);
+ spin_unlock_irqrestore(&snic->snic_lock, flags);
+ atomic64_inc(&snic->s_stats.reset.hba_reset_fail);
+ ret = FAILED;
+
+ goto reset_end;
+ }
+
+ ret = SUCCESS;
+
+reset_end:
+ return ret;
+} /* end of snic_reset */
+
+/*
+ * SCSI Error handling calls driver's eh_host_reset if all prior
+ * error handling levels return FAILED.
+ *
+ * Host Reset is the highest level of error recovery. If this fails, then
+ * host is offlined by SCSI.
+ */
+int
+snic_host_reset(struct scsi_cmnd *sc)
+{
+ struct Scsi_Host *shost = sc->device->host;
+ u32 start_time = jiffies;
+ int ret = FAILED;
+
+ SNIC_SCSI_DBG(shost,
+ "host reset:sc %p sc_cmd 0x%x req %p tag %d flags 0x%llx\n",
+ sc, sc->cmnd[0], sc->request,
+ snic_cmd_tag(sc), CMD_FLAGS(sc));
+
+ ret = snic_reset(shost, sc);
+
+ SNIC_TRC(shost->host_no, snic_cmd_tag(sc), sc,
+ jiffies_to_msecs(jiffies - start_time),
+ 0, SNIC_TRC_CMD(sc), SNIC_TRC_CMD_STATE_FLAGS(sc));
+
+ return ret;
+} /* end of snic_host_reset */
+
+/*
+ * snic_cmpl_pending_tmreq : Caller should hold io_lock
+ */
+static void
+snic_cmpl_pending_tmreq(struct snic *snic, struct scsi_cmnd *sc)
+{
+ struct snic_req_info *rqi = NULL;
+
+ SNIC_SCSI_DBG(snic->shost,
+ "Completing Pending TM Req sc %p, state %s flags 0x%llx\n",
+ sc, snic_io_status_to_str(CMD_STATE(sc)), CMD_FLAGS(sc));
+
+ rqi = (struct snic_req_info *) CMD_SP(sc);
+ if (!rqi)
+ return;
+
+ if (rqi->dr_done)
+ complete(rqi->dr_done);
+ else if (rqi->abts_done)
+ complete(rqi->abts_done);
+}
+
+/*
+ * snic_scsi_cleanup: Walks through tag map and releases the reqs
+ */
+static void
+snic_scsi_cleanup(struct snic *snic, int ex_tag)
+{
+ struct snic_req_info *rqi = NULL;
+ struct scsi_cmnd *sc = NULL;
+ spinlock_t *io_lock = NULL;
+ unsigned long flags;
+ int tag;
+ u64 st_time = 0;
+
+ SNIC_SCSI_DBG(snic->shost, "sc_clean: scsi cleanup.\n");
+
+ for (tag = 0; tag < snic->max_tag_id; tag++) {
+ /* Skip ex_tag */
+ if (tag == ex_tag)
+ continue;
+
+ io_lock = snic_io_lock_tag(snic, tag);
+ spin_lock_irqsave(io_lock, flags);
+ sc = scsi_host_find_tag(snic->shost, tag);
+ if (!sc) {
+ spin_unlock_irqrestore(io_lock, flags);
+
+ continue;
+ }
+
+ if (unlikely(snic_tmreq_pending(sc))) {
+ /*
+ * When FW Completes reset w/o sending completions
+ * for outstanding ios.
+ */
+ snic_cmpl_pending_tmreq(snic, sc);
+ spin_unlock_irqrestore(io_lock, flags);
+
+ continue;
+ }
+
+ rqi = (struct snic_req_info *) CMD_SP(sc);
+ if (!rqi) {
+ spin_unlock_irqrestore(io_lock, flags);
+
+ goto cleanup;
+ }
+
+ SNIC_SCSI_DBG(snic->shost,
+ "sc_clean: sc %p, rqi %p, tag %d flags 0x%llx\n",
+ sc, rqi, tag, CMD_FLAGS(sc));
+
+ CMD_SP(sc) = NULL;
+ CMD_FLAGS(sc) |= SNIC_SCSI_CLEANUP;
+ spin_unlock_irqrestore(io_lock, flags);
+ st_time = rqi->start_time;
+
+ SNIC_HOST_INFO(snic->shost,
+ "sc_clean: Releasing rqi %p : flags 0x%llx\n",
+ rqi, CMD_FLAGS(sc));
+
+ snic_release_req_buf(snic, rqi, sc);
+
+cleanup:
+ sc->result = DID_TRANSPORT_DISRUPTED << 16;
+ SNIC_HOST_INFO(snic->shost,
+ "sc_clean: DID_TRANSPORT_DISRUPTED for sc %p. rqi %p duration %llu msecs\n",
+ sc, rqi, (jiffies - st_time));
+
+ /* Update IO stats */
+ snic_stats_update_io_cmpl(&snic->s_stats);
+
+ if (sc->scsi_done) {
+ SNIC_TRC(snic->shost->host_no, tag, sc,
+ jiffies_to_msecs(jiffies - st_time), 0,
+ SNIC_TRC_CMD(sc),
+ SNIC_TRC_CMD_STATE_FLAGS(sc));
+
+ sc->scsi_done(sc);
+ }
+ }
+} /* end of snic_scsi_cleanup */
+
+void
+snic_shutdown_scsi_cleanup(struct snic *snic)
+{
+ SNIC_HOST_INFO(snic->shost, "Shutdown time SCSI Cleanup.\n");
+
+ snic_scsi_cleanup(snic, SCSI_NO_TAG);
+} /* end of snic_shutdown_scsi_cleanup */
+
+/*
+ * snic_internal_abort_io
+ * called by : snic_tgt_scsi_abort_io
+ */
+static int
+snic_internal_abort_io(struct snic *snic, struct scsi_cmnd *sc, int tmf)
+{
+ struct snic_req_info *rqi = NULL;
+ spinlock_t *io_lock = NULL;
+ unsigned long flags;
+ u32 sv_state = 0;
+ int ret = 0;
+
+ io_lock = snic_io_lock_hash(snic, sc);
+ spin_lock_irqsave(io_lock, flags);
+ rqi = (struct snic_req_info *) CMD_SP(sc);
+ if (!rqi)
+ goto skip_internal_abts;
+
+ if (CMD_STATE(sc) == SNIC_IOREQ_ABTS_PENDING)
+ goto skip_internal_abts;
+
+ if ((CMD_FLAGS(sc) & SNIC_DEVICE_RESET) &&
+ (!(CMD_FLAGS(sc) & SNIC_DEV_RST_ISSUED))) {
+
+ SNIC_SCSI_DBG(snic->shost,
+ "internal_abts: dev rst not pending sc 0x%p\n",
+ sc);
+
+ goto skip_internal_abts;
+ }
+
+
+ if (!(CMD_FLAGS(sc) & SNIC_IO_ISSUED)) {
+ SNIC_SCSI_DBG(snic->shost,
+ "internal_abts: IO not yet issued sc 0x%p tag 0x%x flags 0x%llx state %d\n",
+ sc, snic_cmd_tag(sc), CMD_FLAGS(sc), CMD_STATE(sc));
+
+ goto skip_internal_abts;
+ }
+
+ sv_state = CMD_STATE(sc);
+ CMD_STATE(sc) = SNIC_IOREQ_ABTS_PENDING;
+ CMD_ABTS_STATUS(sc) = SNIC_INVALID_CODE;
+ CMD_FLAGS(sc) |= SNIC_IO_INTERNAL_TERM_PENDING;
+
+ if (CMD_FLAGS(sc) & SNIC_DEVICE_RESET) {
+ /* stats */
+ rqi->tm_tag = SNIC_TAG_DEV_RST;
+ SNIC_SCSI_DBG(snic->shost, "internal_abts:dev rst sc %p\n", sc);
+ }
+
+ SNIC_SCSI_DBG(snic->shost, "internal_abts: Issuing abts tag %x\n",
+ snic_cmd_tag(sc));
+ SNIC_BUG_ON(rqi->abts_done);
+ spin_unlock_irqrestore(io_lock, flags);
+
+ ret = snic_queue_abort_req(snic, rqi, sc, tmf);
+ if (ret) {
+ SNIC_HOST_ERR(snic->shost,
+ "internal_abts: Tag = %x , Failed w/ err = %d\n",
+ snic_cmd_tag(sc), ret);
+
+ spin_lock_irqsave(io_lock, flags);
+
+ if (CMD_STATE(sc) == SNIC_IOREQ_ABTS_PENDING)
+ CMD_STATE(sc) = sv_state;
+
+ goto skip_internal_abts;
+ }
+
+ spin_lock_irqsave(io_lock, flags);
+ if (CMD_FLAGS(sc) & SNIC_DEVICE_RESET)
+ CMD_FLAGS(sc) |= SNIC_DEV_RST_TERM_ISSUED;
+ else
+ CMD_FLAGS(sc) |= SNIC_IO_INTERNAL_TERM_ISSUED;
+
+ ret = SUCCESS;
+
+skip_internal_abts:
+ SNIC_BUG_ON(!spin_is_locked(io_lock));
+ spin_unlock_irqrestore(io_lock, flags);
+
+ return ret;
+} /* end of snic_internal_abort_io */
+
+/*
+ * snic_tgt_scsi_abort_io : called by snic_tgt_del
+ */
+int
+snic_tgt_scsi_abort_io(struct snic_tgt *tgt)
+{
+ struct snic *snic = NULL;
+ struct scsi_cmnd *sc = NULL;
+ struct snic_tgt *sc_tgt = NULL;
+ spinlock_t *io_lock = NULL;
+ unsigned long flags;
+ int ret = 0, tag, abt_cnt = 0, tmf = 0;
+
+ if (!tgt)
+ return -1;
+
+ snic = shost_priv(snic_tgt_to_shost(tgt));
+ SNIC_SCSI_DBG(snic->shost, "tgt_abt_io: Cleaning Pending IOs.\n");
+
+ if (tgt->tdata.typ == SNIC_TGT_DAS)
+ tmf = SNIC_ITMF_ABTS_TASK;
+ else
+ tmf = SNIC_ITMF_ABTS_TASK_TERM;
+
+ for (tag = 0; tag < snic->max_tag_id; tag++) {
+ io_lock = snic_io_lock_tag(snic, tag);
+
+ spin_lock_irqsave(io_lock, flags);
+ sc = scsi_host_find_tag(snic->shost, tag);
+ if (!sc) {
+ spin_unlock_irqrestore(io_lock, flags);
+
+ continue;
+ }
+
+ sc_tgt = starget_to_tgt(scsi_target(sc->device));
+ if (sc_tgt != tgt) {
+ spin_unlock_irqrestore(io_lock, flags);
+
+ continue;
+ }
+ spin_unlock_irqrestore(io_lock, flags);
+
+ ret = snic_internal_abort_io(snic, sc, tmf);
+ if (ret < 0) {
+ SNIC_HOST_ERR(snic->shost,
+ "tgt_abt_io: Tag %x, Failed w err = %d\n",
+ tag, ret);
+
+ continue;
+ }
+
+ if (ret == SUCCESS)
+ abt_cnt++;
+ }
+
+ SNIC_SCSI_DBG(snic->shost, "tgt_abt_io: abt_cnt = %d\n", abt_cnt);
+
+ return 0;
+} /* end of snic_tgt_scsi_abort_io */
--
1.8.5.4
^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH v4 6/9] snic:Add low level queuing interfaces
2015-04-09 11:49 [PATCH v4 0/9] snic:initial submission of snic driver for Cisco SCSI HBA Narsimhulu Musini
` (4 preceding siblings ...)
2015-04-09 11:49 ` [PATCH v4 5/9] snic:add SCSI handling, AEN, and fwreset handling Narsimhulu Musini
@ 2015-04-09 11:49 ` Narsimhulu Musini
2015-04-09 13:23 ` Hannes Reinecke
2015-04-09 11:49 ` [PATCH v4 7/9] snic:Add sysfs entries to list stats and trace data Narsimhulu Musini
` (2 subsequent siblings)
8 siblings, 1 reply; 24+ messages in thread
From: Narsimhulu Musini @ 2015-04-09 11:49 UTC (permalink / raw)
To: JBottomley, linux-scsi, hare, hch; +Cc: Narsimhulu Musini, Sesidhar Baddela
These files contain low level queueing interfaces includes
hardware queues, and management of hardware features.
Signed-off-by: Narsimhulu Musini <nmusini@cisco.com>
Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
---
* v3
- Cleaned up unused functions.
* v2
- driver supports x86-64 arch, so removed cpu_to_XX API to maintain consistency.
drivers/scsi/snic/cq_desc.h | 76 ++++
drivers/scsi/snic/cq_enet_desc.h | 38 ++
drivers/scsi/snic/vnic_cq.c | 86 +++++
drivers/scsi/snic/vnic_cq.h | 120 ++++++
drivers/scsi/snic/vnic_cq_fw.h | 62 ++++
drivers/scsi/snic/vnic_dev.c | 749 ++++++++++++++++++++++++++++++++++++++
drivers/scsi/snic/vnic_dev.h | 140 +++++++
drivers/scsi/snic/vnic_devcmd.h | 270 ++++++++++++++
drivers/scsi/snic/vnic_intr.c | 59 +++
drivers/scsi/snic/vnic_intr.h | 119 ++++++
drivers/scsi/snic/vnic_resource.h | 68 ++++
drivers/scsi/snic/vnic_snic.h | 54 +++
drivers/scsi/snic/vnic_stats.h | 68 ++++
drivers/scsi/snic/vnic_wq.c | 236 ++++++++++++
drivers/scsi/snic/vnic_wq.h | 187 ++++++++++
drivers/scsi/snic/wq_enet_desc.h | 91 +++++
16 files changed, 2423 insertions(+)
create mode 100644 drivers/scsi/snic/cq_desc.h
create mode 100644 drivers/scsi/snic/cq_enet_desc.h
create mode 100644 drivers/scsi/snic/vnic_cq.c
create mode 100644 drivers/scsi/snic/vnic_cq.h
create mode 100644 drivers/scsi/snic/vnic_cq_fw.h
create mode 100644 drivers/scsi/snic/vnic_dev.c
create mode 100644 drivers/scsi/snic/vnic_dev.h
create mode 100644 drivers/scsi/snic/vnic_devcmd.h
create mode 100644 drivers/scsi/snic/vnic_intr.c
create mode 100644 drivers/scsi/snic/vnic_intr.h
create mode 100644 drivers/scsi/snic/vnic_resource.h
create mode 100644 drivers/scsi/snic/vnic_snic.h
create mode 100644 drivers/scsi/snic/vnic_stats.h
create mode 100644 drivers/scsi/snic/vnic_wq.c
create mode 100644 drivers/scsi/snic/vnic_wq.h
create mode 100644 drivers/scsi/snic/wq_enet_desc.h
diff --git a/drivers/scsi/snic/cq_desc.h b/drivers/scsi/snic/cq_desc.h
new file mode 100644
index 0000000..630edfa
--- /dev/null
+++ b/drivers/scsi/snic/cq_desc.h
@@ -0,0 +1,76 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef _CQ_DESC_H_
+#define _CQ_DESC_H_
+
+/*
+ * Completion queue descriptor types
+ */
+enum cq_desc_types {
+ CQ_DESC_TYPE_WQ_ENET = 0,
+ CQ_DESC_TYPE_DESC_COPY = 1,
+ CQ_DESC_TYPE_WQ_EXCH = 2,
+ CQ_DESC_TYPE_RQ_ENET = 3,
+ CQ_DESC_TYPE_RQ_FCP = 4,
+};
+
+/* Completion queue descriptor: 16B
+ *
+ * All completion queues have this basic layout. The
+ * type_specific area is unique for each completion
+ * queue type.
+ */
+struct cq_desc {
+ u16 completed_index;
+ u16 q_number;
+ u8 type_specific[11];
+ u8 type_color;
+};
+
+#define CQ_DESC_TYPE_BITS 4
+#define CQ_DESC_TYPE_MASK ((1 << CQ_DESC_TYPE_BITS) - 1)
+#define CQ_DESC_COLOR_MASK 1
+#define CQ_DESC_COLOR_SHIFT 7
+#define CQ_DESC_Q_NUM_BITS 10
+#define CQ_DESC_Q_NUM_MASK ((1 << CQ_DESC_Q_NUM_BITS) - 1)
+#define CQ_DESC_COMP_NDX_BITS 12
+#define CQ_DESC_COMP_NDX_MASK ((1 << CQ_DESC_COMP_NDX_BITS) - 1)
+
+static inline void cq_desc_dec(const struct cq_desc *desc_arg,
+ u8 *type, u8 *color, u16 *q_number, u16 *completed_index)
+{
+ const struct cq_desc *desc = desc_arg;
+ const u8 type_color = desc->type_color;
+
+ *color = (type_color >> CQ_DESC_COLOR_SHIFT) & CQ_DESC_COLOR_MASK;
+
+ /*
+ * Make sure color bit is read from desc *before* other fields
+ * are read from desc. Hardware guarantees color bit is last
+ * bit (byte) written. Adding the rmb() prevents the compiler
+ * and/or CPU from reordering the reads which would potentially
+ * result in reading stale values.
+ */
+ rmb();
+
+ *type = type_color & CQ_DESC_TYPE_MASK;
+ *q_number = desc->q_number & CQ_DESC_Q_NUM_MASK;
+ *completed_index = desc->completed_index & CQ_DESC_COMP_NDX_MASK;
+}
+
+#endif /* _CQ_DESC_H_ */
diff --git a/drivers/scsi/snic/cq_enet_desc.h b/drivers/scsi/snic/cq_enet_desc.h
new file mode 100644
index 0000000..99ecd20
--- /dev/null
+++ b/drivers/scsi/snic/cq_enet_desc.h
@@ -0,0 +1,38 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef _CQ_ENET_DESC_H_
+#define _CQ_ENET_DESC_H_
+
+#include "cq_desc.h"
+
+/* Ethernet completion queue descriptor: 16B */
+struct cq_enet_wq_desc {
+ u16 completed_index;
+ u16 q_number;
+ u8 reserved[11];
+ u8 type_color;
+};
+
+static inline void cq_enet_wq_desc_dec(struct cq_enet_wq_desc *desc,
+ u8 *type, u8 *color, u16 *q_number, u16 *completed_index)
+{
+ cq_desc_dec((struct cq_desc *)desc, type,
+ color, q_number, completed_index);
+}
+
+#endif /* _CQ_ENET_DESC_H_ */
diff --git a/drivers/scsi/snic/vnic_cq.c b/drivers/scsi/snic/vnic_cq.c
new file mode 100644
index 0000000..88d4537
--- /dev/null
+++ b/drivers/scsi/snic/vnic_cq.c
@@ -0,0 +1,86 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/errno.h>
+#include <linux/types.h>
+#include <linux/pci.h>
+#include "vnic_dev.h"
+#include "vnic_cq.h"
+
+void vnic_cq_free(struct vnic_cq *cq)
+{
+ vnic_dev_free_desc_ring(cq->vdev, &cq->ring);
+
+ cq->ctrl = NULL;
+}
+
+int vnic_cq_alloc(struct vnic_dev *vdev, struct vnic_cq *cq, unsigned int index,
+ unsigned int desc_count, unsigned int desc_size)
+{
+ int err;
+
+ cq->index = index;
+ cq->vdev = vdev;
+
+ cq->ctrl = vnic_dev_get_res(vdev, RES_TYPE_CQ, index);
+ if (!cq->ctrl) {
+ pr_err("Failed to hook CQ[%d] resource\n", index);
+
+ return -EINVAL;
+ }
+
+ err = vnic_dev_alloc_desc_ring(vdev, &cq->ring, desc_count, desc_size);
+ if (err)
+ return err;
+
+ return 0;
+}
+
+void vnic_cq_init(struct vnic_cq *cq, unsigned int flow_control_enable,
+ unsigned int color_enable, unsigned int cq_head, unsigned int cq_tail,
+ unsigned int cq_tail_color, unsigned int interrupt_enable,
+ unsigned int cq_entry_enable, unsigned int cq_message_enable,
+ unsigned int interrupt_offset, u64 cq_message_addr)
+{
+ u64 paddr;
+
+ paddr = (u64)cq->ring.base_addr | VNIC_PADDR_TARGET;
+ writeq(paddr, &cq->ctrl->ring_base);
+ iowrite32(cq->ring.desc_count, &cq->ctrl->ring_size);
+ iowrite32(flow_control_enable, &cq->ctrl->flow_control_enable);
+ iowrite32(color_enable, &cq->ctrl->color_enable);
+ iowrite32(cq_head, &cq->ctrl->cq_head);
+ iowrite32(cq_tail, &cq->ctrl->cq_tail);
+ iowrite32(cq_tail_color, &cq->ctrl->cq_tail_color);
+ iowrite32(interrupt_enable, &cq->ctrl->interrupt_enable);
+ iowrite32(cq_entry_enable, &cq->ctrl->cq_entry_enable);
+ iowrite32(cq_message_enable, &cq->ctrl->cq_message_enable);
+ iowrite32(interrupt_offset, &cq->ctrl->interrupt_offset);
+ writeq(cq_message_addr, &cq->ctrl->cq_message_addr);
+}
+
+void vnic_cq_clean(struct vnic_cq *cq)
+{
+ cq->to_clean = 0;
+ cq->last_color = 0;
+
+ iowrite32(0, &cq->ctrl->cq_head);
+ iowrite32(0, &cq->ctrl->cq_tail);
+ iowrite32(1, &cq->ctrl->cq_tail_color);
+
+ vnic_dev_clear_desc_ring(&cq->ring);
+}
diff --git a/drivers/scsi/snic/vnic_cq.h b/drivers/scsi/snic/vnic_cq.h
new file mode 100644
index 0000000..fb2dc61
--- /dev/null
+++ b/drivers/scsi/snic/vnic_cq.h
@@ -0,0 +1,120 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef _VNIC_CQ_H_
+#define _VNIC_CQ_H_
+
+#include "cq_desc.h"
+#include "vnic_dev.h"
+
+/*
+ * These defines avoid symbol clash between fnic and enic (Cisco 10G Eth
+ * Driver) when both are built with CONFIG options =y
+ */
+#define vnic_cq_service snic_cq_service
+#define vnic_cq_free snic_cq_free
+#define vnic_cq_alloc snic_cq_alloc
+#define vnic_cq_init snic_cq_init
+#define vnic_cq_clean snic_cq_clean
+
+/* Completion queue control */
+struct vnic_cq_ctrl {
+ u64 ring_base; /* 0x00 */
+ u32 ring_size; /* 0x08 */
+ u32 pad0;
+ u32 flow_control_enable; /* 0x10 */
+ u32 pad1;
+ u32 color_enable; /* 0x18 */
+ u32 pad2;
+ u32 cq_head; /* 0x20 */
+ u32 pad3;
+ u32 cq_tail; /* 0x28 */
+ u32 pad4;
+ u32 cq_tail_color; /* 0x30 */
+ u32 pad5;
+ u32 interrupt_enable; /* 0x38 */
+ u32 pad6;
+ u32 cq_entry_enable; /* 0x40 */
+ u32 pad7;
+ u32 cq_message_enable; /* 0x48 */
+ u32 pad8;
+ u32 interrupt_offset; /* 0x50 */
+ u32 pad9;
+ u64 cq_message_addr; /* 0x58 */
+ u32 pad10;
+};
+
+struct vnic_cq {
+ unsigned int index;
+ struct vnic_dev *vdev;
+ struct vnic_cq_ctrl __iomem *ctrl; /* memory-mapped */
+ struct vnic_dev_ring ring;
+ unsigned int to_clean;
+ unsigned int last_color;
+};
+
+static inline unsigned int vnic_cq_service(struct vnic_cq *cq,
+ unsigned int work_to_do,
+ int (*q_service)(struct vnic_dev *vdev, struct cq_desc *cq_desc,
+ u8 type, u16 q_number, u16 completed_index, void *opaque),
+ void *opaque)
+{
+ struct cq_desc *cq_desc;
+ unsigned int work_done = 0;
+ u16 q_number, completed_index;
+ u8 type, color;
+
+ cq_desc = (struct cq_desc *)((u8 *)cq->ring.descs +
+ cq->ring.desc_size * cq->to_clean);
+ cq_desc_dec(cq_desc, &type, &color,
+ &q_number, &completed_index);
+
+ while (color != cq->last_color) {
+
+ if ((*q_service)(cq->vdev, cq_desc, type,
+ q_number, completed_index, opaque))
+ break;
+
+ cq->to_clean++;
+ if (cq->to_clean == cq->ring.desc_count) {
+ cq->to_clean = 0;
+ cq->last_color = cq->last_color ? 0 : 1;
+ }
+
+ cq_desc = (struct cq_desc *)((u8 *)cq->ring.descs +
+ cq->ring.desc_size * cq->to_clean);
+ cq_desc_dec(cq_desc, &type, &color,
+ &q_number, &completed_index);
+
+ work_done++;
+ if (work_done >= work_to_do)
+ break;
+ }
+
+ return work_done;
+}
+
+void vnic_cq_free(struct vnic_cq *cq);
+int vnic_cq_alloc(struct vnic_dev *vdev, struct vnic_cq *cq, unsigned int index,
+ unsigned int desc_count, unsigned int desc_size);
+void vnic_cq_init(struct vnic_cq *cq, unsigned int flow_control_enable,
+ unsigned int color_enable, unsigned int cq_head, unsigned int cq_tail,
+ unsigned int cq_tail_color, unsigned int interrupt_enable,
+ unsigned int cq_entry_enable, unsigned int message_enable,
+ unsigned int interrupt_offset, u64 message_addr);
+void vnic_cq_clean(struct vnic_cq *cq);
+#endif /* _VNIC_CQ_H_ */
diff --git a/drivers/scsi/snic/vnic_cq_fw.h b/drivers/scsi/snic/vnic_cq_fw.h
new file mode 100644
index 0000000..c2d1bbd
--- /dev/null
+++ b/drivers/scsi/snic/vnic_cq_fw.h
@@ -0,0 +1,62 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef _VNIC_CQ_FW_H_
+#define _VNIC_CQ_FW_H_
+
+#include "snic_fwint.h"
+
+static inline unsigned int
+vnic_cq_fw_service(struct vnic_cq *cq,
+ int (*q_service)(struct vnic_dev *vdev,
+ unsigned int index,
+ struct snic_fw_req *desc),
+ unsigned int work_to_do)
+
+{
+ struct snic_fw_req *desc;
+ unsigned int work_done = 0;
+ u8 color;
+
+ desc = (struct snic_fw_req *)((u8 *)cq->ring.descs +
+ cq->ring.desc_size * cq->to_clean);
+ snic_color_dec(desc, &color);
+
+ while (color != cq->last_color) {
+
+ if ((*q_service)(cq->vdev, cq->index, desc))
+ break;
+
+ cq->to_clean++;
+ if (cq->to_clean == cq->ring.desc_count) {
+ cq->to_clean = 0;
+ cq->last_color = cq->last_color ? 0 : 1;
+ }
+
+ desc = (struct snic_fw_req *)((u8 *)cq->ring.descs +
+ cq->ring.desc_size * cq->to_clean);
+ snic_color_dec(desc, &color);
+
+ work_done++;
+ if (work_done >= work_to_do)
+ break;
+ }
+
+ return work_done;
+}
+
+#endif /* _VNIC_CQ_FW_H_ */
diff --git a/drivers/scsi/snic/vnic_dev.c b/drivers/scsi/snic/vnic_dev.c
new file mode 100644
index 0000000..68c63d5
--- /dev/null
+++ b/drivers/scsi/snic/vnic_dev.c
@@ -0,0 +1,749 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/kernel.h>
+#include <linux/errno.h>
+#include <linux/types.h>
+#include <linux/pci.h>
+#include <linux/delay.h>
+#include <linux/if_ether.h>
+#include <linux/slab.h>
+#include "vnic_resource.h"
+#include "vnic_devcmd.h"
+#include "vnic_dev.h"
+#include "vnic_stats.h"
+#include "vnic_wq.h"
+
+#define VNIC_DVCMD_TMO 10000 /* Devcmd Timeout value */
+#define VNIC_NOTIFY_INTR_MASK 0x0000ffff00000000ULL
+
+struct devcmd2_controller {
+ struct vnic_wq_ctrl __iomem *wq_ctrl;
+ struct vnic_dev_ring results_ring;
+ struct vnic_wq wq;
+ struct vnic_devcmd2 *cmd_ring;
+ struct devcmd2_result *result;
+ u16 next_result;
+ u16 result_size;
+ int color;
+};
+
+struct vnic_res {
+ void __iomem *vaddr;
+ unsigned int count;
+};
+
+struct vnic_dev {
+ void *priv;
+ struct pci_dev *pdev;
+ struct vnic_res res[RES_TYPE_MAX];
+ enum vnic_dev_intr_mode intr_mode;
+ struct vnic_devcmd __iomem *devcmd;
+ struct vnic_devcmd_notify *notify;
+ struct vnic_devcmd_notify notify_copy;
+ dma_addr_t notify_pa;
+ u32 *linkstatus;
+ dma_addr_t linkstatus_pa;
+ struct vnic_stats *stats;
+ dma_addr_t stats_pa;
+ struct vnic_devcmd_fw_info *fw_info;
+ dma_addr_t fw_info_pa;
+ u64 args[VNIC_DEVCMD_NARGS];
+ struct devcmd2_controller *devcmd2;
+
+ int (*devcmd_rtn)(struct vnic_dev *vdev, enum vnic_devcmd_cmd cmd,
+ int wait);
+};
+
+#define VNIC_MAX_RES_HDR_SIZE \
+ (sizeof(struct vnic_resource_header) + \
+ sizeof(struct vnic_resource) * RES_TYPE_MAX)
+#define VNIC_RES_STRIDE 128
+
+void *vnic_dev_priv(struct vnic_dev *vdev)
+{
+ return vdev->priv;
+}
+
+static int vnic_dev_discover_res(struct vnic_dev *vdev,
+ struct vnic_dev_bar *bar, unsigned int num_bars)
+{
+ struct vnic_resource_header __iomem *rh;
+ struct vnic_resource __iomem *r;
+ u8 type;
+
+ if (num_bars == 0)
+ return -EINVAL;
+
+ if (bar->len < VNIC_MAX_RES_HDR_SIZE) {
+ pr_err("vNIC BAR0 res hdr length error\n");
+
+ return -EINVAL;
+ }
+
+ rh = bar->vaddr;
+ if (!rh) {
+ pr_err("vNIC BAR0 res hdr not mem-mapped\n");
+
+ return -EINVAL;
+ }
+
+ if (ioread32(&rh->magic) != VNIC_RES_MAGIC ||
+ ioread32(&rh->version) != VNIC_RES_VERSION) {
+ pr_err("vNIC BAR0 res magic/version error exp (%lx/%lx) curr (%x/%x)\n",
+ VNIC_RES_MAGIC, VNIC_RES_VERSION,
+ ioread32(&rh->magic), ioread32(&rh->version));
+
+ return -EINVAL;
+ }
+
+ r = (struct vnic_resource __iomem *)(rh + 1);
+
+ while ((type = ioread8(&r->type)) != RES_TYPE_EOL) {
+
+ u8 bar_num = ioread8(&r->bar);
+ u32 bar_offset = ioread32(&r->bar_offset);
+ u32 count = ioread32(&r->count);
+ u32 len;
+
+ r++;
+
+ if (bar_num >= num_bars)
+ continue;
+
+ if (!bar[bar_num].len || !bar[bar_num].vaddr)
+ continue;
+
+ switch (type) {
+ case RES_TYPE_WQ:
+ case RES_TYPE_RQ:
+ case RES_TYPE_CQ:
+ case RES_TYPE_INTR_CTRL:
+ /* each count is stride bytes long */
+ len = count * VNIC_RES_STRIDE;
+ if (len + bar_offset > bar->len) {
+ pr_err("vNIC BAR0 resource %d out-of-bounds, offset 0x%x + size 0x%x > bar len 0x%lx\n",
+ type, bar_offset,
+ len,
+ bar->len);
+
+ return -EINVAL;
+ }
+ break;
+
+ case RES_TYPE_INTR_PBA_LEGACY:
+ case RES_TYPE_DEVCMD:
+ case RES_TYPE_DEVCMD2:
+ len = count;
+ break;
+
+ default:
+ continue;
+ }
+
+ vdev->res[type].count = count;
+ vdev->res[type].vaddr = (char __iomem *)bar->vaddr + bar_offset;
+ }
+
+ return 0;
+}
+
+unsigned int vnic_dev_get_res_count(struct vnic_dev *vdev,
+ enum vnic_res_type type)
+{
+ return vdev->res[type].count;
+}
+
+void __iomem *vnic_dev_get_res(struct vnic_dev *vdev, enum vnic_res_type type,
+ unsigned int index)
+{
+ if (!vdev->res[type].vaddr)
+ return NULL;
+
+ switch (type) {
+ case RES_TYPE_WQ:
+ case RES_TYPE_RQ:
+ case RES_TYPE_CQ:
+ case RES_TYPE_INTR_CTRL:
+ return (char __iomem *)vdev->res[type].vaddr +
+ index * VNIC_RES_STRIDE;
+
+ default:
+ return (char __iomem *)vdev->res[type].vaddr;
+ }
+}
+
+unsigned int vnic_dev_desc_ring_size(struct vnic_dev_ring *ring,
+ unsigned int desc_count,
+ unsigned int desc_size)
+{
+ /* The base address of the desc rings must be 512 byte aligned.
+ * Descriptor count is aligned to groups of 32 descriptors. A
+ * count of 0 means the maximum 4096 descriptors. Descriptor
+ * size is aligned to 16 bytes.
+ */
+
+ unsigned int count_align = 32;
+ unsigned int desc_align = 16;
+
+ ring->base_align = 512;
+
+ if (desc_count == 0)
+ desc_count = 4096;
+
+ ring->desc_count = ALIGN(desc_count, count_align);
+
+ ring->desc_size = ALIGN(desc_size, desc_align);
+
+ ring->size = ring->desc_count * ring->desc_size;
+ ring->size_unaligned = ring->size + ring->base_align;
+
+ return ring->size_unaligned;
+}
+
+void vnic_dev_clear_desc_ring(struct vnic_dev_ring *ring)
+{
+ memset(ring->descs, 0, ring->size);
+}
+
+int vnic_dev_alloc_desc_ring(struct vnic_dev *vdev, struct vnic_dev_ring *ring,
+ unsigned int desc_count, unsigned int desc_size)
+{
+ vnic_dev_desc_ring_size(ring, desc_count, desc_size);
+
+ ring->descs_unaligned = pci_alloc_consistent(vdev->pdev,
+ ring->size_unaligned,
+ &ring->base_addr_unaligned);
+
+ if (!ring->descs_unaligned) {
+ pr_err("Failed to allocate ring (size=%d), aborting\n",
+ (int)ring->size);
+
+ return -ENOMEM;
+ }
+
+ ring->base_addr = ALIGN(ring->base_addr_unaligned,
+ ring->base_align);
+ ring->descs = (u8 *)ring->descs_unaligned +
+ (ring->base_addr - ring->base_addr_unaligned);
+
+ vnic_dev_clear_desc_ring(ring);
+
+ ring->desc_avail = ring->desc_count - 1;
+
+ return 0;
+}
+
+void vnic_dev_free_desc_ring(struct vnic_dev *vdev, struct vnic_dev_ring *ring)
+{
+ if (ring->descs) {
+ pci_free_consistent(vdev->pdev,
+ ring->size_unaligned,
+ ring->descs_unaligned,
+ ring->base_addr_unaligned);
+ ring->descs = NULL;
+ }
+}
+
+static int _vnic_dev_cmd2(struct vnic_dev *vdev, enum vnic_devcmd_cmd cmd,
+ int wait)
+{
+ struct devcmd2_controller *dc2c = vdev->devcmd2;
+ struct devcmd2_result *result = dc2c->result + dc2c->next_result;
+ unsigned int i;
+ int delay;
+ int err;
+ u32 posted;
+ u32 new_posted;
+
+ posted = ioread32(&dc2c->wq_ctrl->posted_index);
+
+ if (posted == 0xFFFFFFFF) { /* check for hardware gone */
+ /* Hardware surprise removal: return error */
+ return -ENODEV;
+ }
+
+ new_posted = (posted + 1) % DEVCMD2_RING_SIZE;
+ dc2c->cmd_ring[posted].cmd = cmd;
+ dc2c->cmd_ring[posted].flags = 0;
+
+ if ((_CMD_FLAGS(cmd) & _CMD_FLAGS_NOWAIT))
+ dc2c->cmd_ring[posted].flags |= DEVCMD2_FNORESULT;
+
+ if (_CMD_DIR(cmd) & _CMD_DIR_WRITE) {
+ for (i = 0; i < VNIC_DEVCMD_NARGS; i++)
+ dc2c->cmd_ring[posted].args[i] = vdev->args[i];
+ }
+ /* Adding write memory barrier prevents compiler and/or CPU
+ * reordering, thus avoiding descriptor posting before
+ * descriptor is initialized. Otherwise, hardware can read
+ * stale descriptor fields.
+ */
+ wmb();
+ iowrite32(new_posted, &dc2c->wq_ctrl->posted_index);
+
+ if (dc2c->cmd_ring[posted].flags & DEVCMD2_FNORESULT)
+ return 0;
+
+ for (delay = 0; delay < wait; delay++) {
+ udelay(100);
+ if (result->color == dc2c->color) {
+ dc2c->next_result++;
+ if (dc2c->next_result == dc2c->result_size) {
+ dc2c->next_result = 0;
+ dc2c->color = dc2c->color ? 0 : 1;
+ }
+ if (result->error) {
+ err = (int) result->error;
+ if (err != ERR_ECMDUNKNOWN ||
+ cmd != CMD_CAPABILITY)
+ pr_err("Error %d devcmd %d\n",
+ err, _CMD_N(cmd));
+
+ return err;
+ }
+ if (_CMD_DIR(cmd) & _CMD_DIR_READ) {
+ /*
+ * Adding the rmb() prevents the compiler
+ * and/or CPU from reordering the reads which
+ * would potentially result in reading stale
+ * values.
+ */
+ rmb();
+ for (i = 0; i < VNIC_DEVCMD_NARGS; i++)
+ vdev->args[i] = result->results[i];
+ }
+
+ return 0;
+ }
+ }
+
+ pr_err("Timed out devcmd %d\n", _CMD_N(cmd));
+
+ return -ETIMEDOUT;
+}
+
+static int vnic_dev_init_devcmd2(struct vnic_dev *vdev)
+{
+ struct devcmd2_controller *dc2c = NULL;
+ unsigned int fetch_idx;
+ int ret;
+ void __iomem *p;
+
+ if (vdev->devcmd2)
+ return 0;
+
+ p = vnic_dev_get_res(vdev, RES_TYPE_DEVCMD2, 0);
+ if (!p)
+ return -ENODEV;
+
+ dc2c = kzalloc(sizeof(*dc2c), GFP_ATOMIC);
+ if (!dc2c)
+ return -ENOMEM;
+
+ vdev->devcmd2 = dc2c;
+
+ dc2c->color = 1;
+ dc2c->result_size = DEVCMD2_RING_SIZE;
+
+ ret = vnic_wq_devcmd2_alloc(vdev,
+ &dc2c->wq,
+ DEVCMD2_RING_SIZE,
+ DEVCMD2_DESC_SIZE);
+ if (ret)
+ goto err_free_devcmd2;
+
+ fetch_idx = ioread32(&dc2c->wq.ctrl->fetch_index);
+ if (fetch_idx == 0xFFFFFFFF) { /* check for hardware gone */
+ /* Hardware surprise removal: reset fetch_index */
+ fetch_idx = 0;
+ }
+
+ /*
+ * Don't change fetch_index ever and
+ * set posted_index same as fetch_index
+ * when setting up the WQ for devcmd2.
+ */
+ vnic_wq_init_start(&dc2c->wq, 0, fetch_idx, fetch_idx, 0, 0);
+ vnic_wq_enable(&dc2c->wq);
+ ret = vnic_dev_alloc_desc_ring(vdev,
+ &dc2c->results_ring,
+ DEVCMD2_RING_SIZE,
+ DEVCMD2_DESC_SIZE);
+ if (ret)
+ goto err_free_wq;
+
+ dc2c->result = (struct devcmd2_result *) dc2c->results_ring.descs;
+ dc2c->cmd_ring = (struct vnic_devcmd2 *) dc2c->wq.ring.descs;
+ dc2c->wq_ctrl = dc2c->wq.ctrl;
+ vdev->args[0] = (u64) dc2c->results_ring.base_addr | VNIC_PADDR_TARGET;
+ vdev->args[1] = DEVCMD2_RING_SIZE;
+
+ ret = _vnic_dev_cmd2(vdev, CMD_INITIALIZE_DEVCMD2, VNIC_DVCMD_TMO);
+ if (ret < 0)
+ goto err_free_desc_ring;
+
+ vdev->devcmd_rtn = &_vnic_dev_cmd2;
+ pr_info("DEVCMD2 Initialized.\n");
+
+ return ret;
+
+err_free_desc_ring:
+ vnic_dev_free_desc_ring(vdev, &dc2c->results_ring);
+
+err_free_wq:
+ vnic_wq_disable(&dc2c->wq);
+ vnic_wq_free(&dc2c->wq);
+
+err_free_devcmd2:
+ kfree(dc2c);
+ vdev->devcmd2 = NULL;
+
+ return ret;
+} /* end of vnic_dev_init_devcmd2 */
+
+static void vnic_dev_deinit_devcmd2(struct vnic_dev *vdev)
+{
+ struct devcmd2_controller *dc2c = vdev->devcmd2;
+
+ vdev->devcmd2 = NULL;
+ vdev->devcmd_rtn = NULL;
+
+ vnic_dev_free_desc_ring(vdev, &dc2c->results_ring);
+ vnic_wq_disable(&dc2c->wq);
+ vnic_wq_free(&dc2c->wq);
+ kfree(dc2c);
+}
+
+int vnic_dev_cmd(struct vnic_dev *vdev, enum vnic_devcmd_cmd cmd,
+ u64 *a0, u64 *a1, int wait)
+{
+ int err;
+
+ memset(vdev->args, 0, sizeof(vdev->args));
+ vdev->args[0] = *a0;
+ vdev->args[1] = *a1;
+
+ err = (*vdev->devcmd_rtn)(vdev, cmd, wait);
+
+ *a0 = vdev->args[0];
+ *a1 = vdev->args[1];
+
+ return err;
+}
+
+int vnic_dev_fw_info(struct vnic_dev *vdev,
+ struct vnic_devcmd_fw_info **fw_info)
+{
+ u64 a0, a1 = 0;
+ int wait = VNIC_DVCMD_TMO;
+ int err = 0;
+
+ if (!vdev->fw_info) {
+ vdev->fw_info = pci_alloc_consistent(vdev->pdev,
+ sizeof(struct vnic_devcmd_fw_info),
+ &vdev->fw_info_pa);
+ if (!vdev->fw_info)
+ return -ENOMEM;
+
+ a0 = vdev->fw_info_pa;
+
+ /* only get fw_info once and cache it */
+ err = vnic_dev_cmd(vdev, CMD_MCPU_FW_INFO, &a0, &a1, wait);
+ }
+
+ *fw_info = vdev->fw_info;
+
+ return err;
+}
+
+int vnic_dev_spec(struct vnic_dev *vdev, unsigned int offset, unsigned int size,
+ void *value)
+{
+ u64 a0, a1;
+ int wait = VNIC_DVCMD_TMO;
+ int err;
+
+ a0 = offset;
+ a1 = size;
+
+ err = vnic_dev_cmd(vdev, CMD_DEV_SPEC, &a0, &a1, wait);
+
+ switch (size) {
+ case 1:
+ *(u8 *)value = (u8)a0;
+ break;
+ case 2:
+ *(u16 *)value = (u16)a0;
+ break;
+ case 4:
+ *(u32 *)value = (u32)a0;
+ break;
+ case 8:
+ *(u64 *)value = a0;
+ break;
+ default:
+ BUG();
+ break;
+ }
+
+ return err;
+}
+
+int vnic_dev_stats_clear(struct vnic_dev *vdev)
+{
+ u64 a0 = 0, a1 = 0;
+ int wait = VNIC_DVCMD_TMO;
+
+ return vnic_dev_cmd(vdev, CMD_STATS_CLEAR, &a0, &a1, wait);
+}
+
+int vnic_dev_stats_dump(struct vnic_dev *vdev, struct vnic_stats **stats)
+{
+ u64 a0, a1;
+ int wait = VNIC_DVCMD_TMO;
+
+ if (!vdev->stats) {
+ vdev->stats = pci_alloc_consistent(vdev->pdev,
+ sizeof(struct vnic_stats), &vdev->stats_pa);
+ if (!vdev->stats)
+ return -ENOMEM;
+ }
+
+ *stats = vdev->stats;
+ a0 = vdev->stats_pa;
+ a1 = sizeof(struct vnic_stats);
+
+ return vnic_dev_cmd(vdev, CMD_STATS_DUMP, &a0, &a1, wait);
+}
+
+int vnic_dev_close(struct vnic_dev *vdev)
+{
+ u64 a0 = 0, a1 = 0;
+ int wait = VNIC_DVCMD_TMO;
+
+ return vnic_dev_cmd(vdev, CMD_CLOSE, &a0, &a1, wait);
+}
+
+int vnic_dev_enable_wait(struct vnic_dev *vdev)
+{
+ u64 a0 = 0, a1 = 0;
+ int wait = VNIC_DVCMD_TMO;
+ int err = 0;
+
+ err = vnic_dev_cmd(vdev, CMD_ENABLE_WAIT, &a0, &a1, wait);
+ if (err == ERR_ECMDUNKNOWN)
+ return vnic_dev_cmd(vdev, CMD_ENABLE, &a0, &a1, wait);
+
+ return err;
+}
+
+int vnic_dev_disable(struct vnic_dev *vdev)
+{
+ u64 a0 = 0, a1 = 0;
+ int wait = VNIC_DVCMD_TMO;
+
+ return vnic_dev_cmd(vdev, CMD_DISABLE, &a0, &a1, wait);
+}
+
+int vnic_dev_open(struct vnic_dev *vdev, int arg)
+{
+ u64 a0 = (u32)arg, a1 = 0;
+ int wait = VNIC_DVCMD_TMO;
+
+ return vnic_dev_cmd(vdev, CMD_OPEN, &a0, &a1, wait);
+}
+
+int vnic_dev_open_done(struct vnic_dev *vdev, int *done)
+{
+ u64 a0 = 0, a1 = 0;
+ int wait = VNIC_DVCMD_TMO;
+ int err;
+
+ *done = 0;
+
+ err = vnic_dev_cmd(vdev, CMD_OPEN_STATUS, &a0, &a1, wait);
+ if (err)
+ return err;
+
+ *done = (a0 == 0);
+
+ return 0;
+}
+
+int vnic_dev_notify_set(struct vnic_dev *vdev, u16 intr)
+{
+ u64 a0, a1;
+ int wait = VNIC_DVCMD_TMO;
+
+ if (!vdev->notify) {
+ vdev->notify = pci_alloc_consistent(vdev->pdev,
+ sizeof(struct vnic_devcmd_notify),
+ &vdev->notify_pa);
+ if (!vdev->notify)
+ return -ENOMEM;
+ }
+
+ a0 = vdev->notify_pa;
+ a1 = ((u64)intr << 32) & VNIC_NOTIFY_INTR_MASK;
+ a1 += sizeof(struct vnic_devcmd_notify);
+
+ return vnic_dev_cmd(vdev, CMD_NOTIFY, &a0, &a1, wait);
+}
+
+void vnic_dev_notify_unset(struct vnic_dev *vdev)
+{
+ u64 a0, a1;
+ int wait = VNIC_DVCMD_TMO;
+
+ a0 = 0; /* paddr = 0 to unset notify buffer */
+ a1 = VNIC_NOTIFY_INTR_MASK; /* intr num = -1 to unreg for intr */
+ a1 += sizeof(struct vnic_devcmd_notify);
+
+ vnic_dev_cmd(vdev, CMD_NOTIFY, &a0, &a1, wait);
+}
+
+static int vnic_dev_notify_ready(struct vnic_dev *vdev)
+{
+ u32 *words;
+ unsigned int nwords = sizeof(struct vnic_devcmd_notify) / 4;
+ unsigned int i;
+ u32 csum;
+
+ if (!vdev->notify)
+ return 0;
+
+ do {
+ csum = 0;
+ memcpy(&vdev->notify_copy, vdev->notify,
+ sizeof(struct vnic_devcmd_notify));
+ words = (u32 *)&vdev->notify_copy;
+ for (i = 1; i < nwords; i++)
+ csum += words[i];
+ } while (csum != words[0]);
+
+ return 1;
+}
+
+int vnic_dev_init(struct vnic_dev *vdev, int arg)
+{
+ u64 a0 = (u32)arg, a1 = 0;
+ int wait = VNIC_DVCMD_TMO;
+
+ return vnic_dev_cmd(vdev, CMD_INIT, &a0, &a1, wait);
+}
+
+int vnic_dev_link_status(struct vnic_dev *vdev)
+{
+ if (vdev->linkstatus)
+ return *vdev->linkstatus;
+
+ if (!vnic_dev_notify_ready(vdev))
+ return 0;
+
+ return vdev->notify_copy.link_state;
+}
+
+u32 vnic_dev_link_down_cnt(struct vnic_dev *vdev)
+{
+ if (!vnic_dev_notify_ready(vdev))
+ return 0;
+
+ return vdev->notify_copy.link_down_cnt;
+}
+
+void vnic_dev_set_intr_mode(struct vnic_dev *vdev,
+ enum vnic_dev_intr_mode intr_mode)
+{
+ vdev->intr_mode = intr_mode;
+}
+
+enum vnic_dev_intr_mode vnic_dev_get_intr_mode(
+ struct vnic_dev *vdev)
+{
+ return vdev->intr_mode;
+}
+
+void vnic_dev_unregister(struct vnic_dev *vdev)
+{
+ if (vdev) {
+ if (vdev->notify)
+ pci_free_consistent(vdev->pdev,
+ sizeof(struct vnic_devcmd_notify),
+ vdev->notify,
+ vdev->notify_pa);
+ if (vdev->linkstatus)
+ pci_free_consistent(vdev->pdev,
+ sizeof(u32),
+ vdev->linkstatus,
+ vdev->linkstatus_pa);
+ if (vdev->stats)
+ pci_free_consistent(vdev->pdev,
+ sizeof(struct vnic_stats),
+ vdev->stats, vdev->stats_pa);
+ if (vdev->fw_info)
+ pci_free_consistent(vdev->pdev,
+ sizeof(struct vnic_devcmd_fw_info),
+ vdev->fw_info, vdev->fw_info_pa);
+ if (vdev->devcmd2)
+ vnic_dev_deinit_devcmd2(vdev);
+ kfree(vdev);
+ }
+}
+
+struct vnic_dev *vnic_dev_alloc_discover(struct vnic_dev *vdev,
+ void *priv,
+ struct pci_dev *pdev,
+ struct vnic_dev_bar *bar,
+ unsigned int num_bars)
+{
+ if (!vdev) {
+ vdev = kzalloc(sizeof(struct vnic_dev), GFP_ATOMIC);
+ if (!vdev)
+ return NULL;
+ }
+
+ vdev->priv = priv;
+ vdev->pdev = pdev;
+
+ if (vnic_dev_discover_res(vdev, bar, num_bars))
+ goto err_out;
+
+ return vdev;
+
+err_out:
+ vnic_dev_unregister(vdev);
+
+ return NULL;
+} /* end of vnic_dev_alloc_discover */
+
+/*
+ * fallback option is left to keep the interface common for other vnics.
+ */
+int vnic_dev_cmd_init(struct vnic_dev *vdev, int fallback)
+{
+ int err = -ENODEV;
+ void __iomem *p;
+
+ p = vnic_dev_get_res(vdev, RES_TYPE_DEVCMD2, 0);
+ if (p)
+ err = vnic_dev_init_devcmd2(vdev);
+ else
+ pr_err("DEVCMD2 resource not found.\n");
+
+ return err;
+} /* end of vnic_dev_cmd_init */
diff --git a/drivers/scsi/snic/vnic_dev.h b/drivers/scsi/snic/vnic_dev.h
new file mode 100644
index 0000000..19f3e76
--- /dev/null
+++ b/drivers/scsi/snic/vnic_dev.h
@@ -0,0 +1,140 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef _VNIC_DEV_H_
+#define _VNIC_DEV_H_
+
+#include "vnic_resource.h"
+#include "vnic_devcmd.h"
+
+/*
+ * These defines avoid symbol clash between fnic and enic (Cisco 10G Eth
+ * Driver) when both are built with CONFIG options =y
+ */
+#define vnic_dev_priv snic_dev_priv
+#define vnic_dev_get_res_count snic_dev_get_res_count
+#define vnic_dev_get_res snic_dev_get_res
+#define vnic_dev_desc_ring_size snic_dev_desc_ring_siz
+#define vnic_dev_clear_desc_ring snic_dev_clear_desc_ring
+#define vnic_dev_alloc_desc_ring snic_dev_alloc_desc_ring
+#define vnic_dev_free_desc_ring snic_dev_free_desc_ring
+#define vnic_dev_cmd snic_dev_cmd
+#define vnic_dev_fw_info snic_dev_fw_info
+#define vnic_dev_spec snic_dev_spec
+#define vnic_dev_stats_clear snic_dev_stats_clear
+#define vnic_dev_stats_dump snic_dev_stats_dump
+#define vnic_dev_notify_set snic_dev_notify_set
+#define vnic_dev_notify_unset snic_dev_notify_unset
+#define vnic_dev_link_status snic_dev_link_status
+#define vnic_dev_link_down_cnt snic_dev_link_down_cnt
+#define vnic_dev_close snic_dev_close
+#define vnic_dev_enable_wait snic_dev_enable_wait
+#define vnic_dev_disable snic_dev_disable
+#define vnic_dev_open snic_dev_open
+#define vnic_dev_open_done snic_dev_open_done
+#define vnic_dev_init snic_dev_init
+#define vnic_dev_set_intr_mode snic_dev_set_intr_mode
+#define vnic_dev_get_intr_mode snic_dev_get_intr_mode
+#define vnic_dev_unregister snic_dev_unregister
+
+#ifndef VNIC_PADDR_TARGET
+#define VNIC_PADDR_TARGET 0x0000000000000000ULL
+#endif
+
+#ifndef readq
+static inline u64 readq(void __iomem *reg)
+{
+ return ((u64)readl(reg + 0x4UL) << 32) | (u64)readl(reg);
+}
+
+static inline void writeq(u64 val, void __iomem *reg)
+{
+ writel(lower_32_bits(val), reg);
+ writel(upper_32_bits(val), reg + 0x4UL);
+}
+#endif
+
+enum vnic_dev_intr_mode {
+ VNIC_DEV_INTR_MODE_UNKNOWN,
+ VNIC_DEV_INTR_MODE_INTX,
+ VNIC_DEV_INTR_MODE_MSI,
+ VNIC_DEV_INTR_MODE_MSIX,
+};
+
+struct vnic_dev_bar {
+ void __iomem *vaddr;
+ dma_addr_t bus_addr;
+ unsigned long len;
+};
+
+struct vnic_dev_ring {
+ void *descs;
+ size_t size;
+ dma_addr_t base_addr;
+ size_t base_align;
+ void *descs_unaligned;
+ size_t size_unaligned;
+ dma_addr_t base_addr_unaligned;
+ unsigned int desc_size;
+ unsigned int desc_count;
+ unsigned int desc_avail;
+};
+
+struct vnic_dev;
+struct vnic_stats;
+
+void *vnic_dev_priv(struct vnic_dev *vdev);
+unsigned int vnic_dev_get_res_count(struct vnic_dev *vdev,
+ enum vnic_res_type type);
+void __iomem *vnic_dev_get_res(struct vnic_dev *vdev, enum vnic_res_type type,
+ unsigned int index);
+unsigned int vnic_dev_desc_ring_size(struct vnic_dev_ring *ring,
+ unsigned int desc_count,
+ unsigned int desc_size);
+void vnic_dev_clear_desc_ring(struct vnic_dev_ring *ring);
+int vnic_dev_alloc_desc_ring(struct vnic_dev *vdev, struct vnic_dev_ring *ring,
+ unsigned int desc_count, unsigned int desc_size);
+void vnic_dev_free_desc_ring(struct vnic_dev *vdev,
+ struct vnic_dev_ring *ring);
+int vnic_dev_cmd(struct vnic_dev *vdev, enum vnic_devcmd_cmd cmd,
+ u64 *a0, u64 *a1, int wait);
+int vnic_dev_fw_info(struct vnic_dev *vdev,
+ struct vnic_devcmd_fw_info **fw_info);
+int vnic_dev_spec(struct vnic_dev *vdev, unsigned int offset,
+ unsigned int size, void *value);
+int vnic_dev_stats_clear(struct vnic_dev *vdev);
+int vnic_dev_stats_dump(struct vnic_dev *vdev, struct vnic_stats **stats);
+int vnic_dev_notify_set(struct vnic_dev *vdev, u16 intr);
+void vnic_dev_notify_unset(struct vnic_dev *vdev);
+int vnic_dev_link_status(struct vnic_dev *vdev);
+u32 vnic_dev_link_down_cnt(struct vnic_dev *vdev);
+int vnic_dev_close(struct vnic_dev *vdev);
+int vnic_dev_enable_wait(struct vnic_dev *vdev);
+int vnic_dev_disable(struct vnic_dev *vdev);
+int vnic_dev_open(struct vnic_dev *vdev, int arg);
+int vnic_dev_open_done(struct vnic_dev *vdev, int *done);
+int vnic_dev_init(struct vnic_dev *vdev, int arg);
+struct vnic_dev *vnic_dev_alloc_discover(struct vnic_dev *vdev,
+ void *priv, struct pci_dev *pdev,
+ struct vnic_dev_bar *bar,
+ unsigned int num_bars);
+void vnic_dev_set_intr_mode(struct vnic_dev *vdev,
+ enum vnic_dev_intr_mode intr_mode);
+enum vnic_dev_intr_mode vnic_dev_get_intr_mode(struct vnic_dev *vdev);
+void vnic_dev_unregister(struct vnic_dev *vdev);
+int vnic_dev_cmd_init(struct vnic_dev *vdev, int fallback);
+#endif /* _VNIC_DEV_H_ */
diff --git a/drivers/scsi/snic/vnic_devcmd.h b/drivers/scsi/snic/vnic_devcmd.h
new file mode 100644
index 0000000..d81b4f0
--- /dev/null
+++ b/drivers/scsi/snic/vnic_devcmd.h
@@ -0,0 +1,270 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef _VNIC_DEVCMD_H_
+#define _VNIC_DEVCMD_H_
+
+#define _CMD_NBITS 14
+#define _CMD_VTYPEBITS 10
+#define _CMD_FLAGSBITS 6
+#define _CMD_DIRBITS 2
+
+#define _CMD_NMASK ((1 << _CMD_NBITS)-1)
+#define _CMD_VTYPEMASK ((1 << _CMD_VTYPEBITS)-1)
+#define _CMD_FLAGSMASK ((1 << _CMD_FLAGSBITS)-1)
+#define _CMD_DIRMASK ((1 << _CMD_DIRBITS)-1)
+
+#define _CMD_NSHIFT 0
+#define _CMD_VTYPESHIFT (_CMD_NSHIFT+_CMD_NBITS)
+#define _CMD_FLAGSSHIFT (_CMD_VTYPESHIFT+_CMD_VTYPEBITS)
+#define _CMD_DIRSHIFT (_CMD_FLAGSSHIFT+_CMD_FLAGSBITS)
+
+/*
+ * Direction bits (from host perspective).
+ */
+#define _CMD_DIR_NONE 0U
+#define _CMD_DIR_WRITE 1U
+#define _CMD_DIR_READ 2U
+#define _CMD_DIR_RW (_CMD_DIR_WRITE | _CMD_DIR_READ)
+
+/*
+ * Flag bits.
+ */
+#define _CMD_FLAGS_NONE 0U
+#define _CMD_FLAGS_NOWAIT 1U
+
+/*
+ * vNIC type bits.
+ */
+#define _CMD_VTYPE_NONE 0U
+#define _CMD_VTYPE_ENET 1U
+#define _CMD_VTYPE_FC 2U
+#define _CMD_VTYPE_SCSI 4U
+#define _CMD_VTYPE_ALL (_CMD_VTYPE_ENET | _CMD_VTYPE_FC | _CMD_VTYPE_SCSI)
+
+/*
+ * Used to create cmds..
+*/
+#define _CMDCF(dir, flags, vtype, nr) \
+ (((dir) << _CMD_DIRSHIFT) | \
+ ((flags) << _CMD_FLAGSSHIFT) | \
+ ((vtype) << _CMD_VTYPESHIFT) | \
+ ((nr) << _CMD_NSHIFT))
+#define _CMDC(dir, vtype, nr) _CMDCF(dir, 0, vtype, nr)
+#define _CMDCNW(dir, vtype, nr) _CMDCF(dir, _CMD_FLAGS_NOWAIT, vtype, nr)
+
+/*
+ * Used to decode cmds..
+*/
+#define _CMD_DIR(cmd) (((cmd) >> _CMD_DIRSHIFT) & _CMD_DIRMASK)
+#define _CMD_FLAGS(cmd) (((cmd) >> _CMD_FLAGSSHIFT) & _CMD_FLAGSMASK)
+#define _CMD_VTYPE(cmd) (((cmd) >> _CMD_VTYPESHIFT) & _CMD_VTYPEMASK)
+#define _CMD_N(cmd) (((cmd) >> _CMD_NSHIFT) & _CMD_NMASK)
+
+enum vnic_devcmd_cmd {
+ CMD_NONE = _CMDC(_CMD_DIR_NONE, _CMD_VTYPE_NONE, 0),
+
+ /* mcpu fw info in mem: (u64)a0=paddr to struct vnic_devcmd_fw_info */
+ CMD_MCPU_FW_INFO = _CMDC(_CMD_DIR_WRITE, _CMD_VTYPE_ALL, 1),
+
+ /* dev-specific block member:
+ * in: (u16)a0=offset,(u8)a1=size
+ * out: a0=value */
+ CMD_DEV_SPEC = _CMDC(_CMD_DIR_RW, _CMD_VTYPE_ALL, 2),
+
+ /* stats clear */
+ CMD_STATS_CLEAR = _CMDCNW(_CMD_DIR_NONE, _CMD_VTYPE_ALL, 3),
+
+ /* stats dump in mem: (u64)a0=paddr to stats area,
+ * (u16)a1=sizeof stats area */
+ CMD_STATS_DUMP = _CMDC(_CMD_DIR_WRITE, _CMD_VTYPE_ALL, 4),
+
+ /* nic_cfg in (u32)a0 */
+ CMD_NIC_CFG = _CMDCNW(_CMD_DIR_WRITE, _CMD_VTYPE_ALL, 16),
+
+ /* set struct vnic_devcmd_notify buffer in mem:
+ * in:
+ * (u64)a0=paddr to notify (set paddr=0 to unset)
+ * (u32)a1 & 0x00000000ffffffff=sizeof(struct vnic_devcmd_notify)
+ * (u16)a1 & 0x0000ffff00000000=intr num (-1 for no intr)
+ * out:
+ * (u32)a1 = effective size
+ */
+ CMD_NOTIFY = _CMDC(_CMD_DIR_RW, _CMD_VTYPE_ALL, 21),
+
+ /* initiate open sequence (u32)a0=flags (see CMD_OPENF_*) */
+ CMD_OPEN = _CMDCNW(_CMD_DIR_WRITE, _CMD_VTYPE_ALL, 23),
+
+ /* open status:
+ * out: a0=0 open complete, a0=1 open in progress */
+ CMD_OPEN_STATUS = _CMDC(_CMD_DIR_READ, _CMD_VTYPE_ALL, 24),
+
+ /* close vnic */
+ CMD_CLOSE = _CMDC(_CMD_DIR_NONE, _CMD_VTYPE_ALL, 25),
+
+ /* initialize virtual link: (u32)a0=flags (see CMD_INITF_*) */
+ CMD_INIT = _CMDCNW(_CMD_DIR_READ, _CMD_VTYPE_ALL, 26),
+
+ /* enable virtual link */
+ CMD_ENABLE = _CMDCNW(_CMD_DIR_WRITE, _CMD_VTYPE_ALL, 28),
+
+ /* enable virtual link, waiting variant. */
+ CMD_ENABLE_WAIT = _CMDC(_CMD_DIR_WRITE, _CMD_VTYPE_ALL, 28),
+
+ /* disable virtual link */
+ CMD_DISABLE = _CMDC(_CMD_DIR_NONE, _CMD_VTYPE_ALL, 29),
+
+ /* stats dump all vnics on uplink in mem: (u64)a0=paddr (u32)a1=uif */
+ CMD_STATS_DUMP_ALL = _CMDC(_CMD_DIR_WRITE, _CMD_VTYPE_ALL, 30),
+
+ /* init status:
+ * out: a0=0 init complete, a0=1 init in progress
+ * if a0=0, a1=errno */
+ CMD_INIT_STATUS = _CMDC(_CMD_DIR_READ, _CMD_VTYPE_ALL, 31),
+
+ /* undo initialize of virtual link */
+ CMD_DEINIT = _CMDCNW(_CMD_DIR_NONE, _CMD_VTYPE_ALL, 34),
+
+ /* check fw capability of a cmd:
+ * in: (u32)a0=cmd
+ * out: (u32)a0=errno, 0:valid cmd, a1=supported VNIC_STF_* bits */
+ CMD_CAPABILITY = _CMDC(_CMD_DIR_RW, _CMD_VTYPE_ALL, 36),
+
+ /*
+ * Initialization for the devcmd2 interface.
+ * in: (u64) a0=host result buffer physical address
+ * in: (u16) a1=number of entries in result buffer
+ */
+ CMD_INITIALIZE_DEVCMD2 = _CMDC(_CMD_DIR_WRITE, _CMD_VTYPE_ALL, 57)
+};
+
+/* flags for CMD_OPEN */
+#define CMD_OPENF_OPROM 0x1 /* open coming from option rom */
+
+/* flags for CMD_INIT */
+#define CMD_INITF_DEFAULT_MAC 0x1 /* init with default mac addr */
+
+/* flags for CMD_PACKET_FILTER */
+#define CMD_PFILTER_DIRECTED 0x01
+#define CMD_PFILTER_MULTICAST 0x02
+#define CMD_PFILTER_BROADCAST 0x04
+#define CMD_PFILTER_PROMISCUOUS 0x08
+#define CMD_PFILTER_ALL_MULTICAST 0x10
+
+enum vnic_devcmd_status {
+ STAT_NONE = 0,
+ STAT_BUSY = 1 << 0, /* cmd in progress */
+ STAT_ERROR = 1 << 1, /* last cmd caused error (code in a0) */
+};
+
+enum vnic_devcmd_error {
+ ERR_SUCCESS = 0,
+ ERR_EINVAL = 1,
+ ERR_EFAULT = 2,
+ ERR_EPERM = 3,
+ ERR_EBUSY = 4,
+ ERR_ECMDUNKNOWN = 5,
+ ERR_EBADSTATE = 6,
+ ERR_ENOMEM = 7,
+ ERR_ETIMEDOUT = 8,
+ ERR_ELINKDOWN = 9,
+};
+
+struct vnic_devcmd_fw_info {
+ char fw_version[32];
+ char fw_build[32];
+ char hw_version[32];
+ char hw_serial_number[32];
+};
+
+struct vnic_devcmd_notify {
+ u32 csum; /* checksum over following words */
+
+ u32 link_state; /* link up == 1 */
+ u32 port_speed; /* effective port speed (rate limit) */
+ u32 mtu; /* MTU */
+ u32 msglvl; /* requested driver msg lvl */
+ u32 uif; /* uplink interface */
+ u32 status; /* status bits (see VNIC_STF_*) */
+ u32 error; /* error code (see ERR_*) for first ERR */
+ u32 link_down_cnt; /* running count of link down transitions */
+};
+#define VNIC_STF_FATAL_ERR 0x0001 /* fatal fw error */
+
+struct vnic_devcmd_provinfo {
+ u8 oui[3];
+ u8 type;
+ u8 data[0];
+};
+
+/*
+ * Writing cmd register causes STAT_BUSY to get set in status register.
+ * When cmd completes, STAT_BUSY will be cleared.
+ *
+ * If cmd completed successfully STAT_ERROR will be clear
+ * and args registers contain cmd-specific results.
+ *
+ * If cmd error, STAT_ERROR will be set and args[0] contains error code.
+ *
+ * status register is read-only. While STAT_BUSY is set,
+ * all other register contents are read-only.
+ */
+
+/* Make sizeof(vnic_devcmd) a power-of-2 for I/O BAR. */
+#define VNIC_DEVCMD_NARGS 15
+struct vnic_devcmd {
+ u32 status; /* RO */
+ u32 cmd; /* RW */
+ u64 args[VNIC_DEVCMD_NARGS]; /* RW cmd args (little-endian) */
+};
+
+
+/*
+ * Version 2 of the interface.
+ *
+ * Some things are carried over, notably the vnic_devcmd_cmd enum.
+ */
+
+/*
+ * Flags for vnic_devcmd2.flags
+ */
+
+#define DEVCMD2_FNORESULT 0x1 /* Don't copy result to host */
+
+#define VNIC_DEVCMD2_NARGS VNIC_DEVCMD_NARGS
+struct vnic_devcmd2 {
+ u16 pad;
+ u16 flags;
+ u32 cmd; /* same command #defines as original */
+ u64 args[VNIC_DEVCMD2_NARGS];
+};
+
+#define VNIC_DEVCMD2_NRESULTS VNIC_DEVCMD_NARGS
+struct devcmd2_result {
+ u64 results[VNIC_DEVCMD2_NRESULTS];
+ u32 pad;
+ u16 completed_index; /* into copy WQ */
+ u8 error; /* same error codes as original */
+ u8 color; /* 0 or 1 as with completion queues */
+};
+
+#define DEVCMD2_RING_SIZE 32
+#define DEVCMD2_DESC_SIZE 128
+
+#define DEVCMD2_RESULTS_SIZE_MAX ((1 << 16) - 1)
+
+#endif /* _VNIC_DEVCMD_H_ */
diff --git a/drivers/scsi/snic/vnic_intr.c b/drivers/scsi/snic/vnic_intr.c
new file mode 100644
index 0000000..fbbfd90
--- /dev/null
+++ b/drivers/scsi/snic/vnic_intr.c
@@ -0,0 +1,59 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/kernel.h>
+#include <linux/errno.h>
+#include <linux/types.h>
+#include <linux/pci.h>
+#include <linux/delay.h>
+#include "vnic_dev.h"
+#include "vnic_intr.h"
+
+void vnic_intr_free(struct vnic_intr *intr)
+{
+ intr->ctrl = NULL;
+}
+
+int vnic_intr_alloc(struct vnic_dev *vdev, struct vnic_intr *intr,
+ unsigned int index)
+{
+ intr->index = index;
+ intr->vdev = vdev;
+
+ intr->ctrl = vnic_dev_get_res(vdev, RES_TYPE_INTR_CTRL, index);
+ if (!intr->ctrl) {
+ pr_err("Failed to hook INTR[%d].ctrl resource\n",
+ index);
+ return -EINVAL;
+ }
+
+ return 0;
+}
+
+void vnic_intr_init(struct vnic_intr *intr, unsigned int coalescing_timer,
+ unsigned int coalescing_type, unsigned int mask_on_assertion)
+{
+ iowrite32(coalescing_timer, &intr->ctrl->coalescing_timer);
+ iowrite32(coalescing_type, &intr->ctrl->coalescing_type);
+ iowrite32(mask_on_assertion, &intr->ctrl->mask_on_assertion);
+ iowrite32(0, &intr->ctrl->int_credits);
+}
+
+void vnic_intr_clean(struct vnic_intr *intr)
+{
+ iowrite32(0, &intr->ctrl->int_credits);
+}
diff --git a/drivers/scsi/snic/vnic_intr.h b/drivers/scsi/snic/vnic_intr.h
new file mode 100644
index 0000000..1712db6
--- /dev/null
+++ b/drivers/scsi/snic/vnic_intr.h
@@ -0,0 +1,119 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef _VNIC_INTR_H_
+#define _VNIC_INTR_H_
+
+#include <linux/pci.h>
+#include "vnic_dev.h"
+
+/*
+ * These defines avoid symbol clash between fnic and enic (Cisco 10G Eth
+ * Driver) when both are built with CONFIG options =y
+ */
+#define vnic_intr_unmask snic_intr_unmask
+#define vnic_intr_mask snic_intr_mask
+#define vnic_intr_return_credits snic_intr_return_credits
+#define vnic_intr_credits snic_intr_credits
+#define vnic_intr_return_all_credits snic_intr_return_all_credits
+#define vnic_intr_free snic_intr_free
+#define vnic_intr_alloc snic_intr_alloc
+#define vnic_intr_init snic_intr_init
+#define vnic_intr_clean snic_intr_clean
+
+#define VNIC_INTR_TIMER_MAX 0xffff
+
+#define VNIC_INTR_TIMER_TYPE_ABS 0
+#define VNIC_INTR_TIMER_TYPE_QUIET 1
+
+/* Interrupt control */
+struct vnic_intr_ctrl {
+ u32 coalescing_timer; /* 0x00 */
+ u32 pad0;
+ u32 coalescing_value; /* 0x08 */
+ u32 pad1;
+ u32 coalescing_type; /* 0x10 */
+ u32 pad2;
+ u32 mask_on_assertion; /* 0x18 */
+ u32 pad3;
+ u32 mask; /* 0x20 */
+ u32 pad4;
+ u32 int_credits; /* 0x28 */
+ u32 pad5;
+ u32 int_credit_return; /* 0x30 */
+ u32 pad6;
+};
+
+struct vnic_intr {
+ unsigned int index;
+ struct vnic_dev *vdev;
+ struct vnic_intr_ctrl __iomem *ctrl; /* memory-mapped */
+};
+
+static inline void
+vnic_intr_unmask(struct vnic_intr *intr)
+{
+ iowrite32(0, &intr->ctrl->mask);
+}
+
+static inline void
+vnic_intr_mask(struct vnic_intr *intr)
+{
+ iowrite32(1, &intr->ctrl->mask);
+}
+
+static inline void
+vnic_intr_return_credits(struct vnic_intr *intr,
+ unsigned int credits,
+ int unmask,
+ int reset_timer)
+{
+#define VNIC_INTR_UNMASK_SHIFT 16
+#define VNIC_INTR_RESET_TIMER_SHIFT 17
+
+ u32 int_credit_return = (credits & 0xffff) |
+ (unmask ? (1 << VNIC_INTR_UNMASK_SHIFT) : 0) |
+ (reset_timer ? (1 << VNIC_INTR_RESET_TIMER_SHIFT) : 0);
+
+ iowrite32(int_credit_return, &intr->ctrl->int_credit_return);
+}
+
+static inline unsigned int
+vnic_intr_credits(struct vnic_intr *intr)
+{
+ return ioread32(&intr->ctrl->int_credits);
+}
+
+static inline void
+vnic_intr_return_all_credits(struct vnic_intr *intr)
+{
+ unsigned int credits = vnic_intr_credits(intr);
+ int unmask = 1;
+ int reset_timer = 1;
+
+ vnic_intr_return_credits(intr, credits, unmask, reset_timer);
+}
+
+void vnic_intr_free(struct vnic_intr *);
+int vnic_intr_alloc(struct vnic_dev *, struct vnic_intr *, unsigned int);
+void vnic_intr_init(struct vnic_intr *intr,
+ unsigned int coalescing_timer,
+ unsigned int coalescing_type,
+ unsigned int mask_on_assertion);
+void vnic_intr_clean(struct vnic_intr *);
+
+#endif /* _VNIC_INTR_H_ */
diff --git a/drivers/scsi/snic/vnic_resource.h b/drivers/scsi/snic/vnic_resource.h
new file mode 100644
index 0000000..9713d68
--- /dev/null
+++ b/drivers/scsi/snic/vnic_resource.h
@@ -0,0 +1,68 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef _VNIC_RESOURCE_H_
+#define _VNIC_RESOURCE_H_
+
+#define VNIC_RES_MAGIC 0x766E6963L /* 'vnic' */
+#define VNIC_RES_VERSION 0x00000000L
+
+/* vNIC resource types */
+enum vnic_res_type {
+ RES_TYPE_EOL, /* End-of-list */
+ RES_TYPE_WQ, /* Work queues */
+ RES_TYPE_RQ, /* Receive queues */
+ RES_TYPE_CQ, /* Completion queues */
+ RES_TYPE_RSVD1,
+ RES_TYPE_NIC_CFG, /* Enet NIC config registers */
+ RES_TYPE_RSVD2,
+ RES_TYPE_RSVD3,
+ RES_TYPE_RSVD4,
+ RES_TYPE_RSVD5,
+ RES_TYPE_INTR_CTRL, /* Interrupt ctrl table */
+ RES_TYPE_INTR_TABLE, /* MSI/MSI-X Interrupt table */
+ RES_TYPE_INTR_PBA, /* MSI/MSI-X PBA table */
+ RES_TYPE_INTR_PBA_LEGACY, /* Legacy intr status */
+ RES_TYPE_RSVD6,
+ RES_TYPE_RSVD7,
+ RES_TYPE_DEVCMD, /* Device command region */
+ RES_TYPE_PASS_THRU_PAGE, /* Pass-thru page */
+ RES_TYPE_SUBVNIC, /* subvnic resource type */
+ RES_TYPE_MQ_WQ, /* MQ Work queues */
+ RES_TYPE_MQ_RQ, /* MQ Receive queues */
+ RES_TYPE_MQ_CQ, /* MQ Completion queues */
+ RES_TYPE_DEPRECATED1, /* Old version of devcmd 2 */
+ RES_TYPE_DEPRECATED2, /* Old version of devcmd 2 */
+ RES_TYPE_DEVCMD2, /* Device control region */
+
+ RES_TYPE_MAX, /* Count of resource types */
+};
+
+struct vnic_resource_header {
+ u32 magic;
+ u32 version;
+};
+
+struct vnic_resource {
+ u8 type;
+ u8 bar;
+ u8 pad[2];
+ u32 bar_offset;
+ u32 count;
+};
+
+#endif /* _VNIC_RESOURCE_H_ */
diff --git a/drivers/scsi/snic/vnic_snic.h b/drivers/scsi/snic/vnic_snic.h
new file mode 100644
index 0000000..514d39f
--- /dev/null
+++ b/drivers/scsi/snic/vnic_snic.h
@@ -0,0 +1,54 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef _VNIC_SNIC_H_
+#define _VNIC_SNIC_H_
+
+#define VNIC_SNIC_WQ_DESCS_MIN 64
+#define VNIC_SNIC_WQ_DESCS_MAX 1024
+
+#define VNIC_SNIC_MAXDATAFIELDSIZE_MIN 256
+#define VNIC_SNIC_MAXDATAFIELDSIZE_MAX 2112
+
+#define VNIC_SNIC_IO_THROTTLE_COUNT_MIN 1
+#define VNIC_SNIC_IO_THROTTLE_COUNT_MAX 1024
+
+#define VNIC_SNIC_PORT_DOWN_TIMEOUT_MIN 0
+#define VNIC_SNIC_PORT_DOWN_TIMEOUT_MAX 240000
+
+#define VNIC_SNIC_PORT_DOWN_IO_RETRIES_MIN 0
+#define VNIC_SNIC_PORT_DOWN_IO_RETRIES_MAX 255
+
+#define VNIC_SNIC_LUNS_PER_TARGET_MIN 1
+#define VNIC_SNIC_LUNS_PER_TARGET_MAX 1024
+
+/* Device-specific region: scsi configuration */
+struct vnic_snic_config {
+ u32 flags;
+ u32 wq_enet_desc_count;
+ u32 io_throttle_count;
+ u32 port_down_timeout;
+ u32 port_down_io_retries;
+ u32 luns_per_tgt;
+ u16 maxdatafieldsize;
+ u16 intr_timer;
+ u8 intr_timer_type;
+ u8 _resvd2;
+ u8 xpt_type;
+ u8 hid;
+};
+#endif /* _VNIC_SNIC_H_ */
diff --git a/drivers/scsi/snic/vnic_stats.h b/drivers/scsi/snic/vnic_stats.h
new file mode 100644
index 0000000..370a37c
--- /dev/null
+++ b/drivers/scsi/snic/vnic_stats.h
@@ -0,0 +1,68 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef _VNIC_STATS_H_
+#define _VNIC_STATS_H_
+
+/* Tx statistics */
+struct vnic_tx_stats {
+ u64 tx_frames_ok;
+ u64 tx_unicast_frames_ok;
+ u64 tx_multicast_frames_ok;
+ u64 tx_broadcast_frames_ok;
+ u64 tx_bytes_ok;
+ u64 tx_unicast_bytes_ok;
+ u64 tx_multicast_bytes_ok;
+ u64 tx_broadcast_bytes_ok;
+ u64 tx_drops;
+ u64 tx_errors;
+ u64 tx_tso;
+ u64 rsvd[16];
+};
+
+/* Rx statistics */
+struct vnic_rx_stats {
+ u64 rx_frames_ok;
+ u64 rx_frames_total;
+ u64 rx_unicast_frames_ok;
+ u64 rx_multicast_frames_ok;
+ u64 rx_broadcast_frames_ok;
+ u64 rx_bytes_ok;
+ u64 rx_unicast_bytes_ok;
+ u64 rx_multicast_bytes_ok;
+ u64 rx_broadcast_bytes_ok;
+ u64 rx_drop;
+ u64 rx_no_bufs;
+ u64 rx_errors;
+ u64 rx_rss;
+ u64 rx_crc_errors;
+ u64 rx_frames_64;
+ u64 rx_frames_127;
+ u64 rx_frames_255;
+ u64 rx_frames_511;
+ u64 rx_frames_1023;
+ u64 rx_frames_1518;
+ u64 rx_frames_to_max;
+ u64 rsvd[16];
+};
+
+struct vnic_stats {
+ struct vnic_tx_stats tx;
+ struct vnic_rx_stats rx;
+};
+
+#endif /* _VNIC_STATS_H_ */
diff --git a/drivers/scsi/snic/vnic_wq.c b/drivers/scsi/snic/vnic_wq.c
new file mode 100644
index 0000000..b323c10
--- /dev/null
+++ b/drivers/scsi/snic/vnic_wq.c
@@ -0,0 +1,236 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/errno.h>
+#include <linux/types.h>
+#include <linux/pci.h>
+#include <linux/delay.h>
+#include <linux/slab.h>
+#include "vnic_dev.h"
+#include "vnic_wq.h"
+
+static inline int vnic_wq_get_ctrl(struct vnic_dev *vdev, struct vnic_wq *wq,
+ unsigned int index, enum vnic_res_type res_type)
+{
+ wq->ctrl = vnic_dev_get_res(vdev, res_type, index);
+ if (!wq->ctrl)
+ return -EINVAL;
+
+ return 0;
+}
+
+static inline int vnic_wq_alloc_ring(struct vnic_dev *vdev, struct vnic_wq *wq,
+ unsigned int index, unsigned int desc_count, unsigned int desc_size)
+{
+ return vnic_dev_alloc_desc_ring(vdev, &wq->ring, desc_count, desc_size);
+}
+
+static int vnic_wq_alloc_bufs(struct vnic_wq *wq)
+{
+ struct vnic_wq_buf *buf;
+ unsigned int i, j, count = wq->ring.desc_count;
+ unsigned int blks = VNIC_WQ_BUF_BLKS_NEEDED(count);
+
+ for (i = 0; i < blks; i++) {
+ wq->bufs[i] = kzalloc(VNIC_WQ_BUF_BLK_SZ, GFP_ATOMIC);
+ if (!wq->bufs[i]) {
+ pr_err("Failed to alloc wq_bufs\n");
+
+ return -ENOMEM;
+ }
+ }
+
+ for (i = 0; i < blks; i++) {
+ buf = wq->bufs[i];
+ for (j = 0; j < VNIC_WQ_BUF_DFLT_BLK_ENTRIES; j++) {
+ buf->index = i * VNIC_WQ_BUF_DFLT_BLK_ENTRIES + j;
+ buf->desc = (u8 *)wq->ring.descs +
+ wq->ring.desc_size * buf->index;
+ if (buf->index + 1 == count) {
+ buf->next = wq->bufs[0];
+ break;
+ } else if (j + 1 == VNIC_WQ_BUF_DFLT_BLK_ENTRIES) {
+ buf->next = wq->bufs[i + 1];
+ } else {
+ buf->next = buf + 1;
+ buf++;
+ }
+ }
+ }
+
+ wq->to_use = wq->to_clean = wq->bufs[0];
+
+ return 0;
+}
+
+void vnic_wq_free(struct vnic_wq *wq)
+{
+ struct vnic_dev *vdev;
+ unsigned int i;
+
+ vdev = wq->vdev;
+
+ vnic_dev_free_desc_ring(vdev, &wq->ring);
+
+ for (i = 0; i < VNIC_WQ_BUF_BLKS_MAX; i++) {
+ kfree(wq->bufs[i]);
+ wq->bufs[i] = NULL;
+ }
+
+ wq->ctrl = NULL;
+
+}
+
+int vnic_wq_devcmd2_alloc(struct vnic_dev *vdev, struct vnic_wq *wq,
+ unsigned int desc_count, unsigned int desc_size)
+{
+ int err;
+
+ wq->index = 0;
+ wq->vdev = vdev;
+
+ err = vnic_wq_get_ctrl(vdev, wq, 0, RES_TYPE_DEVCMD2);
+ if (err) {
+ pr_err("Failed to get devcmd2 resource\n");
+
+ return err;
+ }
+
+ vnic_wq_disable(wq);
+
+ err = vnic_wq_alloc_ring(vdev, wq, 0, desc_count, desc_size);
+ if (err)
+ return err;
+
+ return 0;
+}
+
+int vnic_wq_alloc(struct vnic_dev *vdev, struct vnic_wq *wq, unsigned int index,
+ unsigned int desc_count, unsigned int desc_size)
+{
+ int err;
+
+ wq->index = index;
+ wq->vdev = vdev;
+
+ err = vnic_wq_get_ctrl(vdev, wq, index, RES_TYPE_WQ);
+ if (err) {
+ pr_err("Failed to hook WQ[%d] resource\n", index);
+
+ return err;
+ }
+
+ vnic_wq_disable(wq);
+
+ err = vnic_wq_alloc_ring(vdev, wq, index, desc_count, desc_size);
+ if (err)
+ return err;
+
+ err = vnic_wq_alloc_bufs(wq);
+ if (err) {
+ vnic_wq_free(wq);
+
+ return err;
+ }
+
+ return 0;
+}
+
+void vnic_wq_init_start(struct vnic_wq *wq, unsigned int cq_index,
+ unsigned int fetch_index, unsigned int posted_index,
+ unsigned int error_interrupt_enable,
+ unsigned int error_interrupt_offset)
+{
+ u64 paddr;
+ unsigned int count = wq->ring.desc_count;
+
+ paddr = (u64)wq->ring.base_addr | VNIC_PADDR_TARGET;
+ writeq(paddr, &wq->ctrl->ring_base);
+ iowrite32(count, &wq->ctrl->ring_size);
+ iowrite32(fetch_index, &wq->ctrl->fetch_index);
+ iowrite32(posted_index, &wq->ctrl->posted_index);
+ iowrite32(cq_index, &wq->ctrl->cq_index);
+ iowrite32(error_interrupt_enable, &wq->ctrl->error_interrupt_enable);
+ iowrite32(error_interrupt_offset, &wq->ctrl->error_interrupt_offset);
+ iowrite32(0, &wq->ctrl->error_status);
+
+ wq->to_use = wq->to_clean =
+ &wq->bufs[fetch_index / VNIC_WQ_BUF_BLK_ENTRIES(count)]
+ [fetch_index % VNIC_WQ_BUF_BLK_ENTRIES(count)];
+}
+
+void vnic_wq_init(struct vnic_wq *wq, unsigned int cq_index,
+ unsigned int error_interrupt_enable,
+ unsigned int error_interrupt_offset)
+{
+ vnic_wq_init_start(wq, cq_index, 0, 0, error_interrupt_enable,
+ error_interrupt_offset);
+}
+
+unsigned int vnic_wq_error_status(struct vnic_wq *wq)
+{
+ return ioread32(&wq->ctrl->error_status);
+}
+
+void vnic_wq_enable(struct vnic_wq *wq)
+{
+ iowrite32(1, &wq->ctrl->enable);
+}
+
+int vnic_wq_disable(struct vnic_wq *wq)
+{
+ unsigned int wait;
+
+ iowrite32(0, &wq->ctrl->enable);
+
+ /* Wait for HW to ACK disable request */
+ for (wait = 0; wait < 100; wait++) {
+ if (!(ioread32(&wq->ctrl->running)))
+ return 0;
+ udelay(1);
+ }
+
+ pr_err("Failed to disable WQ[%d]\n", wq->index);
+
+ return -ETIMEDOUT;
+}
+
+void vnic_wq_clean(struct vnic_wq *wq,
+ void (*buf_clean)(struct vnic_wq *wq, struct vnic_wq_buf *buf))
+{
+ struct vnic_wq_buf *buf;
+
+ BUG_ON(ioread32(&wq->ctrl->enable));
+
+ buf = wq->to_clean;
+
+ while (vnic_wq_desc_used(wq) > 0) {
+
+ (*buf_clean)(wq, buf);
+
+ buf = wq->to_clean = buf->next;
+ wq->ring.desc_avail++;
+ }
+
+ wq->to_use = wq->to_clean = wq->bufs[0];
+
+ iowrite32(0, &wq->ctrl->fetch_index);
+ iowrite32(0, &wq->ctrl->posted_index);
+ iowrite32(0, &wq->ctrl->error_status);
+
+ vnic_dev_clear_desc_ring(&wq->ring);
+}
diff --git a/drivers/scsi/snic/vnic_wq.h b/drivers/scsi/snic/vnic_wq.h
new file mode 100644
index 0000000..843e9fd
--- /dev/null
+++ b/drivers/scsi/snic/vnic_wq.h
@@ -0,0 +1,187 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef _VNIC_WQ_H_
+#define _VNIC_WQ_H_
+
+#include <linux/pci.h>
+#include "vnic_dev.h"
+#include "vnic_cq.h"
+
+/*
+ * These defines avoid symbol clash between fnic and enic (Cisco 10G Eth
+ * Driver) when both are built with CONFIG options =y
+ */
+#define vnic_wq_desc_avail snic_wq_desc_avail
+#define vnic_wq_desc_used snic_wq_desc_used
+#define vnic_wq_next_desc fni_cwq_next_desc
+#define vnic_wq_post snic_wq_post
+#define vnic_wq_service snic_wq_service
+#define vnic_wq_free snic_wq_free
+#define vnic_wq_alloc snic_wq_alloc
+#define vnic_wq_init snic_wq_init
+#define vnic_wq_error_status snic_wq_error_status
+#define vnic_wq_enable snic_wq_enable
+#define vnic_wq_disable snic_wq_disable
+#define vnic_wq_clean snic_wq_clean
+
+/* Work queue control */
+struct vnic_wq_ctrl {
+ u64 ring_base; /* 0x00 */
+ u32 ring_size; /* 0x08 */
+ u32 pad0;
+ u32 posted_index; /* 0x10 */
+ u32 pad1;
+ u32 cq_index; /* 0x18 */
+ u32 pad2;
+ u32 enable; /* 0x20 */
+ u32 pad3;
+ u32 running; /* 0x28 */
+ u32 pad4;
+ u32 fetch_index; /* 0x30 */
+ u32 pad5;
+ u32 dca_value; /* 0x38 */
+ u32 pad6;
+ u32 error_interrupt_enable; /* 0x40 */
+ u32 pad7;
+ u32 error_interrupt_offset; /* 0x48 */
+ u32 pad8;
+ u32 error_status; /* 0x50 */
+ u32 pad9;
+};
+
+struct vnic_wq_buf {
+ struct vnic_wq_buf *next;
+ dma_addr_t dma_addr;
+ void *os_buf;
+ unsigned int len;
+ unsigned int index;
+ int sop;
+ void *desc;
+};
+
+/* Break the vnic_wq_buf allocations into blocks of 64 entries */
+#define VNIC_WQ_BUF_MIN_BLK_ENTRIES 32
+#define VNIC_WQ_BUF_DFLT_BLK_ENTRIES 64
+#define VNIC_WQ_BUF_BLK_ENTRIES(entries) \
+ ((unsigned int)(entries < VNIC_WQ_BUF_DFLT_BLK_ENTRIES) ? \
+ VNIC_WQ_BUF_MIN_BLK_ENTRIES : VNIC_WQ_BUF_DFLT_BLK_ENTRIES)
+#define VNIC_WQ_BUF_BLK_SZ \
+ (VNIC_WQ_BUF_DFLT_BLK_ENTRIES * sizeof(struct vnic_wq_buf))
+#define VNIC_WQ_BUF_BLKS_NEEDED(entries) \
+ DIV_ROUND_UP(entries, VNIC_WQ_BUF_DFLT_BLK_ENTRIES)
+#define VNIC_WQ_BUF_BLKS_NEEDED(entries) \
+ DIV_ROUND_UP(entries, VNIC_WQ_BUF_DFLT_BLK_ENTRIES)
+#define VNIC_WQ_BUF_BLKS_MAX VNIC_WQ_BUF_BLKS_NEEDED(4096)
+
+struct vnic_wq {
+ unsigned int index;
+ struct vnic_dev *vdev;
+ struct vnic_wq_ctrl __iomem *ctrl; /* memory-mapped */
+ struct vnic_dev_ring ring;
+ struct vnic_wq_buf *bufs[VNIC_WQ_BUF_BLKS_MAX];
+ struct vnic_wq_buf *to_use;
+ struct vnic_wq_buf *to_clean;
+ unsigned int pkts_outstanding;
+};
+
+static inline unsigned int vnic_wq_desc_avail(struct vnic_wq *wq)
+{
+ /* how many does SW own? */
+ return wq->ring.desc_avail;
+}
+
+static inline unsigned int vnic_wq_desc_used(struct vnic_wq *wq)
+{
+ /* how many does HW own? */
+ return wq->ring.desc_count - wq->ring.desc_avail - 1;
+}
+
+static inline void *vnic_wq_next_desc(struct vnic_wq *wq)
+{
+ return wq->to_use->desc;
+}
+
+static inline void vnic_wq_post(struct vnic_wq *wq,
+ void *os_buf, dma_addr_t dma_addr,
+ unsigned int len, int sop, int eop)
+{
+ struct vnic_wq_buf *buf = wq->to_use;
+
+ buf->sop = sop;
+ buf->os_buf = eop ? os_buf : NULL;
+ buf->dma_addr = dma_addr;
+ buf->len = len;
+
+ buf = buf->next;
+ if (eop) {
+ /* Adding write memory barrier prevents compiler and/or CPU
+ * reordering, thus avoiding descriptor posting before
+ * descriptor is initialized. Otherwise, hardware can read
+ * stale descriptor fields.
+ */
+ wmb();
+ iowrite32(buf->index, &wq->ctrl->posted_index);
+ }
+ wq->to_use = buf;
+
+ wq->ring.desc_avail--;
+}
+
+static inline void vnic_wq_service(struct vnic_wq *wq,
+ struct cq_desc *cq_desc, u16 completed_index,
+ void (*buf_service)(struct vnic_wq *wq,
+ struct cq_desc *cq_desc, struct vnic_wq_buf *buf, void *opaque),
+ void *opaque)
+{
+ struct vnic_wq_buf *buf;
+
+ buf = wq->to_clean;
+ while (1) {
+
+ (*buf_service)(wq, cq_desc, buf, opaque);
+
+ wq->ring.desc_avail++;
+
+ wq->to_clean = buf->next;
+
+ if (buf->index == completed_index)
+ break;
+
+ buf = wq->to_clean;
+ }
+}
+
+void vnic_wq_free(struct vnic_wq *wq);
+int vnic_wq_alloc(struct vnic_dev *vdev, struct vnic_wq *wq, unsigned int index,
+ unsigned int desc_count, unsigned int desc_size);
+int vnic_wq_devcmd2_alloc(struct vnic_dev *vdev, struct vnic_wq *wq,
+ unsigned int desc_count, unsigned int desc_size);
+void vnic_wq_init_start(struct vnic_wq *wq, unsigned int cq_index,
+ unsigned int fetch_index, unsigned int post_index,
+ unsigned int error_interrupt_enable,
+ unsigned int error_interrupt_offset);
+
+void vnic_wq_init(struct vnic_wq *wq, unsigned int cq_index,
+ unsigned int error_interrupt_enable,
+ unsigned int error_interrupt_offset);
+unsigned int vnic_wq_error_status(struct vnic_wq *wq);
+void vnic_wq_enable(struct vnic_wq *wq);
+int vnic_wq_disable(struct vnic_wq *wq);
+void vnic_wq_clean(struct vnic_wq *wq,
+ void (*buf_clean)(struct vnic_wq *wq, struct vnic_wq_buf *buf));
+#endif /* _VNIC_WQ_H_ */
diff --git a/drivers/scsi/snic/wq_enet_desc.h b/drivers/scsi/snic/wq_enet_desc.h
new file mode 100644
index 0000000..2caaf82
--- /dev/null
+++ b/drivers/scsi/snic/wq_enet_desc.h
@@ -0,0 +1,91 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef _WQ_ENET_DESC_H_
+#define _WQ_ENET_DESC_H_
+
+/* Ethernet work queue descriptor: 16B */
+struct wq_enet_desc {
+ u64 address;
+ u16 length;
+ u16 mss_loopback;
+ u16 header_length_flags;
+ u16 vlan_tag;
+};
+
+#define WQ_ENET_ADDR_BITS 64
+#define WQ_ENET_LEN_BITS 14
+#define WQ_ENET_LEN_MASK ((1 << WQ_ENET_LEN_BITS) - 1)
+#define WQ_ENET_MSS_BITS 14
+#define WQ_ENET_MSS_MASK ((1 << WQ_ENET_MSS_BITS) - 1)
+#define WQ_ENET_MSS_SHIFT 2
+#define WQ_ENET_LOOPBACK_SHIFT 1
+#define WQ_ENET_HDRLEN_BITS 10
+#define WQ_ENET_HDRLEN_MASK ((1 << WQ_ENET_HDRLEN_BITS) - 1)
+#define WQ_ENET_FLAGS_OM_BITS 2
+#define WQ_ENET_FLAGS_OM_MASK ((1 << WQ_ENET_FLAGS_OM_BITS) - 1)
+#define WQ_ENET_FLAGS_EOP_SHIFT 12
+#define WQ_ENET_FLAGS_CQ_ENTRY_SHIFT 13
+#define WQ_ENET_FLAGS_FCOE_ENCAP_SHIFT 14
+#define WQ_ENET_FLAGS_VLAN_TAG_INSERT_SHIFT 15
+
+#define WQ_ENET_OFFLOAD_MODE_CSUM 0
+#define WQ_ENET_OFFLOAD_MODE_RESERVED 1
+#define WQ_ENET_OFFLOAD_MODE_CSUM_L4 2
+#define WQ_ENET_OFFLOAD_MODE_TSO 3
+
+static inline void wq_enet_desc_enc(struct wq_enet_desc *desc,
+ u64 address, u16 length, u16 mss, u16 header_length,
+ u8 offload_mode, u8 eop, u8 cq_entry, u8 fcoe_encap,
+ u8 vlan_tag_insert, u16 vlan_tag, u8 loopback)
+{
+ desc->address = address;
+ desc->length = length & WQ_ENET_LEN_MASK;
+ desc->mss_loopback = (mss & WQ_ENET_MSS_MASK) <<
+ WQ_ENET_MSS_SHIFT | (loopback & 1) << WQ_ENET_LOOPBACK_SHIFT;
+ desc->header_length_flags = (header_length & WQ_ENET_HDRLEN_MASK) |
+ (offload_mode & WQ_ENET_FLAGS_OM_MASK) << WQ_ENET_HDRLEN_BITS |
+ (eop & 1) << WQ_ENET_FLAGS_EOP_SHIFT |
+ (cq_entry & 1) << WQ_ENET_FLAGS_CQ_ENTRY_SHIFT |
+ (fcoe_encap & 1) << WQ_ENET_FLAGS_FCOE_ENCAP_SHIFT |
+ (vlan_tag_insert & 1) << WQ_ENET_FLAGS_VLAN_TAG_INSERT_SHIFT;
+ desc->vlan_tag = vlan_tag;
+}
+
+static inline void wq_enet_desc_dec(struct wq_enet_desc *desc,
+ u64 *address, u16 *length, u16 *mss, u16 *header_length,
+ u8 *offload_mode, u8 *eop, u8 *cq_entry, u8 *fcoe_encap,
+ u8 *vlan_tag_insert, u16 *vlan_tag, u8 *loopback)
+{
+ *address = desc->address;
+ *length = desc->length & WQ_ENET_LEN_MASK;
+ *mss = (desc->mss_loopback >> WQ_ENET_MSS_SHIFT) & WQ_ENET_MSS_MASK;
+ *loopback = (u8)((desc->mss_loopback >> WQ_ENET_LOOPBACK_SHIFT) & 1);
+ *header_length = desc->header_length_flags & WQ_ENET_HDRLEN_MASK;
+ *offload_mode = (u8)((desc->header_length_flags >>
+ WQ_ENET_HDRLEN_BITS) & WQ_ENET_FLAGS_OM_MASK);
+ *eop = (u8)((desc->header_length_flags >> WQ_ENET_FLAGS_EOP_SHIFT) & 1);
+ *cq_entry = (u8)((desc->header_length_flags >>
+ WQ_ENET_FLAGS_CQ_ENTRY_SHIFT) & 1);
+ *fcoe_encap = (u8)((desc->header_length_flags >>
+ WQ_ENET_FLAGS_FCOE_ENCAP_SHIFT) & 1);
+ *vlan_tag_insert = (u8)((desc->header_length_flags >>
+ WQ_ENET_FLAGS_VLAN_TAG_INSERT_SHIFT) & 1);
+ *vlan_tag = desc->vlan_tag;
+}
+
+#endif /* _WQ_ENET_DESC_H_ */
--
1.8.5.4
^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH v4 7/9] snic:Add sysfs entries to list stats and trace data
2015-04-09 11:49 [PATCH v4 0/9] snic:initial submission of snic driver for Cisco SCSI HBA Narsimhulu Musini
` (5 preceding siblings ...)
2015-04-09 11:49 ` [PATCH v4 6/9] snic:Add low level queuing interfaces Narsimhulu Musini
@ 2015-04-09 11:49 ` Narsimhulu Musini
2015-04-09 13:23 ` Hannes Reinecke
2015-04-09 11:49 ` [PATCH v4 8/9] snic:Add event tracing to capture IO events Narsimhulu Musini
2015-04-09 11:49 ` [PATCH v4 9/9] snic:Add Makefile, patch Kconfig, MAINTAINERS Narsimhulu Musini
8 siblings, 1 reply; 24+ messages in thread
From: Narsimhulu Musini @ 2015-04-09 11:49 UTC (permalink / raw)
To: JBottomley, linux-scsi, hare, hch; +Cc: Narsimhulu Musini, Sesidhar Baddela
snic_stats.h contains stats structre definitions for various events
in snic driver.
snic_debugfs.c contains setup and cleanup of sysfs entries for listing stats,
resetting stats, enabling/disabling trace, and listing trace data.
Signed-off-by: Narsimhulu Musini <nmusini@cisco.com>
Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
---
drivers/scsi/snic/snic_debugfs.c | 562 +++++++++++++++++++++++++++++++++++++++
drivers/scsi/snic/snic_stats.h | 123 +++++++++
2 files changed, 685 insertions(+)
create mode 100644 drivers/scsi/snic/snic_debugfs.c
create mode 100644 drivers/scsi/snic/snic_stats.h
diff --git a/drivers/scsi/snic/snic_debugfs.c b/drivers/scsi/snic/snic_debugfs.c
new file mode 100644
index 0000000..31b2642
--- /dev/null
+++ b/drivers/scsi/snic/snic_debugfs.c
@@ -0,0 +1,562 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/module.h>
+#include <linux/errno.h>
+#include <linux/debugfs.h>
+
+#include "snic.h"
+
+/*
+ * snic_debugfs_init - Initialize debugfs for snic debug logging
+ *
+ * Description:
+ * When Debugfs is configured this routine sets up fnic debugfs
+ * filesystem. If not already created. this routine will crate the
+ * fnic directory and statistics directory for trace buffer and
+ * stats logging
+ */
+
+int
+snic_debugfs_init(void)
+{
+ int rc = -1;
+ struct dentry *de = NULL;
+
+ de = debugfs_create_dir("snic", NULL);
+ if (!de) {
+ SNIC_DBG("Cannot create debugfs root\n");
+
+ return rc;
+ }
+ snic_glob->trc_root = de;
+
+ de = debugfs_create_dir("statistics", snic_glob->trc_root);
+ if (!de) {
+ SNIC_DBG("Cannot create Statistics directory\n");
+
+ return rc;
+ }
+ snic_glob->stats_root = de;
+
+ rc = 0;
+
+ return rc;
+} /* end of snic_debugfs_init */
+
+/*
+ * snic_debugfs_term - Tear down debugfs intrastructure
+ *
+ * Description:
+ * When Debufs is configured this routine removes debugfs file system
+ * elements that are specific to snic
+ */
+void
+snic_debugfs_term(void)
+{
+ debugfs_remove(snic_glob->stats_root);
+ snic_glob->stats_root = NULL;
+
+ debugfs_remove(snic_glob->trc_root);
+ snic_glob->trc_root = NULL;
+}
+
+/*
+ * snic_reset_stats_open - Open the reset_stats file
+ */
+static int
+snic_reset_stats_open(struct inode *inode, struct file *filp)
+{
+ SNIC_BUG_ON(!inode->i_private);
+ filp->private_data = inode->i_private;
+
+ return 0;
+}
+
+/*
+ * snic_reset_stats_read - Read a reset_stats debugfs file
+ * @filp: The file pointer to read from.
+ * @ubuf: The buffer tocopy the data to.
+ * @cnt: The number of bytes to read.
+ * @ppos: The position in the file to start reading frm.
+ *
+ * Description:
+ * This routine reads value of variable reset_stats
+ * and stores into local @buf. It will start reading file @ppos and
+ * copy up to @cnt of data to @ubuf from @buf.
+ *
+ * Returns:
+ * This function returns the amount of data that was read.
+ */
+static ssize_t
+snic_reset_stats_read(struct file *filp,
+ char __user *ubuf,
+ size_t cnt,
+ loff_t *ppos)
+{
+ struct snic *snic = (struct snic *) filp->private_data;
+ char buf[64];
+ int len;
+
+ len = sprintf(buf, "%u\n", snic->reset_stats);
+
+ return simple_read_from_buffer(ubuf, cnt, ppos, buf, len);
+}
+
+/*
+ * snic_reset_stats_write - Write to reset_stats debugfs file
+ * @filp: The file pointer to write from
+ * @ubuf: The buffer to copy the data from.
+ * @cnt: The number of bytes to write.
+ * @ppos: The position in the file to start writing to.
+ *
+ * Description:
+ * This routine writes data from user buffer @ubuf to buffer @buf and
+ * resets cumulative stats of snic.
+ *
+ * Returns:
+ * This function returns the amount of data that was written.
+ */
+static ssize_t
+snic_reset_stats_write(struct file *filp,
+ const char __user *ubuf,
+ size_t cnt,
+ loff_t *ppos)
+{
+ struct snic *snic = (struct snic *) filp->private_data;
+ struct snic_stats *stats = &snic->s_stats;
+ u64 *io_stats_p = (u64 *) &stats->io;
+ u64 *fw_stats_p = (u64 *) &stats->fw;
+ char buf[64];
+ unsigned long val;
+ int ret;
+
+ if (cnt >= sizeof(buf))
+ return -EINVAL;
+
+ if (copy_from_user(&buf, ubuf, cnt))
+ return -EFAULT;
+
+ /* TODO: null character and not 0 */
+ buf[cnt] = 0;
+
+ ret = kstrtoul(buf, 10, &val);
+ if (ret < 0)
+ return ret;
+
+ snic->reset_stats = val;
+
+ if (snic->reset_stats) {
+ /* Skip variable is used to avoid descrepancies to Num IOs
+ * and IO Completions stats. Skip incrementing No IO Compls
+ * for pending active IOs after reset_stats
+ */
+ atomic64_set(&snic->io_cmpl_skip,
+ atomic64_read(&stats->io.active));
+ memset(&stats->abts, 0, sizeof(struct snic_abort_stats));
+ memset(&stats->reset, 0, sizeof(struct snic_reset_stats));
+ memset(&stats->misc, 0, sizeof(struct snic_misc_stats));
+ memset(io_stats_p+1,
+ 0,
+ sizeof(struct snic_io_stats) - sizeof(u64));
+ memset(fw_stats_p+1,
+ 0,
+ sizeof(struct snic_fw_stats) - sizeof(u64));
+ }
+
+ (*ppos)++;
+
+ SNIC_HOST_INFO(snic->shost, "Reset Op: Driver statistics.\n");
+
+ return cnt;
+}
+
+static int
+snic_reset_stats_release(struct inode *inode, struct file *filp)
+{
+ filp->private_data = NULL;
+
+ return 0;
+}
+
+/*
+ * snic_stats_show - Formats and prints per host specific driver stats.
+ */
+static int
+snic_stats_show(struct seq_file *sfp, void *data)
+{
+ struct snic *snic = (struct snic *) sfp->private;
+ struct snic_stats *stats = &snic->s_stats;
+ struct timespec last_isr_tms, last_ack_tms;
+ u64 maxio_tm;
+ int i;
+
+ /* Dump IO Stats */
+ seq_printf(sfp,
+ "------------------------------------------\n"
+ "\t\t IO Statistics\n"
+ "------------------------------------------\n");
+
+ maxio_tm = (u64) atomic64_read(&stats->io.max_time);
+ seq_printf(sfp,
+ "Active IOs : %lld\n"
+ "Max Active IOs : %lld\n"
+ "Total IOs : %lld\n"
+ "IOs Completed : %lld\n"
+ "IOs Failed : %lld\n"
+ "IOs Not Found : %lld\n"
+ "Memory Alloc Failures : %lld\n"
+ "REQs Null : %lld\n"
+ "SCSI Cmd Pointers Null : %lld\n"
+ "Max SGL for any IO : %lld\n"
+ "Max IO Size : %lld Sectors\n"
+ "Max Queuing Time : %lld\n"
+ "Max Completion Time : %lld\n"
+ "Max IO Process Time(FW) : %lld (%u msec)\n",
+ (u64) atomic64_read(&stats->io.active),
+ (u64) atomic64_read(&stats->io.max_active),
+ (u64) atomic64_read(&stats->io.num_ios),
+ (u64) atomic64_read(&stats->io.compl),
+ (u64) atomic64_read(&stats->io.fail),
+ (u64) atomic64_read(&stats->io.io_not_found),
+ (u64) atomic64_read(&stats->io.alloc_fail),
+ (u64) atomic64_read(&stats->io.req_null),
+ (u64) atomic64_read(&stats->io.sc_null),
+ (u64) atomic64_read(&stats->io.max_sgl),
+ (u64) atomic64_read(&stats->io.max_io_sz),
+ (u64) atomic64_read(&stats->io.max_qtime),
+ (u64) atomic64_read(&stats->io.max_cmpl_time),
+ maxio_tm,
+ jiffies_to_msecs(maxio_tm));
+
+ seq_puts(sfp, "\nSGL Counters\n");
+
+ for (i = 0; i < SNIC_MAX_SG_DESC_CNT; i++) {
+ seq_printf(sfp,
+ "%10lld ",
+ (u64) atomic64_read(&stats->io.sgl_cnt[i]));
+
+ if ((i + 1) % 8 == 0)
+ seq_puts(sfp, "\n");
+ }
+
+ /* Dump Abort Stats */
+ seq_printf(sfp,
+ "\n-------------------------------------------\n"
+ "\t\t Abort Statistics\n"
+ "---------------------------------------------\n");
+
+ seq_printf(sfp,
+ "Aborts : %lld\n"
+ "Aborts Fail : %lld\n"
+ "Aborts Driver Timeout : %lld\n"
+ "Abort FW Timeout : %lld\n"
+ "Abort IO NOT Found : %lld\n",
+ (u64) atomic64_read(&stats->abts.num),
+ (u64) atomic64_read(&stats->abts.fail),
+ (u64) atomic64_read(&stats->abts.drv_tmo),
+ (u64) atomic64_read(&stats->abts.fw_tmo),
+ (u64) atomic64_read(&stats->abts.io_not_found));
+
+ /* Dump Reset Stats */
+ seq_printf(sfp,
+ "\n-------------------------------------------\n"
+ "\t\t Reset Statistics\n"
+ "---------------------------------------------\n");
+
+ seq_printf(sfp,
+ "HBA Resets : %lld\n"
+ "HBA Reset Cmpls : %lld\n"
+ "HBA Reset Fail : %lld\n",
+ (u64) atomic64_read(&stats->reset.hba_resets),
+ (u64) atomic64_read(&stats->reset.hba_reset_cmpl),
+ (u64) atomic64_read(&stats->reset.hba_reset_fail));
+
+ /* Dump Firmware Stats */
+ seq_printf(sfp,
+ "\n-------------------------------------------\n"
+ "\t\t Firmware Statistics\n"
+ "---------------------------------------------\n");
+
+ seq_printf(sfp,
+ "Active FW Requests : %lld\n"
+ "Max FW Requests : %lld\n"
+ "FW Out Of Resource Errs : %lld\n"
+ "FW IO Errors : %lld\n"
+ "FW SCSI Errors : %lld\n",
+ (u64) atomic64_read(&stats->fw.actv_reqs),
+ (u64) atomic64_read(&stats->fw.max_actv_reqs),
+ (u64) atomic64_read(&stats->fw.out_of_res),
+ (u64) atomic64_read(&stats->fw.io_errs),
+ (u64) atomic64_read(&stats->fw.scsi_errs));
+
+
+ /* Dump Miscellenous Stats */
+ seq_printf(sfp,
+ "\n---------------------------------------------\n"
+ "\t\t Other Statistics\n"
+ "\n---------------------------------------------\n");
+
+ jiffies_to_timespec(stats->misc.last_isr_time, &last_isr_tms);
+ jiffies_to_timespec(stats->misc.last_ack_time, &last_ack_tms);
+
+ seq_printf(sfp,
+ "Last ISR Time : %llu (%8lu.%8lu)\n"
+ "Last Ack Time : %llu (%8lu.%8lu)\n"
+ "ISRs : %llu\n"
+ "Max CQ Entries : %lld\n"
+ "Data Count Mismatch : %lld\n"
+ "IOs w/ Timeout Status : %lld\n"
+ "IOs w/ Aborted Status : %lld\n"
+ "IOs w/ SGL Invalid Stat : %lld\n"
+ "WQ Desc Alloc Fail : %lld\n"
+ "Queue Full : %lld\n"
+ "Target Not Ready : %lld\n",
+ (u64) stats->misc.last_isr_time,
+ last_isr_tms.tv_sec, last_isr_tms.tv_nsec,
+ (u64)stats->misc.last_ack_time,
+ last_ack_tms.tv_sec, last_ack_tms.tv_nsec,
+ (u64) atomic64_read(&stats->misc.isr_cnt),
+ (u64) atomic64_read(&stats->misc.max_cq_ents),
+ (u64) atomic64_read(&stats->misc.data_cnt_mismat),
+ (u64) atomic64_read(&stats->misc.io_tmo),
+ (u64) atomic64_read(&stats->misc.io_aborted),
+ (u64) atomic64_read(&stats->misc.sgl_inval),
+ (u64) atomic64_read(&stats->misc.wq_alloc_fail),
+ (u64) atomic64_read(&stats->misc.qfull),
+ (u64) atomic64_read(&stats->misc.tgt_not_rdy));
+
+ return 0;
+}
+
+/*
+ * snic_stats_open - Open the stats file for specific host
+ *
+ * Description:
+ * This routine opens a debugfs file stats of specific host
+ */
+static int
+snic_stats_open(struct inode *inode, struct file *filp)
+{
+ return single_open(filp, snic_stats_show, inode->i_private);
+}
+
+static const struct file_operations snic_stats_fops __read_mostly = {
+ .owner = THIS_MODULE,
+ .open = snic_stats_open,
+ .read = seq_read,
+ .llseek = seq_lseek,
+ .release = single_release,
+};
+
+static const struct file_operations snic_reset_stats_fops __read_mostly = {
+ .owner = THIS_MODULE,
+ .open = snic_reset_stats_open,
+ .read = snic_reset_stats_read,
+ .write = snic_reset_stats_write,
+ .release = snic_reset_stats_release,
+};
+
+/*
+ * snic_stats_init - Initialize stats struct and create stats file
+ * per snic
+ *
+ * Description:
+ * When debugfs is cofigured this routine sets up the stats file per snic
+ * It will create file stats and reset_stats under statistics/host# directory
+ * to log per snic stats
+ */
+int
+snic_stats_debugfs_init(struct snic *snic)
+{
+ int rc = -1;
+ char name[16];
+ struct dentry *de = NULL;
+
+ snprintf(name, sizeof(name), "host%d", snic->shost->host_no);
+ if (!snic_glob->stats_root) {
+ SNIC_DBG("snic_stats root doesn't exist\n");
+
+ return rc;
+ }
+
+ de = debugfs_create_dir(name, snic_glob->stats_root);
+ if (!de) {
+ SNIC_DBG("Cannot create host directory\n");
+
+ return rc;
+ }
+ snic->stats_host = de;
+
+ /* TODO: do we need write permissions ?? */
+ de = debugfs_create_file("stats",
+ S_IFREG|S_IRUGO|S_IWUSR,
+ snic->stats_host,
+ snic,
+ &snic_stats_fops);
+ if (!de) {
+ SNIC_DBG("Cannot create host's stats file\n");
+
+ return rc;
+ }
+ snic->stats_file = de;
+
+ de = debugfs_create_file("reset_stats",
+ S_IFREG|S_IRUGO|S_IWUSR,
+ snic->stats_host,
+ snic,
+ &snic_reset_stats_fops);
+
+ if (!de) {
+ SNIC_DBG("Cannot create host's reset_stats file\n");
+
+ return rc;
+ }
+ snic->reset_stats_file = de;
+ rc = 0;
+
+ return rc;
+} /* end of snic_stats_debugfs_init */
+
+/*
+ * snic_stats_debugfs_remove - Tear down debugfs infrastructure of stats
+ *
+ * Description:
+ * When Debufs is configured this routine removes debugfs file system
+ * elements that are specific to to snic stats
+ */
+void
+snic_stats_debugfs_remove(struct snic *snic)
+{
+ debugfs_remove(snic->stats_file);
+ snic->stats_file = NULL;
+
+ debugfs_remove(snic->reset_stats_file);
+ snic->reset_stats_file = NULL;
+
+ debugfs_remove(snic->stats_host);
+ snic->stats_host = NULL;
+}
+
+/* Trace Facility related API */
+static void *
+snic_trc_seq_start(struct seq_file *sfp, loff_t *pos)
+{
+ return &snic_glob->trc;
+}
+
+static void *
+snic_trc_seq_next(struct seq_file *sfp, void *data, loff_t *pos)
+{
+ return NULL;
+}
+
+static void
+snic_trc_seq_stop(struct seq_file *sfp, void *data)
+{
+}
+
+#define SNIC_TRC_PBLEN 256
+static int
+snic_trc_seq_show(struct seq_file *sfp, void *data)
+{
+ char buf[SNIC_TRC_PBLEN];
+
+ if (snic_get_trc_data(buf, SNIC_TRC_PBLEN) > 0)
+ seq_printf(sfp, "%s\n", buf);
+
+ return 0;
+}
+
+static const struct seq_operations snic_trc_seq_ops __read_mostly = {
+ .start = snic_trc_seq_start,
+ .next = snic_trc_seq_next,
+ .stop = snic_trc_seq_stop,
+ .show = snic_trc_seq_show,
+};
+
+static int
+snic_trc_open(struct inode *inode, struct file *filp)
+{
+ return seq_open(filp, &snic_trc_seq_ops);
+}
+
+static const struct file_operations snic_trc_fops __read_mostly = {
+ .owner = THIS_MODULE,
+ .open = snic_trc_open,
+ .read = seq_read,
+ .llseek = seq_lseek,
+ .release = seq_release,
+};
+
+/*
+ * snic_trc_debugfs_init : creates trace/tracing_enable files for trace
+ * under debugfs
+ */
+int
+snic_trc_debugfs_init(void)
+{
+ struct dentry *de = NULL;
+ int ret = -1;
+
+ if (!snic_glob->trc_root) {
+ SNIC_ERR("Debugfs root directory for snic doesn't exist.\n");
+
+ return ret;
+ }
+
+ de = debugfs_create_bool("tracing_enable",
+ S_IFREG | S_IRUGO | S_IWUSR,
+ snic_glob->trc_root,
+ &snic_glob->trc.enable);
+
+ if (!de) {
+ SNIC_ERR("Can't create trace_enable file.\n");
+
+ return ret;
+ }
+ snic_glob->trc.trc_enable = de;
+
+ de = debugfs_create_file("trace",
+ S_IFREG | S_IRUGO | S_IWUSR,
+ snic_glob->trc_root,
+ NULL,
+ &snic_trc_fops);
+
+ if (!de) {
+ SNIC_ERR("Cann't create trace file.\n");
+
+ return ret;
+ }
+ snic_glob->trc.trc_file = de;
+ ret = 0;
+
+ return ret;
+} /* end of snic_trc_debugfs_init */
+
+/*
+ * snic_trc_debugfs_term : cleans up the files created for trace under debugfs
+ */
+void
+snic_trc_debugfs_term(void)
+{
+ debugfs_remove(snic_glob->trc.trc_file);
+ snic_glob->trc.trc_file = NULL;
+
+ debugfs_remove(snic_glob->trc.trc_enable);
+ snic_glob->trc.trc_enable = NULL;
+}
diff --git a/drivers/scsi/snic/snic_stats.h b/drivers/scsi/snic/snic_stats.h
new file mode 100644
index 0000000..11e6148
--- /dev/null
+++ b/drivers/scsi/snic/snic_stats.h
@@ -0,0 +1,123 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef __SNIC_STATS_H
+#define __SNIC_STATS_H
+
+struct snic_io_stats {
+ atomic64_t active; /* Active IOs */
+ atomic64_t max_active; /* Max # active IOs */
+ atomic64_t max_sgl; /* Max # SGLs for any IO */
+ atomic64_t max_time; /* Max time to process IO */
+ atomic64_t max_qtime; /* Max time to Queue the IO */
+ atomic64_t max_cmpl_time; /* Max time to complete the IO */
+ atomic64_t sgl_cnt[SNIC_MAX_SG_DESC_CNT]; /* SGL Counters */
+ atomic64_t max_io_sz; /* Max IO Size */
+ atomic64_t compl; /* IO Completions */
+ atomic64_t fail; /* IO Failures */
+ atomic64_t req_null; /* req or req info is NULL */
+ atomic64_t alloc_fail; /* Alloc Failures */
+ atomic64_t sc_null;
+ atomic64_t io_not_found; /* IO Not Found */
+ atomic64_t num_ios; /* Number of IOs */
+};
+
+struct snic_abort_stats {
+ atomic64_t num; /* Abort counter */
+ atomic64_t fail; /* Abort Failure Counter */
+ atomic64_t drv_tmo; /* Abort Driver Timeouts */
+ atomic64_t fw_tmo; /* Abort Firmware Timeouts */
+ atomic64_t io_not_found;/* Abort IO Not Found */
+};
+
+struct snic_reset_stats {
+ atomic64_t dev_resets; /* Device Reset Counter */
+ atomic64_t dev_reset_fail; /* Device Reset Failures */
+ atomic64_t dev_reset_aborts; /* Device Reset Aborts */
+ atomic64_t dev_reset_tmo; /* Device Reset Timeout */
+ atomic64_t dev_reset_terms; /* Device Reset terminate */
+ atomic64_t hba_resets; /* hba/firmware resets */
+ atomic64_t hba_reset_cmpl; /* hba/firmware reset completions */
+ atomic64_t hba_reset_fail; /* hba/firmware failures */
+ atomic64_t snic_resets; /* snic resets */
+ atomic64_t snic_reset_compl; /* snic reset completions */
+ atomic64_t snic_reset_fail; /* snic reset failures */
+};
+
+struct snic_fw_stats {
+ atomic64_t actv_reqs; /* Active Requests */
+ atomic64_t max_actv_reqs; /* Max Active Requests */
+ atomic64_t out_of_res; /* Firmware Out Of Resources */
+ atomic64_t io_errs; /* Firmware IO Firmware Errors */
+ atomic64_t scsi_errs; /* Target hits check condition */
+};
+
+struct snic_misc_stats {
+ u64 last_isr_time;
+ u64 last_ack_time;
+ atomic64_t isr_cnt;
+ atomic64_t max_cq_ents; /* Max CQ Entries */
+ atomic64_t data_cnt_mismat; /* Data Count Mismatch */
+ atomic64_t io_tmo;
+ atomic64_t io_aborted;
+ atomic64_t sgl_inval; /* SGL Invalid */
+ atomic64_t abts_wq_alloc_fail; /* Abort Path WQ desc alloc failure */
+ atomic64_t devrst_wq_alloc_fail;/* Device Reset - WQ desc alloc fail */
+ atomic64_t wq_alloc_fail; /* IO WQ desc alloc failure */
+ atomic64_t no_icmnd_itmf_cmpls;
+ atomic64_t io_under_run;
+ atomic64_t qfull;
+ atomic64_t tgt_not_rdy;
+};
+
+struct snic_stats {
+ struct snic_io_stats io;
+ struct snic_abort_stats abts;
+ struct snic_reset_stats reset;
+ struct snic_fw_stats fw;
+ struct snic_misc_stats misc;
+ atomic64_t io_cmpl_skip;
+};
+
+int snic_stats_debugfs_init(struct snic *);
+void snic_stats_debugfs_remove(struct snic *);
+
+/* Auxillary function to update active IO counter */
+static inline void
+snic_stats_update_active_ios(struct snic_stats *s_stats)
+{
+ struct snic_io_stats *io = &s_stats->io;
+ u32 nr_active_ios;
+
+ nr_active_ios = atomic64_inc_return(&io->active);
+ if (atomic64_read(&io->max_active) < nr_active_ios)
+ atomic64_set(&io->max_active, nr_active_ios);
+
+ atomic64_inc(&io->num_ios);
+}
+
+/* Auxillary function to update IO completion counter */
+static inline void
+snic_stats_update_io_cmpl(struct snic_stats *s_stats)
+{
+ atomic64_dec(&s_stats->io.active);
+ if (unlikely(atomic64_read(&s_stats->io_cmpl_skip)))
+ atomic64_dec(&s_stats->io_cmpl_skip);
+ else
+ atomic64_inc(&s_stats->io.compl);
+}
+#endif /* __SNIC_STATS_H */
--
1.8.5.4
^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH v4 8/9] snic:Add event tracing to capture IO events.
2015-04-09 11:49 [PATCH v4 0/9] snic:initial submission of snic driver for Cisco SCSI HBA Narsimhulu Musini
` (6 preceding siblings ...)
2015-04-09 11:49 ` [PATCH v4 7/9] snic:Add sysfs entries to list stats and trace data Narsimhulu Musini
@ 2015-04-09 11:49 ` Narsimhulu Musini
2015-04-09 13:24 ` Hannes Reinecke
2015-04-09 11:49 ` [PATCH v4 9/9] snic:Add Makefile, patch Kconfig, MAINTAINERS Narsimhulu Musini
8 siblings, 1 reply; 24+ messages in thread
From: Narsimhulu Musini @ 2015-04-09 11:49 UTC (permalink / raw)
To: JBottomley, linux-scsi, hare, hch; +Cc: Narsimhulu Musini, Sesidhar Baddela
snic_trc.h contains global trace structure definitions for snic driver
snic_trc.c adds tracing functionality to capture various IO events.
It maintains global trace buffer to maintain recent history of IO events.
It helps to understand the sequence of events prior to particular IO event,
or hung, panic, etc,.
Signed-off-by: Narsimhulu Musini <nmusini@cisco.com>
Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
---
* v2
- Added compile time macro SNIC_DEBUG_FS to handle debugfs dependent functionality.
drivers/scsi/snic/snic_trc.c | 182 +++++++++++++++++++++++++++++++++++++++++++
drivers/scsi/snic/snic_trc.h | 121 ++++++++++++++++++++++++++++
2 files changed, 303 insertions(+)
create mode 100644 drivers/scsi/snic/snic_trc.c
create mode 100644 drivers/scsi/snic/snic_trc.h
diff --git a/drivers/scsi/snic/snic_trc.c b/drivers/scsi/snic/snic_trc.c
new file mode 100644
index 0000000..c9e6f96
--- /dev/null
+++ b/drivers/scsi/snic/snic_trc.c
@@ -0,0 +1,182 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/module.h>
+#include <linux/mempool.h>
+#include <linux/errno.h>
+
+#include "snic_io.h"
+#include "snic.h"
+
+/*
+ * snic_get_trc_buf : Allocates a trace record and returns.
+ */
+struct snic_trc_data *
+snic_get_trc_buf(void)
+{
+ struct snic_trc *trc = &snic_glob->trc;
+ struct snic_trc_data *td = NULL;
+ unsigned long flags;
+
+ spin_lock_irqsave(&trc->lock, flags);
+ td = &trc->buf[trc->wr_idx];
+ trc->wr_idx++;
+
+ if (trc->wr_idx == trc->max_idx)
+ trc->wr_idx = 0;
+
+ if (trc->wr_idx != trc->rd_idx) {
+ spin_unlock_irqrestore(&trc->lock, flags);
+
+ goto end;
+ }
+
+ trc->rd_idx++;
+ if (trc->rd_idx == trc->max_idx)
+ trc->rd_idx = 0;
+
+ td->ts = 0; /* Marker for checking the record, for complete data*/
+ spin_unlock_irqrestore(&trc->lock, flags);
+
+end:
+
+ return td;
+} /* end of snic_get_trc_buf */
+
+/*
+ * snic_fmt_trc_data : Formats trace data for printing.
+ */
+static int
+snic_fmt_trc_data(struct snic_trc_data *td, char *buf, int buf_sz)
+{
+ int len = 0;
+ struct timespec tmspec;
+
+ jiffies_to_timespec(td->ts, &tmspec);
+
+ len += snprintf(buf, buf_sz,
+ "%lu.%10lu %-25s %3d %4x %16llx %16llx %16llx %16llx %16llx\n",
+ tmspec.tv_sec,
+ tmspec.tv_nsec,
+ td->fn,
+ td->hno,
+ td->tag,
+ td->data[0], td->data[1], td->data[2], td->data[3],
+ td->data[4]);
+
+ return len;
+} /* end of snic_fmt_trc_data */
+
+/*
+ * snic_get_trc_data : Returns a formatted trace buffer.
+ */
+int
+snic_get_trc_data(char *buf, int buf_sz)
+{
+ struct snic_trc_data *td = NULL;
+ struct snic_trc *trc = &snic_glob->trc;
+ unsigned long flags;
+
+ spin_lock_irqsave(&trc->lock, flags);
+ if (trc->rd_idx == trc->wr_idx) {
+ spin_unlock_irqrestore(&trc->lock, flags);
+
+ return -1;
+ }
+ td = &trc->buf[trc->rd_idx];
+
+ if (td->ts == 0) {
+ /* write in progress. */
+ spin_unlock_irqrestore(&trc->lock, flags);
+
+ return -1;
+ }
+
+ trc->rd_idx++;
+ if (trc->rd_idx == trc->max_idx)
+ trc->rd_idx = 0;
+ spin_unlock_irqrestore(&trc->lock, flags);
+
+ return snic_fmt_trc_data(td, buf, buf_sz);
+} /* end of snic_get_trc_data */
+
+/*
+ * snic_trc_init() : Configures Trace Functionality for snic.
+ */
+int
+snic_trc_init(void)
+{
+ struct snic_trc *trc = &snic_glob->trc;
+ void *tbuf = NULL;
+ int tbuf_sz = 0, ret;
+
+ BUILD_BUG_ON(sizeof(struct snic_trc_data) != 64);
+
+ tbuf_sz = (snic_trace_max_pages * PAGE_SIZE);
+ tbuf = vmalloc(tbuf_sz);
+ if (!tbuf) {
+ SNIC_ERR("Failed to Allocate Trace Buffer Size. %d\n", tbuf_sz);
+ SNIC_ERR("Trace Facility not enabled.\n");
+ ret = -ENOMEM;
+
+ return ret;
+ }
+
+ memset(tbuf, 0, tbuf_sz);
+ trc->buf = (struct snic_trc_data *) tbuf;
+ spin_lock_init(&trc->lock);
+
+ ret = snic_trc_debugfs_init();
+ if (ret) {
+ SNIC_ERR("Failed to create Debugfs Files.\n");
+
+ goto error;
+ }
+
+ trc->max_idx = (tbuf_sz / SNIC_TRC_ENTRY_SZ);
+ trc->rd_idx = trc->wr_idx = 0;
+ trc->enable = 1;
+ SNIC_INFO("Trace Facility Enabled.\n Trace Buffer SZ %lu Pages.\n",
+ tbuf_sz / PAGE_SIZE);
+ ret = 0;
+
+ return ret;
+
+error:
+ snic_trc_free();
+
+ return ret;
+} /* end of snic_trc_init */
+
+/*
+ * snic_trc_free : Releases the trace buffer and disables the tracing.
+ */
+void
+snic_trc_free(void)
+{
+ struct snic_trc *trc = &snic_glob->trc;
+
+ trc->enable = 0;
+ snic_trc_debugfs_term();
+
+ if (trc->buf) {
+ vfree(trc->buf);
+ trc->buf = NULL;
+ }
+
+ SNIC_INFO("Trace Facility Disabled.\n");
+} /* end of snic_trc_free */
diff --git a/drivers/scsi/snic/snic_trc.h b/drivers/scsi/snic/snic_trc.h
new file mode 100644
index 0000000..99de384
--- /dev/null
+++ b/drivers/scsi/snic/snic_trc.h
@@ -0,0 +1,121 @@
+/*
+ * Copyright 2014 Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef __SNIC_TRC_H
+#define __SNIC_TRC_H
+
+#ifdef SNIC_DEBUG_FS
+
+extern ssize_t simple_read_from_buffer(void __user *to,
+ size_t count,
+ loff_t *ppos,
+ const void *from,
+ size_t available);
+
+extern unsigned int snic_trace_max_pages;
+
+/* Global Data structure for trace to manage trace functionality */
+struct snic_trc_data {
+ u64 ts; /* Time Stamp */
+ char *fn; /* Ptr to Function Name */
+ u32 hno; /* SCSI Host ID */
+ u32 tag; /* Command Tag */
+ u64 data[5];
+} __attribute__((__packed__));
+
+#define SNIC_TRC_ENTRY_SZ 64 /* in Bytes */
+
+struct snic_trc {
+ spinlock_t lock;
+ struct snic_trc_data *buf; /* Trace Buffer */
+ u32 max_idx; /* Max Index into trace buffer */
+ u32 rd_idx;
+ u32 wr_idx;
+ u32 enable; /* Control Variable for Tracing */
+
+ struct dentry *trc_enable; /* debugfs file object */
+ struct dentry *trc_file;
+};
+
+int snic_trc_init(void);
+void snic_trc_free(void);
+int snic_trc_debugfs_init(void);
+void snic_trc_debugfs_term(void);
+struct snic_trc_data *snic_get_trc_buf(void);
+int snic_get_trc_data(char *buf, int buf_sz);
+
+int snic_debugfs_init(void);
+void snic_debugfs_term(void);
+
+static inline void
+snic_trace(char *fn, u16 hno, u32 tag, u64 d1, u64 d2, u64 d3, u64 d4, u64 d5)
+{
+ struct snic_trc_data *tr_rec = snic_get_trc_buf();
+
+ if (!tr_rec)
+ return;
+
+ tr_rec->fn = (char *)fn;
+ tr_rec->hno = hno;
+ tr_rec->tag = tag;
+ tr_rec->data[0] = d1;
+ tr_rec->data[1] = d2;
+ tr_rec->data[2] = d3;
+ tr_rec->data[3] = d4;
+ tr_rec->data[4] = d5;
+ tr_rec->ts = jiffies; /* Update time stamp at last */
+}
+
+#define SNIC_TRC(_hno, _tag, d1, d2, d3, d4, d5) \
+ do { \
+ if (unlikely(snic_glob->trc.enable)) \
+ snic_trace((char *)__func__, \
+ (u16)(_hno), \
+ (u32)(_tag), \
+ (u64)(d1), \
+ (u64)(d2), \
+ (u64)(d3), \
+ (u64)(d4), \
+ (u64)(d5)); \
+ } while (0)
+#else
+
+#define SNIC_TRC(_hno, _tag, d1, d2, d3, d4, d5) \
+ do { \
+ if (unlikely(snic_log_level & 0x2)) \
+ SNIC_DBG("SnicTrace: %s %2u %2u %llx %llx %llx %llx %llx", \
+ (char *)__func__, \
+ (u16)(_hno), \
+ (u32)(_tag), \
+ (u64)(d1), \
+ (u64)(d2), \
+ (u64)(d3), \
+ (u64)(d4), \
+ (u64)(d5)); \
+ } while (0);
+#endif /* end of SNIC_DEBUG_FS */
+
+#define SNIC_TRC_CMD(sc) \
+ ((u64)sc->cmnd[0] << 56 | (u64)sc->cmnd[7] << 40 | \
+ (u64)sc->cmnd[8] << 32 | (u64)sc->cmnd[2] << 24 | \
+ (u64)sc->cmnd[3] << 16 | (u64)sc->cmnd[4] << 8 | \
+ (u64)sc->cmnd[5])
+
+#define SNIC_TRC_CMD_STATE_FLAGS(sc) \
+ ((u64) CMD_FLAGS(sc) << 32 | CMD_STATE(sc))
+
+#endif /* end of __SNIC_TRC_H */
--
1.8.5.4
^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH v4 9/9] snic:Add Makefile, patch Kconfig, MAINTAINERS
2015-04-09 11:49 [PATCH v4 0/9] snic:initial submission of snic driver for Cisco SCSI HBA Narsimhulu Musini
` (7 preceding siblings ...)
2015-04-09 11:49 ` [PATCH v4 8/9] snic:Add event tracing to capture IO events Narsimhulu Musini
@ 2015-04-09 11:49 ` Narsimhulu Musini
2015-04-09 13:25 ` Hannes Reinecke
8 siblings, 1 reply; 24+ messages in thread
From: Narsimhulu Musini @ 2015-04-09 11:49 UTC (permalink / raw)
To: JBottomley, linux-scsi, hare, hch; +Cc: Narsimhulu Musini, Sesidhar Baddela
Kconfig for kbuild
Makefile to build snic module
Updated MAINTAINERS file
Signed-off-by: Narsimhulu Musini <nmusini@cisco.com>
Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
---
* v3
- Added additional config section (CONFIG_SNIC_DEBUG_FS) for enabling debugging
functionality.
* v2
- Added compile time flags for debugfs dependent functionality.
MAINTAINERS | 7 +++++++
drivers/scsi/Kconfig | 17 +++++++++++++++++
drivers/scsi/Makefile | 1 +
drivers/scsi/snic/Makefile | 21 +++++++++++++++++++++
4 files changed, 46 insertions(+)
create mode 100644 drivers/scsi/snic/Makefile
diff --git a/MAINTAINERS b/MAINTAINERS
index 2a97e05..368fb76 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -2536,6 +2536,13 @@ L: linux-scsi@vger.kernel.org
S: Supported
F: drivers/scsi/fnic/
+CISCO SCSI HBA DRIVER
+M: Narsimhulu Musini <nmusini@cisco.com>
+M: Sesidhar Baddela <sebaddel@cisco.com>
+L: linux-scsi@vger.kernel.org
+S: Supported
+F: drivers/scsi/snic/
+
CMPC ACPI DRIVER
M: Thadeu Lima de Souza Cascardo <cascardo@holoscopio.com>
M: Daniel Oliveira Nascimento <don@syst.com.br>
diff --git a/drivers/scsi/Kconfig b/drivers/scsi/Kconfig
index 9c92f41..a6d3901 100644
--- a/drivers/scsi/Kconfig
+++ b/drivers/scsi/Kconfig
@@ -634,6 +634,23 @@ config FCOE_FNIC
<file:Documentation/scsi/scsi.txt>.
The module will be called fnic.
+config SCSI_SNIC
+ tristate "Cisco SNIC Driver"
+ depends on PCI && SCSI && X86
+ help
+ This is support for the Cisco PCI-Express SCSI HBA.
+
+ To compile this driver as a module, choose M here and read
+ <file:Documentation/scsi/scsi.txt>.
+ The module will be called snic.
+
+config SCSI_SNIC_DEBUG_FS
+ bool "Cisco SNIC Driver Debugfs Support"
+ depends on SCSI_SNIC && DEBUG_FS
+ help
+ This enables to list debugging information from SNIC Driver
+ available via debugfs file system
+
config SCSI_DMX3191D
tristate "DMX3191D SCSI support"
depends on PCI && SCSI
diff --git a/drivers/scsi/Makefile b/drivers/scsi/Makefile
index 58158f1..f643942 100644
--- a/drivers/scsi/Makefile
+++ b/drivers/scsi/Makefile
@@ -39,6 +39,7 @@ obj-$(CONFIG_LIBFC) += libfc/
obj-$(CONFIG_LIBFCOE) += fcoe/
obj-$(CONFIG_FCOE) += fcoe/
obj-$(CONFIG_FCOE_FNIC) += fnic/
+obj-$(CONFIG_SCSI_SNIC) += snic/
obj-$(CONFIG_SCSI_BNX2X_FCOE) += libfc/ fcoe/ bnx2fc/
obj-$(CONFIG_ISCSI_TCP) += libiscsi.o libiscsi_tcp.o iscsi_tcp.o
obj-$(CONFIG_INFINIBAND_ISER) += libiscsi.o
diff --git a/drivers/scsi/snic/Makefile b/drivers/scsi/snic/Makefile
new file mode 100644
index 0000000..572102a
--- /dev/null
+++ b/drivers/scsi/snic/Makefile
@@ -0,0 +1,21 @@
+obj-$(CONFIG_SCSI_SNIC) += snic.o
+
+snic-y := \
+ snic_attrs.o \
+ snic_main.o \
+ snic_res.o \
+ snic_isr.o \
+ snic_ctl.o \
+ snic_io.o \
+ snic_scsi.o \
+ snic_disc.o \
+ vnic_cq.o \
+ vnic_intr.o \
+ vnic_dev.o \
+ vnic_wq.o
+
+ifeq ($(CONFIG_SCSI_SNIC_DEBUG_FS), y)
+ccflags-y += -DSNIC_DEBUG_FS
+snic-y += snic_debugfs.o \
+ snic_trc.o
+endif
--
1.8.5.4
^ permalink raw reply related [flat|nested] 24+ messages in thread
* Re: [PATCH v4 1/9] snic: snic module infrastructure
2015-04-09 11:49 ` [PATCH v4 1/9] snic: snic module infrastructure Narsimhulu Musini
@ 2015-04-09 12:40 ` Hannes Reinecke
0 siblings, 0 replies; 24+ messages in thread
From: Hannes Reinecke @ 2015-04-09 12:40 UTC (permalink / raw)
To: Narsimhulu Musini, JBottomley, linux-scsi, hch; +Cc: Sesidhar Baddela
On 04/09/2015 01:49 PM, Narsimhulu Musini wrote:
> snic_main.c contains module load and unload, global driver context,
> PCI Registration, PCI probe and remove, SCSI ML registration functionality.
>
> snic.h contains snic structure definition, snic global context, and
> prototypes.
>
> snic_attrs.c contains device attributes to list snic state, link state,
> and driver version under /sys/class/scsi_host/host<id>/
>
> Signed-off-by: Narsimhulu Musini <nmusini@cisco.com>
> Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
> ---
Reviewed-by: Hannes Reinecke <hare@suse.de>
Cheers,
Hannes
--
Dr. Hannes Reinecke zSeries & Storage
hare@suse.de +49 911 74053 688
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg
GF: J. Hawn, J. Guild, F. Imendörffer, HRB 16746 (AG Nürnberg)
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH v4 2/9] snic:Add interrupt, resource firmware interfaces
2015-04-09 11:49 ` [PATCH v4 2/9] snic:Add interrupt, resource firmware interfaces Narsimhulu Musini
@ 2015-04-09 12:46 ` Hannes Reinecke
0 siblings, 0 replies; 24+ messages in thread
From: Hannes Reinecke @ 2015-04-09 12:46 UTC (permalink / raw)
To: Narsimhulu Musini, JBottomley, linux-scsi, hch; +Cc: Sesidhar Baddela
On 04/09/2015 01:49 PM, Narsimhulu Musini wrote:
> snic_fwint.h contains firmware interface structures.
>
> snic_res.h contains firmware request initialization
>
> snic_res.c contains retrieval of resource configuration, and allocation,
> and initialization of HW Queues.
>
> snic_isr.c contains interrupt request, release, and handling
>
> Signed-off-by: Narsimhulu Musini <nmusini@cisco.com>
> Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Cheers,
Hannes
--
Dr. Hannes Reinecke zSeries & Storage
hare@suse.de +49 911 74053 688
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg
GF: J. Hawn, J. Guild, F. Imendörffer, HRB 16746 (AG Nürnberg)
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH v4 3/9] snic:Add meta request, handling of meta requests.
2015-04-09 11:49 ` [PATCH v4 3/9] snic:Add meta request, handling of meta requests Narsimhulu Musini
@ 2015-04-09 12:50 ` Hannes Reinecke
0 siblings, 0 replies; 24+ messages in thread
From: Hannes Reinecke @ 2015-04-09 12:50 UTC (permalink / raw)
To: Narsimhulu Musini, JBottomley, linux-scsi, hch; +Cc: Sesidhar Baddela
On 04/09/2015 01:49 PM, Narsimhulu Musini wrote:
> snic_io.h contains meta request structure definition
> meta request contains high level information about firmware requests.
> such as request information, size, SGLs.
>
> snic_io.c contains interfaces to handle meta request, firmware acknowledgment,
> and high level generic queueing interface.
>
> Signed-off-by: Narsimhulu Musini <nmusini@cisco.com>
> Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
> ---
Reviewed-by: Hannes Reinecke <hare@suse.de>
Cheers,
Hannes
--
Dr. Hannes Reinecke zSeries & Storage
hare@suse.de +49 911 74053 688
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg
GF: J. Hawn, J. Guild, F. Imendörffer, HRB 16746 (AG Nürnberg)
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH v4 4/9] snic:Add snic target discovery
2015-04-09 11:49 ` [PATCH v4 4/9] snic:Add snic target discovery Narsimhulu Musini
@ 2015-04-09 12:59 ` Hannes Reinecke
2015-04-13 5:25 ` Narsimhulu Musini (nmusini)
0 siblings, 1 reply; 24+ messages in thread
From: Hannes Reinecke @ 2015-04-09 12:59 UTC (permalink / raw)
To: Narsimhulu Musini, JBottomley, linux-scsi, hch; +Cc: Sesidhar Baddela
Hi Narsimhulu,
please find some comments inline.
On 04/09/2015 01:49 PM, Narsimhulu Musini wrote:
> snic_disc.h contains snic target structure definition.
>
> snic_disc.c contains target discovery, setup, lookup, and cleanup
>
> snic_ctl.c contains retrieval of snic capabilities includes
> max ios, size, SGs per request, and max concurrent requests.
>
> Signed-off-by: Narsimhulu Musini <nmusini@cisco.com>
> Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
> ---
> * v3
> - Cleaned up redundant comment.
>
> drivers/scsi/snic/snic_ctl.c | 276 +++++++++++++++++++
> drivers/scsi/snic/snic_disc.c | 602 ++++++++++++++++++++++++++++++++++++++++++
> drivers/scsi/snic/snic_disc.h | 124 +++++++++
> 3 files changed, 1002 insertions(+)
> create mode 100644 drivers/scsi/snic/snic_ctl.c
> create mode 100644 drivers/scsi/snic/snic_disc.c
> create mode 100644 drivers/scsi/snic/snic_disc.h
>
> diff --git a/drivers/scsi/snic/snic_ctl.c b/drivers/scsi/snic/snic_ctl.c
> new file mode 100644
> index 0000000..4bc55be
> --- /dev/null
> +++ b/drivers/scsi/snic/snic_ctl.c
> @@ -0,0 +1,276 @@
> +/*
> + * Copyright 2014 Cisco Systems, Inc. All rights reserved.
> + *
> + * This program is free software; you may redistribute it and/or modify
> + * it under the terms of the GNU General Public License as published by
> + * the Free Software Foundation; version 2 of the License.
> + *
> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
> + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
> + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
> + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
> + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
> + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
> + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
> + * SOFTWARE.
> + */
> +
> +#include <linux/errno.h>
> +#include <linux/pci.h>
> +#include <linux/slab.h>
> +
> +#include <linux/interrupt.h>
> +#include <linux/workqueue.h>
> +#include <linux/spinlock.h>
> +#include <linux/mempool.h>
> +#include <scsi/scsi_tcq.h>
> +#include <linux/ctype.h>
> +
> +#include "snic_io.h"
> +#include "snic.h"
> +#include "cq_enet_desc.h"
> +#include "snic_fwint.h"
> +
> +/*
> + * snic_handle_link : Handles link flaps.
> + */
> +void
> +snic_handle_link(struct work_struct *work)
> +{
> + struct snic *snic = container_of(work, struct snic, link_work);
> +
> + if (snic->config.xpt_type != SNIC_DAS) {
> + SNIC_HOST_INFO(snic->shost, "Link Event Received.\n");
> + SNIC_ASSERT_NOT_IMPL(1);
> +
> + return;
> + }
> +
> + snic->link_status = vnic_dev_link_status(snic->vdev);
> + snic->link_down_cnt = vnic_dev_link_down_cnt(snic->vdev);
> + SNIC_HOST_INFO(snic->shost, "Link Event: Link %s.\n",
> + ((snic->link_status) ? "Up" : "Down"));
> +}
> +
> +
> +/*
> + * snic_ver_enc : Encodes version str to int
> + * version string is similar to netmask string
> + */
> +static int
> +snic_ver_enc(const char *s)
> +{
> + int v[4] = {0};
> + int i = 0, x = 0;
> + char c;
> + const char *p = s;
> +
> + /* validate version string */
> + if ((strlen(s) > 15) || (strlen(s) < 7))
> + goto end;
> +
> + while ((c = *p++)) {
> + if (c == '.') {
> + i++;
> + continue;
> + }
> +
> + if (i > 4 || !isdigit(c))
> + goto end;
> +
> + v[i] = v[i] * 10 + (c - '0');
> + }
> +
> + /* validate sub version numbers */
> + for (i = 3; i >= 0; i--)
> + if (v[i] > 0xff)
> + goto end;
> +
> + x |= (v[0] << 24) | v[1] << 16 | v[2] << 8 | v[3];
> +
> +end:
> + if (x == 0) {
> + SNIC_ERR("Invalid version string [%s].\n", s);
> +
> + return -1;
> + }
> +
> + return x;
> +} /* end of snic_ver_enc */
> +
> +/*
> + * snic_qeueue_exch_ver_req :
> + *
> + * Queues Exchange Version Request, to communicate host information
> + * in return, it gets firmware version details
> + */
> +int
> +snic_queue_exch_ver_req(struct snic *snic)
> +{
> + struct snic_req_info *rqi = NULL;
> + struct snic_host_req *req = NULL;
> + int ret = 0;
> +
> + SNIC_HOST_INFO(snic->shost, "Exch Ver Req Preparing...\n");
> +
> + rqi = snic_req_init(snic, 0);
> + if (!rqi) {
> + SNIC_HOST_ERR(snic->shost,
> + "Queuing Exch Ver Req failed, err = %d\n",
> + ret);
> +
> + ret = -ENOMEM;
> + goto error;
> + }
> +
> + req = rqi_to_req(rqi);
> +
> + /* Initialize snic_host_req */
> + snic_io_hdr_enc(&req->hdr, SNIC_REQ_EXCH_VER, 0, SCSI_NO_TAG,
> + snic->config.hid, 0, (u64)rqi);
> + req->u.exch_ver.drvr_ver = snic_ver_enc(SNIC_DRV_VERSION);
> + req->u.exch_ver.os_type = SNIC_OS_LINUX;
> +
> + snic_handle_untagged_req(snic, rqi);
> +
> + ret = snic_queue_wq_desc(snic, req, sizeof(*req));
> + if (ret) {
> + snic_release_untagged_req(snic, rqi);
> + SNIC_HOST_ERR(snic->shost,
> + "Queuing Exch Ver Req failed, err = %d\n",
> + ret);
> + goto error;
> + }
> +
> + SNIC_HOST_INFO(snic->shost, "Exch Ver Req is issued. ret = %d\n", ret);
> +
> +error:
> + return ret;
> +} /* end of snic_queue_exch_ver_req */
> +
> +/*
> + * snic_io_exch_ver_cmpl_handler
> + */
> +int
> +snic_io_exch_ver_cmpl_handler(struct snic *snic, struct snic_fw_req *fwreq)
> +{
> + struct snic_req_info *rqi = NULL;
> + u8 typ, hdr_stat;
> + u32 cmnd_id, hid, max_sgs;
> + u64 ctx = 0;
> + unsigned long flags;
> + int ret = 0;
> +
> + SNIC_HOST_INFO(snic->shost, "Exch Ver Compl Received.\n");
> + snic_io_hdr_dec(&fwreq->hdr, &typ, &hdr_stat, &cmnd_id, &hid, &ctx);
> + SNIC_BUG_ON(snic->config.hid != hid);
> + rqi = (struct snic_req_info *) ctx;
> +
> + if (hdr_stat) {
> + SNIC_HOST_ERR(snic->shost,
> + "Exch Ver Completed w/ err status %d\n",
> + hdr_stat);
> +
> + goto exch_cmpl_end;
> + }
> +
> + spin_lock_irqsave(&snic->snic_lock, flags);
> + snic->fwinfo.fw_ver = fwreq->u.exch_ver_cmpl.version;
> + snic->fwinfo.hid = fwreq->u.exch_ver_cmpl.hid;
> + snic->fwinfo.max_concur_ios = fwreq->u.exch_ver_cmpl.max_concur_ios;
> + snic->fwinfo.max_sgs_per_cmd = fwreq->u.exch_ver_cmpl.max_sgs_per_cmd;
> + snic->fwinfo.max_io_sz = fwreq->u.exch_ver_cmpl.max_io_sz;
> + snic->fwinfo.max_tgts = fwreq->u.exch_ver_cmpl.max_tgts;
> + snic->fwinfo.io_tmo = fwreq->u.exch_ver_cmpl.io_timeout;
> +
> + SNIC_HOST_INFO(snic->shost,
> + "vers %u hid %u max_concur_ios %u max_sgs_per_cmd %u max_io_sz %u max_tgts %u fw tmo %u\n",
> + fwreq->u.exch_ver_cmpl.version,
> + fwreq->u.exch_ver_cmpl.hid,
> + fwreq->u.exch_ver_cmpl.max_concur_ios,
> + fwreq->u.exch_ver_cmpl.max_sgs_per_cmd,
> + fwreq->u.exch_ver_cmpl.max_io_sz,
> + fwreq->u.exch_ver_cmpl.max_tgts,
> + fwreq->u.exch_ver_cmpl.io_timeout);
> +
> + SNIC_HOST_INFO(snic->shost,
> + "HBA Capabilities = 0x%x\n",
> + fwreq->u.exch_ver_cmpl.hba_cap);
> +
> + /* Updating SGList size */
> + max_sgs = snic->fwinfo.max_sgs_per_cmd;
> + if (max_sgs && max_sgs < SNIC_MAX_SG_DESC_CNT) {
> + snic->shost->sg_tablesize = max_sgs;
> + SNIC_HOST_INFO(snic->shost, "Max SGs set to %d\n",
> + snic->shost->sg_tablesize);
> + } else if (max_sgs > snic->shost->sg_tablesize) {
> + SNIC_HOST_INFO(snic->shost,
> + "Target type %d Supports Larger Max SGList %d than driver's Max SG List %d.\n",
> + snic->config.xpt_type, max_sgs,
> + snic->shost->sg_tablesize);
> + }
> +
> + if (snic->shost->can_queue > snic->fwinfo.max_concur_ios)
> + snic->shost->can_queue = snic->fwinfo.max_concur_ios;
> +
> + snic->shost->max_sectors = snic->fwinfo.max_io_sz >> 9;
> + if (snic->fwinfo.wait)
> + complete(snic->fwinfo.wait);
> +
> + spin_unlock_irqrestore(&snic->snic_lock, flags);
> +
> +exch_cmpl_end:
> + snic_release_untagged_req(snic, rqi);
> +
> + SNIC_HOST_INFO(snic->shost, "Exch_cmpl Done, hdr_stat %d.\n", hdr_stat);
> +
> + return ret;
> +} /* end of snic_io_exch_ver_cmpl_handler */
> +
> +/*
> + * snic_get_conf
> + *
> + * Synchronous call, and Retrieves snic params.
> + */
> +int
> +snic_get_conf(struct snic *snic)
> +{
> + DECLARE_COMPLETION_ONSTACK(wait);
> + unsigned long flags;
> + int ret;
> + int nr_retries = 3;
> +
> + SNIC_HOST_INFO(snic->shost, "Retrieving snic params.\n");
> + spin_lock_irqsave(&snic->snic_lock, flags);
> + memset(&snic->fwinfo, 0, sizeof(snic->fwinfo));
> + snic->fwinfo.wait = &wait;
> + spin_unlock_irqrestore(&snic->snic_lock, flags);
> +
> + /* Additional delay to handle HW Resource initialization. */
> + msleep(50);
> +
> + /*
> + * Exch ver req can be ignored by FW, if HW Resource initialization
> + * is in progress, Hence retry.
> + */
> + do {
> + ret = snic_queue_exch_ver_req(snic);
> + if (ret)
> + return ret;
> +
> + wait_for_completion_timeout(&wait, msecs_to_jiffies(2000));
> + spin_lock_irqsave(&snic->snic_lock, flags);
> + ret = (snic->fwinfo.fw_ver != 0) ? 0 : -ETIMEDOUT;
> + if (ret)
> + SNIC_HOST_ERR(snic->shost,
> + "Failed to retrieve snic params,\n");
> +
> + /* Unset fwinfo.wait, on success or on last retry */
> + if (ret == 0 || nr_retries == 1)
> + snic->fwinfo.wait = NULL;
> +
> + spin_unlock_irqrestore(&snic->snic_lock, flags);
> + } while (ret && --nr_retries);
> +
> + return ret;
> +} /* end of snic_get_info */
> diff --git a/drivers/scsi/snic/snic_disc.c b/drivers/scsi/snic/snic_disc.c
> new file mode 100644
> index 0000000..b6b6cca
> --- /dev/null
> +++ b/drivers/scsi/snic/snic_disc.c
> @@ -0,0 +1,602 @@
> +/*
> + * Copyright 2014 Cisco Systems, Inc. All rights reserved.
> + *
> + * This program is free software; you may redistribute it and/or modify
> + * it under the terms of the GNU General Public License as published by
> + * the Free Software Foundation; version 2 of the License.
> + *
> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
> + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
> + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
> + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
> + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
> + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
> + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
> + * SOFTWARE.
> + */
> +
> +#include <linux/errno.h>
> +#include <linux/mempool.h>
> +
> +#include <scsi/scsi_tcq.h>
> +
> +#include "snic_disc.h"
> +#include "snic.h"
> +#include "snic_io.h"
> +
> +
> +/* snic target types */
> +static const char * const snic_tgt_type_str[] = {
> + [SNIC_TGT_DAS] = "DAS",
> + [SNIC_TGT_SAN] = "SAN",
> +};
> +
> +static inline const char *
> +snic_tgt_type_to_str(int typ)
> +{
> + return ((typ > SNIC_TGT_NONE && typ <= SNIC_TGT_SAN) ?
> + snic_tgt_type_str[typ] : "Unknown");
> +}
> +
> +static const char * const snic_tgt_state_str[] = {
> + [SNIC_TGT_STAT_INIT] = "INIT",
> + [SNIC_TGT_STAT_ONLINE] = "ONLINE",
> + [SNIC_TGT_STAT_OFFLINE] = "OFFLINE",
> + [SNIC_TGT_STAT_DEL] = "DELETION IN PROGRESS",
> +};
> +
> +const char *
> +snic_tgt_state_to_str(int state)
> +{
> + return ((state >= SNIC_TGT_STAT_INIT && state <= SNIC_TGT_STAT_DEL) ?
> + snic_tgt_state_str[state] : "UNKNOWN");
> +}
> +
> +/*
> + * Initiate report_tgt req desc
> + */
> +static void
> +snic_report_tgt_init(struct snic_host_req *req, u32 hid, u8 *buf, u32 len,
> + u64 rsp_buf_pa, u64 ctx)
> +{
> + struct snic_sg_desc *sgd = NULL;
> +
> +
> + snic_io_hdr_enc(&req->hdr, SNIC_REQ_REPORT_TGTS, 0, SCSI_NO_TAG, hid,
> + 1, ctx);
> +
> + req->u.rpt_tgts.sg_cnt = 1;
> + sgd = req_to_sgl(req);
> + sgd[0].addr = cpu_to_le64(rsp_buf_pa);
> + sgd[0].len = cpu_to_le32(len);
> + sgd[0]._resvd = 0;
> + req->u.rpt_tgts.sg_addr = (u64) sgd;
> +}
> +
> +/*
> + * snic_queue_report_tgt_req: Queues report target request.
> + */
> +static int
> +snic_queue_report_tgt_req(struct snic *snic)
> +{
> + struct snic_req_info *rqi = NULL;
> + u32 ntgts, buf_len = 0;
> + u8 *buf = NULL;
> + dma_addr_t pa = 0;
> + int ret = 0;
> +
> + rqi = snic_req_init(snic, 1);
> + if (!rqi) {
> + ret = -ENOMEM;
> + goto error;
> + }
> +
> + if (snic->fwinfo.max_tgts)
> + ntgts = min_t(u32, snic->fwinfo.max_tgts, snic->shost->max_id);
> + else
> + ntgts = snic->shost->max_id;
> +
> + /* Allocate Response Buffer */
> + SNIC_BUG_ON(ntgts == 0);
> + buf_len = ntgts * sizeof(struct snic_tgt_id) + SNIC_SG_DESC_ALIGN;
> +
> + buf = kzalloc(buf_len, GFP_KERNEL|GFP_DMA);
> + if (!buf) {
> + snic_req_free(snic, rqi);
> + SNIC_HOST_ERR(snic->shost, "Resp Buf Alloc Failed.\n");
> +
> + ret = -ENOMEM;
> + goto error;
> + }
> +
> + SNIC_BUG_ON((((unsigned long)buf) % SNIC_SG_DESC_ALIGN) != 0);
> +
> + pa = pci_map_single(snic->pdev, buf, buf_len, PCI_DMA_FROMDEVICE);
> + if (pci_dma_mapping_error(snic->pdev, pa)) {
> + kfree(buf);
> + snic_req_free(snic, rqi);
> + SNIC_HOST_ERR(snic->shost,
> + "Rpt-tgt rspbuf %p: PCI DMA Mapping Failed\n",
> + buf);
> + ret = -EINVAL;
> +
> + goto error;
> + }
> +
> +
> + SNIC_BUG_ON(pa == 0);
> + rqi->sge_va = (u64) buf;
> +
> + snic_report_tgt_init(rqi->req,
> + snic->config.hid,
> + buf,
> + buf_len,
> + (u64)pa,
> + (u64)rqi);
> +
> + snic_handle_untagged_req(snic, rqi);
> +
> + ret = snic_queue_wq_desc(snic, rqi->req, rqi->req_len);
> + if (ret) {
> + pci_unmap_single(snic->pdev, pa, buf_len, PCI_DMA_FROMDEVICE);
> + kfree(buf);
> + rqi->sge_va = 0;
> + snic_release_untagged_req(snic, rqi);
> + SNIC_HOST_ERR(snic->shost, "Queuing Report Tgts Failed.\n");
> +
> + goto error;
> + }
> +
> + SNIC_DISC_DBG(snic->shost, "Report Targets Issued.\n");
> +
> + return ret;
> +
> +error:
> + SNIC_HOST_ERR(snic->shost,
> + "Queuing Report Targets Failed, err = %d\n",
> + ret);
> + return ret;
> +} /* end of snic_queue_report_tgt_req */
> +
> +/* call into SML */
> +static void
> +snic_scsi_scan_tgt(struct work_struct *work)
> +{
> + struct snic_tgt *tgt = container_of(work, struct snic_tgt, scan_work);
> + struct Scsi_Host *shost = dev_to_shost(&tgt->dev);
> + unsigned long flags;
> +
> + SNIC_HOST_INFO(shost, "Scanning Target id 0x%x\n", tgt->id);
> + scsi_scan_target(&tgt->dev,
> + tgt->channel,
> + tgt->scsi_tgt_id,
> + SCAN_WILD_CARD,
> + 1);
> +
> + spin_lock_irqsave(shost->host_lock, flags);
> + tgt->flags &= ~SNIC_TGT_SCAN_PENDING;
> + spin_unlock_irqrestore(shost->host_lock, flags);
> +} /* end of snic_scsi_scan_tgt */
> +
> +/*
> + * snic_tgt_lookup :
> + */
> +static struct snic_tgt *
> +snic_tgt_lookup(struct snic *snic, struct snic_tgt_id *tgtid)
> +{
> + struct list_head *cur, *nxt;
> + struct snic_tgt *tgt = NULL;
> +
> + list_for_each_safe(cur, nxt, &snic->disc.tgt_list) {
> + tgt = list_entry(cur, struct snic_tgt, list);
> + if (tgt->id == tgtid->tgt_id)
> + return tgt;
> + tgt = NULL;
> + }
> +
> + return tgt;
> +} /* end of snic_tgt_lookup */
> +
> +/*
> + * snic_tgt_dev_release : Called on dropping last ref for snic_tgt object
> + */
> +void
> +snic_tgt_dev_release(struct device *dev)
> +{
> + struct snic_tgt *tgt = dev_to_tgt(dev);
> +
> + SNIC_HOST_INFO(snic_tgt_to_shost(tgt),
> + "Target Device ID %d (%s) Permanently Deleted.\n",
> + tgt->id,
> + dev_name(dev));
> +
> + SNIC_BUG_ON(!list_empty(&tgt->list));
> + kfree(tgt);
> +}
> +
> +/*
> + * snic_tgt_del : work function to delete snic_tgt
> + */
> +static void
> +snic_tgt_del(struct work_struct *work)
> +{
> + struct snic_tgt *tgt = container_of(work, struct snic_tgt, del_work);
> + struct Scsi_Host *shost = snic_tgt_to_shost(tgt);
> +
> + if (tgt->flags & SNIC_TGT_SCAN_PENDING)
> + scsi_flush_work(shost);
> +
> + /* Block IOs on child devices, stops new IOs */
> + scsi_target_block(&tgt->dev);
> +
> + /* Cleanup IOs */
> + snic_tgt_scsi_abort_io(tgt);
> +
> + /* Unblock IOs now, to flush if there are any. */
> + scsi_target_unblock(&tgt->dev, SDEV_TRANSPORT_OFFLINE);
> +
> + /* Delete SCSI Target and sdevs */
> + scsi_remove_target(&tgt->dev); /* ?? */
> + device_del(&tgt->dev);
> + put_device(&tgt->dev);
> +} /* end of snic_tgt_del */
> +
> +/* snic_tgt_create: checks for existence of snic_tgt, if it doesn't
> + * it creates one.
> + */
> +static struct snic_tgt *
> +snic_tgt_create(struct snic *snic, struct snic_tgt_id *tgtid)
> +{
> + struct snic_tgt *tgt = NULL;
> + unsigned long flags;
> + int ret;
> +
> + tgt = snic_tgt_lookup(snic, tgtid);
> + if (tgt) {
> + /* update the information if required */
> + return tgt;
> + }
> +
> + tgt = kzalloc(sizeof(*tgt), GFP_KERNEL);
> + if (!tgt) {
> + SNIC_HOST_ERR(snic->shost, "Failure to allocate snic_tgt.\n");
> + ret = -ENOMEM;
> +
> + return tgt;
> + }
> +
> + INIT_LIST_HEAD(&tgt->list);
> + tgt->id = tgtid->tgt_id;
> + tgt->channel = 0;
> +
> + SNIC_BUG_ON(tgtid->tgt_type > SNIC_TGT_SAN);
> + tgt->tdata.typ = tgtid->tgt_type;
> +
> + /*
> + * Plugging into SML Device Tree
> + */
> + tgt->tdata.disc_id = 0;
> + tgt->state = SNIC_TGT_STAT_INIT;
> + device_initialize(&tgt->dev);
> + tgt->dev.parent = get_device(&snic->shost->shost_gendev);
> + tgt->dev.release = snic_tgt_dev_release;
Why do you use your own scsi target instantiation here?
If it's equivalent to the scsi target than you should rather use the
'scsi_target' structure here and attach driver-specific information
to either hostdata or starget_data.
Cheers,
Hannes
--
Dr. Hannes Reinecke zSeries & Storage
hare@suse.de +49 911 74053 688
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg
GF: J. Hawn, J. Guild, F. Imendörffer, HRB 16746 (AG Nürnberg)
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH v4 5/9] snic:add SCSI handling, AEN, and fwreset handling
2015-04-09 11:49 ` [PATCH v4 5/9] snic:add SCSI handling, AEN, and fwreset handling Narsimhulu Musini
@ 2015-04-09 13:16 ` Hannes Reinecke
2015-04-13 5:37 ` Narsimhulu Musini (nmusini)
0 siblings, 1 reply; 24+ messages in thread
From: Hannes Reinecke @ 2015-04-09 13:16 UTC (permalink / raw)
To: Narsimhulu Musini, JBottomley, linux-scsi, hch; +Cc: Sesidhar Baddela
Hi Narsimhulu,
please find some comments inline.
On 04/09/2015 01:49 PM, Narsimhulu Musini wrote:
> snic_scsi.c contains scsi handling, includes queuing io, abort, lun reset,
> and host reset. Also it handles asynchronous event notifications from FW.
>
> Signed-off-by: Narsimhulu Musini <nmusini@cisco.com>
> Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
> ---
> * v2
> - Changed queuecommand to lock-free version.
> - Converted custom error codes to standard error codes.
>
> drivers/scsi/snic/snic_scsi.c | 2638 +++++++++++++++++++++++++++++++++++++++++
> 1 file changed, 2638 insertions(+)
> create mode 100644 drivers/scsi/snic/snic_scsi.c
>
> diff --git a/drivers/scsi/snic/snic_scsi.c b/drivers/scsi/snic/snic_scsi.c
> new file mode 100644
> index 0000000..5d877f0
> --- /dev/null
> +++ b/drivers/scsi/snic/snic_scsi.c
> @@ -0,0 +1,2638 @@
> +/*
> + * Copyright 2014 Cisco Systems, Inc. All rights reserved.
> + *
> + * This program is free software; you may redistribute it and/or modify
> + * it under the terms of the GNU General Public License as published by
> + * the Free Software Foundation; version 2 of the License.
> + *
> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
> + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
> + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
> + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
> + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
> + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
> + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
> + * SOFTWARE.
> + */
> +
> +#include <linux/mempool.h>
> +#include <linux/errno.h>
> +#include <linux/init.h>
> +#include <linux/workqueue.h>
> +#include <linux/pci.h>
> +#include <linux/spinlock.h>
> +#include <linux/delay.h>
> +#include <linux/gfp.h>
> +#include <scsi/scsi.h>
> +#include <scsi/scsi_host.h>
> +#include <scsi/scsi_device.h>
> +#include <scsi/scsi_cmnd.h>
> +#include <scsi/scsi_tcq.h>
> +#include <scsi/scsi_dbg.h>
> +
> +#include "snic_io.h"
> +#include "snic.h"
> +
> +#define snic_cmd_tag(sc) (((struct scsi_cmnd *) sc)->request->tag)
> +
> +const char *snic_state_str[] = {
> + [SNIC_INIT] = "SNIC_INIT",
> + [SNIC_ERROR] = "SNIC_ERROR",
> + [SNIC_ONLINE] = "SNIC_ONLINE",
> + [SNIC_OFFLINE] = "SNIC_OFFLINE",
> + [SNIC_FWRESET] = "SNIC_FWRESET",
> +};
> +
> +static const char * const snic_req_state_str[] = {
> + [SNIC_IOREQ_NOT_INITED] = "SNIC_IOREQ_NOT_INITED",
> + [SNIC_IOREQ_PENDING] = "SNIC_IOREQ_PENDING",
> + [SNIC_IOREQ_ABTS_PENDING] = "SNIC_IOREQ_ABTS_PENDING",
> + [SNIC_IOREQ_ABTS_COMPLETE] = "SNIC_IOREQ_ABTS_COMPELTE",
> + [SNIC_IOREQ_LR_PENDING] = "SNIC_IOREQ_LR_PENDING",
> + [SNIC_IOREQ_LR_COMPLETE] = "SNIC_IOREQ_LR_COMPELTE",
> + [SNIC_IOREQ_COMPLETE] = "SNIC_IOREQ_CMD_COMPELTE",
> +};
> +
> +/* snic cmd status strings */
> +static const char * const snic_io_status_str[] = {
> + [SNIC_STAT_IO_SUCCESS] = "SNIC_STAT_IO_SUCCESS", /* 0x0 */
> + [SNIC_STAT_INVALID_HDR] = "SNIC_STAT_INVALID_HDR",
> + [SNIC_STAT_OUT_OF_RES] = "SNIC_STAT_OUT_OF_RES",
> + [SNIC_STAT_INVALID_PARM] = "SNIC_STAT_INVALID_PARM",
> + [SNIC_STAT_REQ_NOT_SUP] = "SNIC_STAT_REQ_NOT_SUP",
> + [SNIC_STAT_IO_NOT_FOUND] = "SNIC_STAT_IO_NOT_FOUND",
> + [SNIC_STAT_ABORTED] = "SNIC_STAT_ABORTED",
> + [SNIC_STAT_TIMEOUT] = "SNIC_STAT_TIMEOUT",
> + [SNIC_STAT_SGL_INVALID] = "SNIC_STAT_SGL_INVALID",
> + [SNIC_STAT_DATA_CNT_MISMATCH] = "SNIC_STAT_DATA_CNT_MISMATCH",
> + [SNIC_STAT_FW_ERR] = "SNIC_STAT_FW_ERR",
> + [SNIC_STAT_ITMF_REJECT] = "SNIC_STAT_ITMF_REJECT",
> + [SNIC_STAT_ITMF_FAIL] = "SNIC_STAT_ITMF_FAIL",
> + [SNIC_STAT_ITMF_INCORRECT_LUN] = "SNIC_STAT_ITMF_INCORRECT_LUN",
> + [SNIC_STAT_CMND_REJECT] = "SNIC_STAT_CMND_REJECT",
> + [SNIC_STAT_DEV_OFFLINE] = "SNIC_STAT_DEV_OFFLINE",
> + [SNIC_STAT_NO_BOOTLUN] = "SNIC_STAT_NO_BOOTLUN",
> + [SNIC_STAT_SCSI_ERR] = "SNIC_STAT_SCSI_ERR",
> + [SNIC_STAT_NOT_READY] = "SNIC_STAT_NOT_READY",
> + [SNIC_STAT_FATAL_ERROR] = "SNIC_STAT_FATAL_ERROR",
> +};
> +
> +static void snic_scsi_cleanup(struct snic *, int);
> +
> +const char *
> +snic_state_to_str(unsigned int state)
> +{
> + if (state >= ARRAY_SIZE(snic_state_str) || !snic_state_str[state])
> + return "Unknown";
> +
> + return snic_state_str[state];
> +}
> +
> +static const char *
> +snic_io_status_to_str(unsigned int state)
> +{
> + if ((state >= ARRAY_SIZE(snic_io_status_str)) ||
> + (!snic_io_status_str[state]))
> + return "Unknown";
> +
> + return snic_io_status_str[state];
> +}
> +
> +static const char *
> +snic_ioreq_state_to_str(unsigned int state)
> +{
> + if (state >= ARRAY_SIZE(snic_req_state_str) ||
> + !snic_req_state_str[state])
> + return "Unknown";
> +
> + return snic_req_state_str[state];
> +}
> +
> +static inline spinlock_t *
> +snic_io_lock_hash(struct snic *snic, struct scsi_cmnd *sc)
> +{
> + u32 hash = snic_cmd_tag(sc) & (SNIC_IO_LOCKS - 1);
> +
> + return &snic->io_req_lock[hash];
> +}
> +
> +static inline spinlock_t *
> +snic_io_lock_tag(struct snic *snic, int tag)
> +{
> + return &snic->io_req_lock[tag & (SNIC_IO_LOCKS - 1)];
> +}
> +
> +/* snic_release_req_buf : Releases snic_req_info */
> +static void
> +snic_release_req_buf(struct snic *snic,
> + struct snic_req_info *rqi,
> + struct scsi_cmnd *sc)
> +{
> + struct snic_host_req *req = rqi_to_req(rqi);
> +
> + /* Freeing cmd without marking completion, not okay */
> + SNIC_BUG_ON(!((CMD_STATE(sc) == SNIC_IOREQ_COMPLETE) ||
> + (CMD_STATE(sc) == SNIC_IOREQ_ABTS_COMPLETE) ||
> + (CMD_FLAGS(sc) & SNIC_DEV_RST_NOTSUP) ||
> + (CMD_FLAGS(sc) & SNIC_IO_INTERNAL_TERM_ISSUED) ||
> + (CMD_FLAGS(sc) & SNIC_DEV_RST_TERM_ISSUED) ||
> + (CMD_FLAGS(sc) & SNIC_SCSI_CLEANUP) ||
> + (CMD_STATE(sc) == SNIC_IOREQ_LR_COMPLETE)));
> +
> + SNIC_SCSI_DBG(snic->shost,
> + "Rel_req:sc %p:tag %x:rqi %p:ioreq %p:abt %p:dr %p: state %s:flags 0x%llx\n",
> + sc, snic_cmd_tag(sc), rqi, rqi->req, rqi->abort_req,
> + rqi->dr_req, snic_ioreq_state_to_str(CMD_STATE(sc)),
> + CMD_FLAGS(sc));
> +
> + if (req->u.icmnd.sense_addr)
> + pci_unmap_single(snic->pdev,
> + req->u.icmnd.sense_addr,
> + SCSI_SENSE_BUFFERSIZE,
> + PCI_DMA_FROMDEVICE);
> +
> + scsi_dma_unmap(sc);
> +
> + snic_req_free(snic, rqi);
> +} /* end of snic_release_req_buf */
> +
> +/*
> + * snic_queue_icmnd_req : Queues snic_icmnd request
> + */
> +static int
> +snic_queue_icmnd_req(struct snic *snic,
> + struct snic_req_info *rqi,
> + struct scsi_cmnd *sc,
> + int sg_cnt)
> +{
> + struct scatterlist *sg;
> + struct snic_sg_desc *sgd;
> + dma_addr_t pa = 0;
> + struct scsi_lun lun;
> + int flags = 0;
> + int ret = 0;
> + unsigned int i;
> +
> + if (sg_cnt) {
> + flags = SNIC_ICMND_ESGL;
> + sgd = (struct snic_sg_desc *) req_to_sgl(rqi->req);
> +
> + for_each_sg(scsi_sglist(sc), sg, sg_cnt, i) {
> + sgd->addr = cpu_to_le64(sg_dma_address(sg));
> + sgd->len = cpu_to_le32(sg_dma_len(sg));
> + sgd->_resvd = 0;
> + sgd++;
> + }
> + }
> +
> + pa = pci_map_single(snic->pdev,
> + sc->sense_buffer,
> + SCSI_SENSE_BUFFERSIZE,
> + PCI_DMA_FROMDEVICE);
> +
> + if (pci_dma_mapping_error(snic->pdev, pa)) {
> + SNIC_HOST_ERR(snic->shost,
> + "QIcmnd:PCI Map Failed for sns buf %p tag %x\n",
> + sc->sense_buffer, snic_cmd_tag(sc));
> + ret = -ENOMEM;
> +
> + return ret;
> + }
> +
> + int_to_scsilun(sc->device->lun, &lun);
> + if (sc->sc_data_direction == DMA_FROM_DEVICE)
> + flags |= SNIC_ICMND_RD;
> + if (sc->sc_data_direction == DMA_TO_DEVICE)
> + flags |= SNIC_ICMND_WR;
> +
> + /* Initialize icmnd */
> + snic_icmnd_init(rqi->req,
> + snic_cmd_tag(sc),
> + snic->config.hid, /* hid */
> + (u64)rqi,
> + flags, /* command flags */
> + rqi->tgt_id,
> + lun.scsi_lun,
> + sc->cmnd,
> + sc->cmd_len,
> + scsi_bufflen(sc),
> + sg_cnt,
> + (u64)req_to_sgl(rqi->req),
> + pa, /* sense buffer pa */
> + SCSI_SENSE_BUFFERSIZE);
> +
> + ret = snic_queue_wq_desc(snic, rqi->req, rqi->req_len);
> + if (ret)
> + SNIC_HOST_ERR(snic->shost,
> + "QIcmnd: Queuing Icmnd Failed. ret = %d\n",
> + ret);
> +
> + return ret;
> +} /* end of snic_queue_icmnd_req */
> +
> +/*
> + * snic_issue_scsi_req : Prepares IO request and Issues to FW.
> + */
> +static int
> +snic_issue_scsi_req(struct snic *snic,
> + struct snic_tgt *tgt,
> + struct scsi_cmnd *sc)
> +{
> + struct snic_req_info *rqi = NULL;
> + int sg_cnt = 0;
> + int ret = 0;
> + u32 tag = snic_cmd_tag(sc);
> + u64 cmd_trc = 0, cmd_st_flags = 0;
> + spinlock_t *io_lock = NULL;
> + unsigned long flags;
> +
> + CMD_STATE(sc) = SNIC_IOREQ_NOT_INITED;
> + CMD_FLAGS(sc) = SNIC_NO_FLAGS;
> + sg_cnt = scsi_dma_map(sc);
> + if (sg_cnt < 0) {
> + SNIC_TRC((u16)snic->shost->host_no, tag, (u64)sc, 0,
> + sc->cmnd[0], sg_cnt, CMD_STATE(sc));
> +
> + SNIC_HOST_ERR(snic->shost, "issue_sc:Failed to map SG List.\n");
> + ret = -ENOMEM;
> +
> + goto issue_sc_end;
> + }
> +
> + rqi = snic_req_init(snic, sg_cnt);
> + if (!rqi) {
> + scsi_dma_unmap(sc);
> + ret = -ENOMEM;
> +
> + goto issue_sc_end;
> + }
> +
> + rqi->tgt_id = tgt->id;
> + rqi->sc = sc;
> +
> + CMD_STATE(sc) = SNIC_IOREQ_PENDING;
> + CMD_SP(sc) = (char *) rqi;
> + cmd_trc = SNIC_TRC_CMD(sc);
> + CMD_FLAGS(sc) |= (SNIC_IO_INITIALIZED | SNIC_IO_ISSUED);
> + cmd_st_flags = SNIC_TRC_CMD_STATE_FLAGS(sc);
> + io_lock = snic_io_lock_hash(snic, sc);
> +
> + /* create wq desc and enqueue it */
> + ret = snic_queue_icmnd_req(snic, rqi, sc, sg_cnt);
> + if (ret) {
> + SNIC_HOST_ERR(snic->shost,
> + "issue_sc: icmnd qing Failed for sc %p, err %d\n",
> + sc, ret);
> +
> + spin_lock_irqsave(io_lock, flags);
> + rqi = (struct snic_req_info *) CMD_SP(sc);
> + CMD_SP(sc) = NULL;
> + CMD_STATE(sc) = SNIC_IOREQ_COMPLETE;
> + CMD_FLAGS(sc) &= ~SNIC_IO_ISSUED; /* turn off the flag */
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + if (rqi)
> + snic_release_req_buf(snic, rqi, sc);
> +
> + SNIC_TRC(snic->shost->host_no, tag, (u64)sc, 0, 0, 0,
> + SNIC_TRC_CMD_STATE_FLAGS(sc));
> + } else {
> + u32 io_sz = scsi_bufflen(sc) >> 9;
> + u32 qtime = jiffies - rqi->start_time;
> + struct snic_io_stats *iostats = &snic->s_stats.io;
> +
> + if (io_sz > atomic64_read(&iostats->max_io_sz))
> + atomic64_set(&iostats->max_io_sz, io_sz);
> +
> + if (qtime > atomic64_read(&iostats->max_qtime))
> + atomic64_set(&iostats->max_qtime, qtime);
> +
> + SNIC_SCSI_DBG(snic->shost,
> + "issue_sc:sc %p, tag %d queued to WQ.\n",
> + sc, tag);
> +
> + SNIC_TRC(snic->shost->host_no, tag, (u64)sc, (u64)rqi, sg_cnt,
> + cmd_trc, cmd_st_flags);
> + }
> +
> +issue_sc_end:
> +
> + return ret;
> +} /* end of snic_issue_scsi_req */
> +
> +
> +/*
> + * snic_queuecommand
> + * Routine to send a scsi cdb to LLD
> + * Called with host_lock held and interrupts disabled
> + */
> +int
> +snic_queuecommand(struct Scsi_Host *shost, struct scsi_cmnd *sc)
> +{
> + struct snic_tgt *tgt = NULL;
> + struct snic *snic = shost_priv(shost);
> + int ret;
> +
> + tgt = starget_to_tgt(scsi_target(sc->device));
> + ret = snic_tgt_chkready(tgt);
> + if (ret) {
> + SNIC_HOST_ERR(shost, "Tgt %p id %d Not Ready.\n", tgt, tgt->id);
> + atomic64_inc(&snic->s_stats.misc.tgt_not_rdy);
> + sc->result = ret;
> + sc->scsi_done(sc);
> +
> + return 0;
> + }
> +
> + if (snic_get_state(snic) != SNIC_ONLINE) {
> + SNIC_HOST_ERR(shost, "snic state is %s\n",
> + snic_state_str[snic_get_state(snic)]);
> +
> + return SCSI_MLQUEUE_HOST_BUSY;
> + }
> + atomic_inc(&snic->ios_inflight);
> +
> + SNIC_SCSI_DBG(shost, "sc %p Tag %d (sc %0x) lun %lld in snic_qcmd\n",
> + sc, snic_cmd_tag(sc), sc->cmnd[0], sc->device->lun);
> +
> + memset(scsi_cmd_priv(sc), 0, sizeof(struct snic_internal_io_state));
> +
> +#ifdef SNIC_DEBUG
> + scsi_print_command(sc);
> +#endif
> +
> + ret = snic_issue_scsi_req(snic, tgt, sc);
> + if (ret) {
> + SNIC_HOST_ERR(shost, "Failed to Q, Scsi Req w/ err %d.\n", ret);
> + ret = SCSI_MLQUEUE_HOST_BUSY;
> + } else
> + snic_stats_update_active_ios(&snic->s_stats);
> +
> + atomic_dec(&snic->ios_inflight);
> +
> + return ret;
> +} /* end of snic_queuecommand */
> +
> +/*
> + * snic_process_abts_pending_state:
> + * caller should hold IO lock
> + */
> +static void
> +snic_proc_tmreq_pending_state(struct snic *snic,
> + struct scsi_cmnd *sc,
> + u8 cmpl_status)
> +{
> + int state = CMD_STATE(sc);
> +
> + if (state == SNIC_IOREQ_ABTS_PENDING)
> + CMD_FLAGS(sc) |= SNIC_IO_ABTS_PENDING;
> + else if (state == SNIC_IOREQ_LR_PENDING)
> + CMD_FLAGS(sc) |= SNIC_DEV_RST_PENDING;
> + else
> + SNIC_BUG_ON(1);
> +
> + switch (cmpl_status) {
> + case SNIC_STAT_IO_SUCCESS:
> + CMD_FLAGS(sc) |= SNIC_IO_DONE;
> + break;
> +
> + case SNIC_STAT_ABORTED:
> + CMD_FLAGS(sc) |= SNIC_IO_ABORTED;
> + break;
> +
> + default:
> + SNIC_BUG_ON(1);
> + }
> +}
> +
> +/*
> + * snic_process_io_failed_state:
> + * Processes IO's error states
> + */
> +static void
> +snic_process_io_failed_state(struct snic *snic,
> + struct snic_icmnd_cmpl *icmnd_cmpl,
> + struct scsi_cmnd *sc,
> + u8 cmpl_stat)
> +{
> + int res = 0;
> +
> + switch (cmpl_stat) {
> + case SNIC_STAT_TIMEOUT: /* Req was timedout */
> + atomic64_inc(&snic->s_stats.misc.io_tmo);
> + res = DID_TIME_OUT;
> + break;
> +
> + case SNIC_STAT_ABORTED: /* Req was aborted */
> + atomic64_inc(&snic->s_stats.misc.io_aborted);
> + res = DID_ERROR;
> + break;
> +
DID_ABORTED?
> + case SNIC_STAT_DATA_CNT_MISMATCH:/* Recv/Sent more/less data than exp */
> + atomic64_inc(&snic->s_stats.misc.data_cnt_mismat);
> + scsi_set_resid(sc, icmnd_cmpl->resid);
> + res = DID_ERROR;
> + break;
> +
> + case SNIC_STAT_OUT_OF_RES: /* Out of resources to complete request */
> + atomic64_inc(&snic->s_stats.fw.out_of_res);
> + res = DID_REQUEUE;
> + break;
> +
> + case SNIC_STAT_IO_NOT_FOUND: /* Requested I/O was not found */
> + atomic64_inc(&snic->s_stats.io.io_not_found);
> + res = DID_ERROR;
> + break;
> +
> + case SNIC_STAT_SGL_INVALID: /* Req was aborted to due to sgl error*/
> + atomic64_inc(&snic->s_stats.misc.sgl_inval);
> + res = DID_ERROR;
> + break;
> +
> + case SNIC_STAT_FW_ERR: /* Req terminated due to FW Error */
> + atomic64_inc(&snic->s_stats.fw.io_errs);
> + res = DID_ERROR;
> + break;
> +
> + case SNIC_STAT_SCSI_ERR: /* FW hits SCSI Error */
> + atomic64_inc(&snic->s_stats.fw.scsi_errs);
> + res = DID_ERROR;
> + break;
> +
SCSI error? What should stat be? If it's the normal SAM_STAT_XXX values
it should be decoded correctly.
> + case SNIC_STAT_INVALID_HDR: /* Hdr contains invalid data */
> + case SNIC_STAT_INVALID_PARM: /* Some param in req is invalid */
> + case SNIC_STAT_REQ_NOT_SUP: /* Req type is not supported */
> + case SNIC_STAT_CMND_REJECT: /* Req rejected */
> + case SNIC_STAT_DEV_OFFLINE: /* Device offline */
> + case SNIC_STAT_NOT_READY: /* XPT yet to initialize */
> + case SNIC_STAT_FATAL_ERROR: /* XPT Error */
> + default:
> + SNIC_SCSI_DBG(snic->shost,
> + "Invalid Hdr/Param or Req Not Supported or Cmnd Rejected or Device Offline. or Unknown\n");
> + res = DID_ERROR;
> + break;
> + }
> +
You should be returning 'DID_NO_CONNECT' for DEV_OFFLINE. If NOT_READY
is temporary you should be returning DID_RETRY or similar.
Otherwise DID_NO_CONNECT is preferred, too.
> + SNIC_HOST_ERR(snic->shost, "fw returns failed status %s flags 0x%llx\n",
> + snic_io_status_to_str(cmpl_stat), CMD_FLAGS(sc));
> +
> + /* Set sc->result */
> + sc->result = (res << 16) | icmnd_cmpl->scsi_status;
> +} /* end of snic_process_io_failed_state */
> +
> +/*
> + * snic_tmreq_pending : is task management in progress.
> + */
> +static int
> +snic_tmreq_pending(struct scsi_cmnd *sc)
> +{
> + int state = CMD_STATE(sc);
> +
> + return ((state == SNIC_IOREQ_ABTS_PENDING) ||
> + (state == SNIC_IOREQ_LR_PENDING));
> +}
> +
> +/*
> + * snic_process_icmnd_cmpl_status:
> + * Caller should hold io_lock
> + */
> +static int
> +snic_process_icmnd_cmpl_status(struct snic *snic,
> + struct snic_icmnd_cmpl *icmnd_cmpl,
> + u8 cmpl_stat,
> + struct scsi_cmnd *sc)
> +{
> + u8 scsi_stat = icmnd_cmpl->scsi_status;
> + u64 xfer_len = 0;
> + int ret = 0;
> +
> + /* Mark the IO as complete */
> + CMD_STATE(sc) = SNIC_IOREQ_COMPLETE;
> +
> + if (likely(cmpl_stat == SNIC_STAT_IO_SUCCESS)) {
> + sc->result = (DID_OK << 16) | scsi_stat;
> +
> + xfer_len = scsi_bufflen(sc);
> +
> + /* Update SCSI Cmd with resid value */
> + scsi_set_resid(sc, icmnd_cmpl->resid);
> +
> + if (icmnd_cmpl->flags & SNIC_ICMND_CMPL_UNDR_RUN) {
> + xfer_len -= icmnd_cmpl->resid;
> + atomic64_inc(&snic->s_stats.misc.io_under_run);
> + }
> +
> + if (icmnd_cmpl->scsi_status == SAM_STAT_TASK_SET_FULL)
> + atomic64_inc(&snic->s_stats.misc.qfull);
> +
> + ret = 0;
> + } else {
> + snic_process_io_failed_state(snic, icmnd_cmpl, sc, cmpl_stat);
> + atomic64_inc(&snic->s_stats.io.fail);
> + SNIC_HOST_ERR(snic->shost,
> + "icmnd_cmpl: IO Failed : Hdr Status %s flags 0x%llx\n",
> + snic_io_status_to_str(cmpl_stat), CMD_FLAGS(sc));
> + ret = 1;
> + }
> +
> + return ret;
> +} /* end of snic_process_icmnd_cmpl_status */
> +
> +
> +/*
> + * snic_icmnd_cmpl_handler
> + * Routine to handle icmnd completions
> + */
> +static void
> +snic_icmnd_cmpl_handler(struct snic *snic, struct snic_fw_req *fwreq)
> +{
> + u8 typ, hdr_stat;
> + u32 cmnd_id, hid;
> + u64 ctx;
> + struct scsi_cmnd *sc = NULL;
> + struct snic_icmnd_cmpl *icmnd_cmpl = NULL;
> + struct snic_host_req *req = NULL;
> + struct snic_req_info *rqi = NULL;
> + unsigned long flags, start_time;
> + spinlock_t *io_lock;
> + u8 sc_stat = 0;
> +
> + snic_io_hdr_dec(&fwreq->hdr, &typ, &hdr_stat, &cmnd_id, &hid, &ctx);
> + icmnd_cmpl = &fwreq->u.icmnd_cmpl;
> + sc_stat = icmnd_cmpl->scsi_status;
> +
> + SNIC_SCSI_DBG(snic->shost,
> + "Icmnd_cmpl: type = %x, hdr_stat = %x, cmnd_id = %x, hid = %x,i ctx = %llx\n",
> + typ, hdr_stat, cmnd_id, hid, ctx);
> +
> + if (cmnd_id >= snic->max_tag_id) {
> + SNIC_HOST_ERR(snic->shost,
> + "Icmnd_cmpl:Tag Error:Out of Range Tag %d, hdr status = %s\n",
> + cmnd_id, snic_io_status_to_str(hdr_stat));
> + return;
> + }
> +
> + sc = scsi_host_find_tag(snic->shost, cmnd_id);
> + WARN_ON_ONCE(!sc);
> +
> + if (!sc) {
> + atomic64_inc(&snic->s_stats.io.sc_null);
> + SNIC_HOST_ERR(snic->shost,
> + "Icmnd_cmpl: Scsi Cmnd Not found, sc = NULL Hdr Status = %s tag = 0x%x fwreq = 0x%p\n",
> + snic_io_status_to_str(hdr_stat),
> + cmnd_id,
> + fwreq);
> +
> + SNIC_TRC(snic->shost->host_no, cmnd_id, 0,
> + ((u64)hdr_stat << 16 |
> + (u64)sc_stat << 8 | (u64)icmnd_cmpl->flags),
> + fwreq, icmnd_cmpl->resid, ctx);
> +
> + return;
> + }
> +
> +
> +#ifdef SNIC_DEBUG
> + scsi_print_result(sc);
> +#endif
> +
> + io_lock = snic_io_lock_hash(snic, sc);
> +
> + spin_lock_irqsave(io_lock, flags);
> + rqi = (struct snic_req_info *) CMD_SP(sc);
> + SNIC_SCSI_DBG(snic->shost,
> + "Icmnd_cmpl:lun %lld sc %p cmd %xtag %d flags 0x%llx rqi %p\n",
> + sc->device->lun, sc, sc->cmnd[0], snic_cmd_tag(sc),
> + CMD_FLAGS(sc), rqi);
> +
> + SNIC_BUG_ON(rqi != (struct snic_req_info *)ctx);
> + WARN_ON_ONCE(req);
> + if (!rqi) {
> + atomic64_inc(&snic->s_stats.io.req_null);
> + CMD_FLAGS(sc) |= SNIC_IO_REQ_NULL;
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + SNIC_HOST_ERR(snic->shost,
> + "Icmnd_cmpl:Host Req Not Found(null), Hdr Status %s, Tag 0x%x, sc 0x%p flags 0x%llx\n",
> + snic_io_status_to_str(hdr_stat),
> + cmnd_id, sc, CMD_FLAGS(sc));
> + return;
> + }
> +
> + rqi = (struct snic_req_info *) ctx;
> + start_time = rqi->start_time;
> +
> + /* firmware completed the io */
> + rqi->io_cmpl = 1;
> +
> + /*
> + * if SCSI-ML has already issued abort on this command,
> + * ignore completion of the IO. The abts path will clean it up
> + */
> + if (unlikely(snic_tmreq_pending(sc))) {
> + snic_proc_tmreq_pending_state(snic, sc, hdr_stat);
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + snic_stats_update_io_cmpl(&snic->s_stats);
> +
> + /* Expected value is SNIC_STAT_ABORTED */
> + if (likely(hdr_stat == SNIC_STAT_ABORTED))
> + return;
> +
> + SNIC_SCSI_DBG(snic->shost,
> + "icmnd_cmpl:TM Req Pending(%s), Hdr Status %s sc 0x%p scsi status %x resid %d flags 0x%llx\n",
> + snic_ioreq_state_to_str(CMD_STATE(sc)),
> + snic_io_status_to_str(hdr_stat),
> + sc, sc_stat, icmnd_cmpl->resid, CMD_FLAGS(sc));
> +
> + SNIC_TRC(snic->shost->host_no, cmnd_id, sc,
> + jiffies_to_msecs(jiffies - start_time), fwreq,
> + SNIC_TRC_CMD(sc), SNIC_TRC_CMD_STATE_FLAGS(sc));
> +
> + return;
> + }
> +
> + if (snic_process_icmnd_cmpl_status(snic, icmnd_cmpl, hdr_stat, sc)) {
> +#ifdef SNIC_DEBUG
> + scsi_print_command(sc);
> +#endif
> + SNIC_HOST_ERR(snic->shost,
> + "icmnd_cmpl:IO Failed, sc 0x%p Tag %d Cmd %x Hdr Status %s flags 0x%llx\n",
> + sc, sc->cmnd[0], cmnd_id,
> + snic_io_status_to_str(hdr_stat), CMD_FLAGS(sc));
> + }
> +
> + /* Break link with the SCSI Command */
> + CMD_SP(sc) = NULL;
> + CMD_FLAGS(sc) |= SNIC_IO_DONE;
> +
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + /* For now, consider only successful IO. */
> + snic_calc_io_process_time(snic, rqi);
> +
> + snic_release_req_buf(snic, rqi, sc);
> +
> + SNIC_TRC(snic->shost->host_no, cmnd_id, (u64)sc,
> + jiffies_to_msecs(jiffies - start_time), (u64)fwreq,
> + SNIC_TRC_CMD(sc), SNIC_TRC_CMD_STATE_FLAGS(sc));
> +
> +
> + if (sc->scsi_done)
> + sc->scsi_done(sc);
> +
> + snic_stats_update_io_cmpl(&snic->s_stats);
> +} /* end of snic_icmnd_cmpl_handler */
> +
> +static void
> +snic_proc_dr_cmpl_locked(struct snic *snic,
> + struct snic_fw_req *fwreq,
> + u8 cmpl_stat,
> + u32 cmnd_id,
> + struct scsi_cmnd *sc)
> +{
> + struct snic_req_info *rqi = (struct snic_req_info *) CMD_SP(sc);
> + u32 start_time = rqi->start_time;
> +
> + CMD_LR_STATUS(sc) = cmpl_stat;
> +
> + SNIC_SCSI_DBG(snic->shost, "itmf_cmpl: Cmd State = %s\n",
> + snic_ioreq_state_to_str(CMD_STATE(sc)));
> +
> + if (CMD_STATE(sc) == SNIC_IOREQ_ABTS_PENDING) {
> + CMD_FLAGS(sc) |= SNIC_DEV_RST_ABTS_PENDING;
> +
> + SNIC_TRC(snic->shost->host_no, cmnd_id, sc,
> + jiffies_to_msecs(jiffies - start_time),
> + fwreq, 0, SNIC_TRC_CMD_STATE_FLAGS(sc));
> +
> + SNIC_SCSI_DBG(snic->shost,
> + "itmf_cmpl: Terminate Pending Dev Reset Cmpl Recvd.id %x, status %s flags 0x%llx\n",
> + (int)(cmnd_id & SNIC_TAG_MASK),
> + snic_io_status_to_str(cmpl_stat),
> + CMD_FLAGS(sc));
> +
> + return;
> + }
> +
> +
> + if (CMD_FLAGS(sc) & SNIC_DEV_RST_TIMEDOUT) {
> + SNIC_TRC(snic->shost->host_no, cmnd_id, sc,
> + jiffies_to_msecs(jiffies - start_time),
> + fwreq, 0, SNIC_TRC_CMD_STATE_FLAGS(sc));
> +
> + SNIC_SCSI_DBG(snic->shost,
> + "itmf_cmpl:Dev Reset Completion Received after timeout. id %d cmpl status %s flags 0x%llx\n",
> + (int)(cmnd_id & SNIC_TAG_MASK),
> + snic_io_status_to_str(cmpl_stat),
> + CMD_FLAGS(sc));
> +
> + return;
> + }
> +
> + CMD_STATE(sc) = SNIC_IOREQ_LR_COMPLETE;
> + CMD_FLAGS(sc) |= SNIC_DEV_RST_DONE;
> +
> + SNIC_SCSI_DBG(snic->shost,
> + "itmf_cmpl:Dev Reset Cmpl Recvd id %d cmpl status %s flags 0x%llx\n",
> + (int)(cmnd_id & SNIC_TAG_MASK),
> + snic_io_status_to_str(cmpl_stat),
> + CMD_FLAGS(sc));
> +
> + if (rqi->dr_done)
> + complete(rqi->dr_done);
> +} /* end of snic_proc_dr_cmpl_locked */
> +
> +/*
> + * snic_update_abort_stats : Updates abort stats based on completion status.
> + */
> +static void
> +snic_update_abort_stats(struct snic *snic, u8 cmpl_stat)
> +{
> + struct snic_abort_stats *abt_stats = &snic->s_stats.abts;
> +
> + SNIC_SCSI_DBG(snic->shost, "Updating Abort stats.\n");
> +
> + switch (cmpl_stat) {
> + case SNIC_STAT_IO_SUCCESS:
> + break;
> +
> + case SNIC_STAT_TIMEOUT:
> + atomic64_inc(&abt_stats->fw_tmo);
> + break;
> +
> + case SNIC_STAT_IO_NOT_FOUND:
> + atomic64_inc(&abt_stats->io_not_found);
> + break;
> +
> + default:
> + atomic64_inc(&abt_stats->fail);
> + break;
> + }
> +}
> +
> +static int
> +snic_process_itmf_cmpl(struct snic *snic,
> + struct snic_fw_req *fwreq,
> + u32 cmnd_id,
> + u8 cmpl_stat,
> + struct scsi_cmnd *sc)
> +{
> + struct snic_req_info *rqi = NULL;
> + u32 tm_tags = 0;
> + spinlock_t *io_lock = NULL;
> + unsigned long flags;
> + u32 start_time = 0;
> + int ret = 0;
> +
> + io_lock = snic_io_lock_hash(snic, sc);
> + spin_lock_irqsave(io_lock, flags);
> + rqi = (struct snic_req_info *) CMD_SP(sc);
> + WARN_ON_ONCE(!rqi);
> +
> + if (!rqi) {
> + atomic64_inc(&snic->s_stats.io.req_null);
> + spin_unlock_irqrestore(io_lock, flags);
> + CMD_FLAGS(sc) |= SNIC_IO_ABTS_TERM_REQ_NULL;
> + SNIC_HOST_ERR(snic->shost,
> + "itmf_cmpl: rqi is null,Hdr stat = %s Tag = 0x%x sc = 0x%p flags 0x%llx\n",
> + snic_io_status_to_str(cmpl_stat), cmnd_id, sc,
> + CMD_FLAGS(sc));
> +
> + return ret;
> + }
> +
> + /* Extract task management flags */
> + tm_tags = cmnd_id & ~(SNIC_TAG_MASK);
> +
> + start_time = rqi->start_time;
> + cmnd_id &= (SNIC_TAG_MASK);
> +
> + switch (tm_tags) {
> + case SNIC_TAG_ABORT:
> + /* Abort only issued on cmd */
> + snic_update_abort_stats(snic, cmpl_stat);
> +
> + if (CMD_STATE(sc) != SNIC_IOREQ_ABTS_PENDING) {
> + /* This is a late completion. Ignore it. */
> + ret = -1;
> + spin_unlock_irqrestore(io_lock, flags);
> + break;
> + }
> +
> + CMD_STATE(sc) = SNIC_IOREQ_ABTS_COMPLETE;
> + CMD_ABTS_STATUS(sc) = cmpl_stat;
> + CMD_FLAGS(sc) |= SNIC_IO_ABTS_TERM_DONE;
> +
> + SNIC_SCSI_DBG(snic->shost,
> + "itmf_cmpl:Abort Cmpl Recvd.Tag 0x%x Status %s flags 0x%llx\n",
> + cmnd_id,
> + snic_io_status_to_str(cmpl_stat),
> + CMD_FLAGS(sc));
> +
> + /*
> + * If scsi_eh thread is blocked waiting for abts complete,
> + * signal completion to it. IO will be cleaned in the thread,
> + * else clean it in this context.
> + */
> + if (rqi->abts_done) {
> + complete(rqi->abts_done);
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + break; /* jump out */
> + }
> +
> + CMD_SP(sc) = NULL;
> + sc->result = (DID_ERROR << 16);
> + SNIC_SCSI_DBG(snic->shost,
> + "itmf_cmpl: Completing IO. sc %p flags 0x%llx\n",
> + sc, CMD_FLAGS(sc));
> +
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + snic_release_req_buf(snic, rqi, sc);
> +
> + if (sc->scsi_done) {
> + SNIC_TRC(snic->shost->host_no, cmnd_id, sc,
> + jiffies_to_msecs(jiffies - start_time),
> + fwreq,
> + SNIC_TRC_CMD(sc),
> + SNIC_TRC_CMD_STATE_FLAGS(sc));
> +
> + sc->scsi_done(sc);
> + }
> +
> + break;
> +
> + case SNIC_TAG_DEV_RST:
> + case SNIC_TAG_DEV_RST | SNIC_TAG_IOCTL_DEV_RST:
> + snic_proc_dr_cmpl_locked(snic, fwreq, cmpl_stat, cmnd_id, sc);
> + spin_unlock_irqrestore(io_lock, flags);
> + ret = 0;
> +
> + break;
> +
> + case SNIC_TAG_ABORT | SNIC_TAG_DEV_RST:
> + /* Abort and terminate completion of device reset req */
> +
> + CMD_STATE(sc) = SNIC_IOREQ_ABTS_COMPLETE;
> + CMD_ABTS_STATUS(sc) = cmpl_stat;
> + CMD_FLAGS(sc) |= SNIC_DEV_RST_DONE;
> +
> + SNIC_SCSI_DBG(snic->shost,
> + "itmf_cmpl:dev reset abts cmpl recvd. id %d status %s flags 0x%llx\n",
> + cmnd_id, snic_io_status_to_str(cmpl_stat),
> + CMD_FLAGS(sc));
> +
> + if (rqi->abts_done)
> + complete(rqi->abts_done);
> +
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + break;
> +
> + default:
> + spin_unlock_irqrestore(io_lock, flags);
> + SNIC_HOST_ERR(snic->shost,
> + "itmf_cmpl: Unknown TM tag bit 0x%x\n", tm_tags);
> +
> + SNIC_HOST_ERR(snic->shost,
> + "itmf_cmpl:Unexpected itmf io stat %s Tag = 0x%x flags 0x%llx\n",
> + snic_ioreq_state_to_str(CMD_STATE(sc)),
> + cmnd_id,
> + CMD_FLAGS(sc));
> + ret = -1;
> + SNIC_BUG_ON(1);
> +
> + break;
> + }
> +
> + return ret;
> +} /* end of snic_process_itmf_cmpl_status */
> +
> +/*
> + * snic_itmf_cmpl_handler.
> + * Routine to handle itmf completions.
> + */
> +static void
> +snic_itmf_cmpl_handler(struct snic *snic, struct snic_fw_req *fwreq)
> +{
> + struct scsi_cmnd *sc = NULL;
> + struct snic_req_info *rqi = NULL;
> + struct snic_itmf_cmpl *itmf_cmpl = NULL;
> + u64 ctx;
> + u32 cmnd_id;
> + u32 hid;
> + u8 typ;
> + u8 hdr_stat;
> +
> + snic_io_hdr_dec(&fwreq->hdr, &typ, &hdr_stat, &cmnd_id, &hid, &ctx);
> + SNIC_SCSI_DBG(snic->shost,
> + "Itmf_cmpl: %s: type = %x, hdr_stat = %x, cmnd_id = %x, hid = %x,ctx = %llx\n",
> + __func__, typ, hdr_stat, cmnd_id, hid, ctx);
> +
> + itmf_cmpl = &fwreq->u.itmf_cmpl;
> + SNIC_SCSI_DBG(snic->shost,
> + "Itmf_cmpl: nterm %u , flags 0x%x\n",
> + itmf_cmpl->nterminated, itmf_cmpl->flags);
> +
> + /* spl case, dev reset issued through ioctl */
> + if (cmnd_id & SNIC_TAG_IOCTL_DEV_RST) {
> + rqi = (struct snic_req_info *) ctx;
> + sc = rqi->sc;
> +
> + goto ioctl_dev_rst;
> + }
> +
> + if ((cmnd_id & SNIC_TAG_MASK) >= snic->max_tag_id) {
> + SNIC_HOST_ERR(snic->shost,
> + "Itmf_cmpl: Tag 0x%x out of Range,HdrStat %s\n",
> + cmnd_id, snic_io_status_to_str(hdr_stat));
> + SNIC_BUG_ON(1);
> +
> + return;
> + }
> +
> + sc = scsi_host_find_tag(snic->shost, cmnd_id & SNIC_TAG_MASK);
> + WARN_ON_ONCE(!sc);
> +
> +ioctl_dev_rst:
> + if (!sc) {
> + atomic64_inc(&snic->s_stats.io.sc_null);
> + SNIC_HOST_ERR(snic->shost,
> + "Itmf_cmpl: sc is NULL - Hdr Stat %s Tag 0x%x\n",
> + snic_io_status_to_str(hdr_stat), cmnd_id);
> +
> + return;
> + }
> +
> + snic_process_itmf_cmpl(snic, fwreq, cmnd_id, hdr_stat, sc);
> +} /* end of snic_itmf_cmpl_handler */
> +
> +
> +
> +static void
> +snic_hba_reset_scsi_cleanup(struct snic *snic, struct scsi_cmnd *sc)
> +{
> + struct snic_stats *st = &snic->s_stats;
> + long act_ios = 0, act_fwreqs = 0;
> +
> + SNIC_SCSI_DBG(snic->shost, "HBA Reset scsi cleanup.\n");
> + snic_scsi_cleanup(snic, snic_cmd_tag(sc));
> +
> + /* Update stats on pending IOs */
> + act_ios = atomic64_read(&st->io.active);
> + atomic64_add(act_ios, &st->io.compl);
> + atomic64_sub(act_ios, &st->io.active);
> +
> + act_fwreqs = atomic64_read(&st->fw.actv_reqs);
> + atomic64_sub(act_fwreqs, &st->fw.actv_reqs);
> +}
> +
> +/*
> + * snic_hba_reset_cmpl_handler :
> + *
> + * Notes :
> + * 1. Cleanup all the scsi cmds, release all snic specific cmds
> + * 2. Issue Report Targets in case of SAN targets
> + */
> +static int
> +snic_hba_reset_cmpl_handler(struct snic *snic, struct snic_fw_req *fwreq)
> +{
> + u64 ctx;
> + u32 cmnd_id;
> + u32 hid;
> + u8 typ;
> + u8 hdr_stat;
> + struct scsi_cmnd *sc = NULL;
> + struct snic_req_info *rqi = NULL;
> + spinlock_t *io_lock = NULL;
> + unsigned long flags, gflags;
> + int ret = 0;
> +
> + SNIC_HOST_INFO(snic->shost,
> + "reset_cmpl:HBA Reset Completion received.\n");
> +
> + snic_io_hdr_dec(&fwreq->hdr, &typ, &hdr_stat, &cmnd_id, &hid, &ctx);
> + SNIC_SCSI_DBG(snic->shost,
> + "reset_cmpl: type = %x, hdr_stat = %x, cmnd_id = %x, hid = %x, ctx = %llx\n",
> + typ, hdr_stat, cmnd_id, hid, ctx);
> +
> + /* spl case, host reset issued through ioctl */
> + if (cmnd_id == SCSI_NO_TAG) {
> + rqi = (struct snic_req_info *) ctx;
> + sc = rqi->sc;
> +
> + goto ioctl_hba_rst;
> + }
> +
> + if (cmnd_id >= snic->max_tag_id) {
> + SNIC_HOST_ERR(snic->shost,
> + "reset_cmpl: Tag 0x%x out of Range,HdrStat %s\n",
> + cmnd_id, snic_io_status_to_str(hdr_stat));
> + SNIC_BUG_ON(1);
> +
> + return 1;
> + }
> +
> + sc = scsi_host_find_tag(snic->shost, cmnd_id);
> +ioctl_hba_rst:
> + if (!sc) {
> + atomic64_inc(&snic->s_stats.io.sc_null);
> + SNIC_HOST_ERR(snic->shost,
> + "reset_cmpl: sc is NULL - Hdr Stat %s Tag 0x%x\n",
> + snic_io_status_to_str(hdr_stat), cmnd_id);
> + ret = 1;
> +
> + return ret;
> + }
> +
> + io_lock = snic_io_lock_hash(snic, sc);
> + spin_lock_irqsave(io_lock, flags);
> +
> + if (!snic->remove_wait) {
> + spin_unlock_irqrestore(io_lock, flags);
> + SNIC_HOST_ERR(snic->shost,
> + "reset_cmpl:host reset completed after timout\n");
> + ret = 1;
> +
> + return ret;
> + }
> +
> + rqi = (struct snic_req_info *) CMD_SP(sc);
> + WARN_ON_ONCE(!rqi);
> +
> + if (!rqi) {
> + atomic64_inc(&snic->s_stats.io.req_null);
> + spin_unlock_irqrestore(io_lock, flags);
> + CMD_FLAGS(sc) |= SNIC_IO_ABTS_TERM_REQ_NULL;
> + SNIC_HOST_ERR(snic->shost,
> + "reset_cmpl: rqi is null,Hdr stat %s Tag 0x%x sc 0x%p flags 0x%llx\n",
> + snic_io_status_to_str(hdr_stat), cmnd_id, sc,
> + CMD_FLAGS(sc));
> +
> + ret = 1;
> +
> + return ret;
> + }
> + /* stats */
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + /* scsi cleanup */
> + snic_hba_reset_scsi_cleanup(snic, sc);
> +
> + SNIC_BUG_ON(snic_get_state(snic) != SNIC_OFFLINE &&
> + snic_get_state(snic) != SNIC_FWRESET);
> +
> + /* Careful locking between snic_lock and io lock */
> + spin_lock_irqsave(io_lock, flags);
> + spin_lock_irqsave(&snic->snic_lock, gflags);
> + if (snic_get_state(snic) == SNIC_FWRESET)
> + snic_set_state(snic, SNIC_ONLINE);
> + spin_unlock_irqrestore(&snic->snic_lock, gflags);
> +
> + if (snic->remove_wait)
> + complete(snic->remove_wait);
> +
> + spin_unlock_irqrestore(io_lock, flags);
> + atomic64_inc(&snic->s_stats.reset.hba_reset_cmpl);
> +
> + ret = 0;
> + /* Rediscovery is for SAN */
> + if (snic->config.xpt_type == SNIC_DAS)
> + return ret;
> +
> + SNIC_SCSI_DBG(snic->shost, "reset_cmpl: Queuing discovery work.\n");
> + queue_work(snic_glob->event_q, &snic->disc_work);
> +
> + return ret;
> +}
So the hba reset relies on the firmware to process the queues?
Hmm. But if the firmware is hosed it wouldn't process commands, right?
I would have really, _really_, designed an out-of-band mechanism
for doing an HBA reset.
Having the HBA reset to rely on functionality which only is established
during startup is ... dangerous, to say the least.
We've been through this with the various RAID controller.
Eventually everyone of them implemented an out-of-band firmware reset...
> +
> +static void
> +snic_msg_ack_handler(struct snic *snic, struct snic_fw_req *fwreq)
> +{
> + SNIC_HOST_INFO(snic->shost, "Message Ack Received.\n");
> +
> + SNIC_ASSERT_NOT_IMPL(1);
> +}
> +
> +static void
> +snic_aen_handler(struct snic *snic, struct snic_fw_req *fwreq)
> +{
> + u8 typ, hdr_stat;
> + u32 cmnd_id, hid;
> + u64 ctx;
> + struct snic_async_evnotify *aen = &fwreq->u.async_ev;
> +
> + snic_io_hdr_dec(&fwreq->hdr, &typ, &hdr_stat, &cmnd_id, &hid, &ctx);
> + SNIC_SCSI_DBG(snic->shost,
> + "aen: type = %x, hdr_stat = %x, cmnd_id = %x, hid = %x, ctx = %llx\n",
> + typ, hdr_stat, cmnd_id, hid, ctx);
> +
> + switch (aen->ev_id) {
> + case SNIC_EV_TGT_OFFLINE:
> + SNIC_HOST_INFO(snic->shost, "aen:TGT_OFFLINE Event Recvd.\n");
> + break;
> +
> + case SNIC_EV_TGT_ONLINE:
> + SNIC_HOST_INFO(snic->shost, "aen:TGT_ONLINE Event Recvd.\n");
> + break;
> +
> + case SNIC_EV_LUN_OFFLINE:
> + SNIC_HOST_INFO(snic->shost, "aen:LUN_OFFLINE Event Recvd.\n");
> + break;
> +
> + case SNIC_EV_LUN_ONLINE:
> + SNIC_HOST_INFO(snic->shost, "aen:LUN_ONLINE Event Recvd.\n");
> + break;
> +
> + case SNIC_EV_CONF_CHG:
> + SNIC_HOST_INFO(snic->shost, "aen:Config Change Event Recvd.\n");
> + break;
> +
> + case SNIC_EV_TGT_ADDED:
> + SNIC_HOST_INFO(snic->shost, "aen:TGT_ADD Event Recvd.\n");
> + break;
> +
> + case SNIC_EV_TGT_DELTD:
> + SNIC_HOST_INFO(snic->shost, "aen:TGT_DEL Event Recvd.\n");
> + break;
> +
> + case SNIC_EV_LUN_ADDED:
> + SNIC_HOST_INFO(snic->shost, "aen:LUN_ADD Event Recvd.\n");
> + break;
> +
> + case SNIC_EV_LUN_DELTD:
> + SNIC_HOST_INFO(snic->shost, "aen:LUN_DEL Event Recvd.\n");
> + break;
> +
> + case SNIC_EV_DISC_CMPL:
> + SNIC_HOST_INFO(snic->shost, "aen:DISC_CMPL Event Recvd.\n");
> + break;
> +
> + default:
> + SNIC_HOST_INFO(snic->shost, "aen:Unknown Event Recvd.\n");
> + SNIC_BUG_ON(1);
> + break;
> + }
> +
> + SNIC_ASSERT_NOT_IMPL(1);
> +} /* end of snic_aen_handler */
> +
> +/*
> + * snic_io_cmpl_handler
> + * Routine to process CQ entries(IO Completions) posted by fw.
> + */
> +static int
> +snic_io_cmpl_handler(struct vnic_dev *vdev,
> + unsigned int cq_idx,
> + struct snic_fw_req *fwreq)
> +{
> + struct snic *snic = vnic_dev_priv(vdev);
> + u64 start = jiffies, cmpl_time;
> +
> + snic_print_desc(__func__, (char *)fwreq, sizeof(*fwreq));
> +
> + /* Update FW Stats */
> + if ((fwreq->hdr.type >= SNIC_RSP_REPORT_TGTS_CMPL) &&
> + (fwreq->hdr.type <= SNIC_RSP_BOOT_LUNS_CMPL))
> + atomic64_dec(&snic->s_stats.fw.actv_reqs);
> +
> + SNIC_BUG_ON((fwreq->hdr.type > SNIC_RSP_BOOT_LUNS_CMPL) &&
> + (fwreq->hdr.type < SNIC_MSG_ASYNC_EVNOTIFY));
> +
> + /* Check for snic subsys errors */
> + switch (fwreq->hdr.status) {
> + case SNIC_STAT_NOT_READY: /* XPT yet to initialize */
> + SNIC_HOST_ERR(snic->shost,
> + "sNIC SubSystem is NOT Ready.\n");
> + break;
> +
> + case SNIC_STAT_FATAL_ERROR: /* XPT Error */
> + SNIC_HOST_ERR(snic->shost,
> + "sNIC SubSystem in Unrecoverable State.\n");
> + break;
> + }
> +
> + switch (fwreq->hdr.type) {
> + case SNIC_RSP_EXCH_VER_CMPL:
> + snic_io_exch_ver_cmpl_handler(snic, fwreq);
> + break;
> +
> + case SNIC_RSP_REPORT_TGTS_CMPL:
> + snic_report_tgt_cmpl_handler(snic, fwreq);
> + break;
> +
> + case SNIC_RSP_ICMND_CMPL:
> + snic_icmnd_cmpl_handler(snic, fwreq);
> + break;
> +
> + case SNIC_RSP_ITMF_CMPL:
> + snic_itmf_cmpl_handler(snic, fwreq);
> + break;
> +
> + case SNIC_RSP_HBA_RESET_CMPL:
> + snic_hba_reset_cmpl_handler(snic, fwreq);
> + break;
> +
> + case SNIC_MSG_ACK:
> + snic_msg_ack_handler(snic, fwreq);
> + break;
> +
> + case SNIC_MSG_ASYNC_EVNOTIFY:
> + snic_aen_handler(snic, fwreq);
> + break;
> +
> + default:
> + SNIC_BUG_ON(1);
> + SNIC_SCSI_DBG(snic->shost,
> + "Unknown Firmwqre completion request type %d\n",
> + fwreq->hdr.type);
> + break;
> + }
> +
> + /* Update Stats */
> + cmpl_time = jiffies - start;
> + if (cmpl_time > atomic64_read(&snic->s_stats.io.max_cmpl_time))
> + atomic64_set(&snic->s_stats.io.max_cmpl_time, cmpl_time);
> +
> + return 0;
> +} /* end of snic_io_cmpl_handler */
> +
> +/*
> + * snic_fwcq_cmpl_handler
> + * Routine to process fwCQ
> + * This CQ is independent, and not associated with wq/rq/wq_copy queues
> + */
> +int
> +snic_fwcq_cmpl_handler(struct snic *snic, int io_cmpl_work)
> +{
> + unsigned int num_ent = 0; /* number cq entries processed */
> + unsigned int cq_idx;
> + unsigned int nent_per_cq;
> + struct snic_misc_stats *misc_stats = &snic->s_stats.misc;
> +
> + for (cq_idx = snic->wq_count; cq_idx < snic->cq_count; cq_idx++) {
> + nent_per_cq = vnic_cq_fw_service(&snic->cq[cq_idx],
> + snic_io_cmpl_handler,
> + io_cmpl_work);
> + num_ent += nent_per_cq;
> +
> + if (nent_per_cq > atomic64_read(&misc_stats->max_cq_ents))
> + atomic64_set(&misc_stats->max_cq_ents, nent_per_cq);
> + }
> +
> + return num_ent;
> +} /* end of snic_fwcq_cmpl_handler */
> +
> +/*
> + * snic_queue_itmf_req: Common API to queue Task Management requests.
> + * Use rqi->tm_tag for passing special tags.
> + * @req_id : aborted request's tag, -1 for lun reset.
> + */
> +static int
> +snic_queue_itmf_req(struct snic *snic,
> + struct snic_host_req *tmreq,
> + struct scsi_cmnd *sc,
> + u32 tmf,
> + u32 req_id)
> +{
> + struct snic_req_info *rqi = req_to_rqi(tmreq);
> + struct scsi_lun lun;
> + int tm_tag = snic_cmd_tag(sc) | rqi->tm_tag;
> + int ret = 0;
> +
> + SNIC_BUG_ON(!rqi);
> + SNIC_BUG_ON(!rqi->tm_tag);
> +
> + /* fill in lun info */
> + int_to_scsilun(sc->device->lun, &lun);
> +
> + /* Initialize snic_host_req: itmf */
> + snic_itmf_init(tmreq,
> + tm_tag,
> + snic->config.hid,
> + (u64) rqi,
> + 0 /* flags */,
> + req_id, /* Command to be aborted. */
> + rqi->tgt_id,
> + lun.scsi_lun,
> + tmf);
> +
> + /*
> + * In case of multiple aborts on same cmd,
> + * use try_wait_for_completion and completion_done() to check
> + * whether it queues aborts even after completion of abort issued
> + * prior.SNIC_BUG_ON(completion_done(&rqi->done));
> + */
> +
> + ret = snic_queue_wq_desc(snic, tmreq, sizeof(*tmreq));
> + if (ret)
> + SNIC_HOST_ERR(snic->shost,
> + "qitmf:Queuing ITMF(%d) Req sc %p, rqi %p, req_id %d tag %d Failed, ret = %d\n",
> + tmf, sc, rqi, req_id, snic_cmd_tag(sc), ret);
> + else
> + SNIC_SCSI_DBG(snic->shost,
> + "qitmf:Queuing ITMF(%d) Req sc %p, rqi %p, req_id %d, tag %d (req_id)- Success.",
> + tmf, sc, rqi, req_id, snic_cmd_tag(sc));
> +
> + return ret;
> +} /* end of snic_queue_itmf_req */
> +
> +static int
> +snic_issue_tm_req(struct snic *snic,
> + struct snic_req_info *rqi,
> + struct scsi_cmnd *sc,
> + int tmf)
> +{
> + struct snic_host_req *tmreq = NULL;
> + int req_id = 0, tag = snic_cmd_tag(sc);
> + int ret = 0;
> +
> + if (snic_get_state(snic) == SNIC_FWRESET)
> + return -EBUSY;
> +
> + atomic_inc(&snic->ios_inflight);
> +
> + SNIC_SCSI_DBG(snic->shost,
> + "issu_tmreq: Task mgmt req %d. rqi %p w/ tag %x\n",
> + tmf, rqi, tag);
> +
> +
> + if (tmf == SNIC_ITMF_LUN_RESET) {
> + tmreq = snic_dr_req_init(snic, rqi);
> + req_id = SCSI_NO_TAG;
> + } else {
> + tmreq = snic_abort_req_init(snic, rqi);
> + req_id = tag;
> + }
> +
> + if (!tmreq) {
> + ret = -ENOMEM;
> +
> + goto tmreq_err;
> + }
> +
> + ret = snic_queue_itmf_req(snic, tmreq, sc, tmf, req_id);
> + if (ret)
> + goto tmreq_err;
> +
> + ret = 0;
> +
> +tmreq_err:
> + if (ret) {
> + SNIC_HOST_ERR(snic->shost,
> + "issu_tmreq: Queing ITMF(%d) Req, sc %p rqi %p req_id %d tag %x fails err = %d\n",
> + tmf, sc, rqi, req_id, tag, ret);
> + } else {
> + SNIC_SCSI_DBG(snic->shost,
> + "issu_tmreq: Queuing ITMF(%d) Req, sc %p, rqi %p, req_id %d tag %x - Success.\n",
> + tmf, sc, rqi, req_id, tag);
> + }
> +
> + atomic_dec(&snic->ios_inflight);
> +
> + return ret;
> +}
> +
> +/*
> + * snic_queue_abort_req : Queues abort req to WQ
> + */
> +static int
> +snic_queue_abort_req(struct snic *snic,
> + struct snic_req_info *rqi,
> + struct scsi_cmnd *sc,
> + int tmf)
> +{
> + SNIC_SCSI_DBG(snic->shost, "q_abtreq: sc %p, rqi %p, tag %x, tmf %d\n",
> + sc, rqi, snic_cmd_tag(sc), tmf);
> +
> + /* Add special tag for abort */
> + rqi->tm_tag |= SNIC_TAG_ABORT;
> +
> + return snic_issue_tm_req(snic, rqi, sc, tmf);
> +}
> +
> +/*
> + * snic_abort_finish : called by snic_abort_cmd on queuing abort successfully.
> + */
> +static int
> +snic_abort_finish(struct snic *snic, struct scsi_cmnd *sc)
> +{
> + struct snic_req_info *rqi = NULL;
> + spinlock_t *io_lock = NULL;
> + unsigned long flags;
> + int ret = 0, tag = snic_cmd_tag(sc);
> +
> + io_lock = snic_io_lock_hash(snic, sc);
> + spin_lock_irqsave(io_lock, flags);
> + rqi = (struct snic_req_info *) CMD_SP(sc);
> + if (!rqi) {
> + atomic64_inc(&snic->s_stats.io.req_null);
> + CMD_FLAGS(sc) |= SNIC_IO_ABTS_TERM_REQ_NULL;
> +
> + SNIC_SCSI_DBG(snic->shost,
> + "abt_fini:req info is null tag 0x%x, sc 0x%p flags 0x%llx\n",
> + tag, sc, CMD_FLAGS(sc));
> + ret = FAILED;
> +
> + goto abort_fail;
> + }
> +
> + rqi->abts_done = NULL;
> +
> + ret = FAILED;
> +
> + /* Check the abort status. */
> + switch (CMD_ABTS_STATUS(sc)) {
> + case SNIC_INVALID_CODE:
> + /* Firmware didn't complete abort req, timedout */
> + CMD_FLAGS(sc) |= SNIC_IO_ABTS_TIMEDOUT;
> + atomic64_inc(&snic->s_stats.abts.drv_tmo);
> + SNIC_SCSI_DBG(snic->shost,
> + "abt_fini:sc %p Tag %x Driver Timeout.flags 0x%llx\n",
> + sc, snic_cmd_tag(sc), CMD_FLAGS(sc));
> + /* do not release snic request in timedout case */
> + rqi = NULL;
> +
> + goto abort_fail;
> +
> + case SNIC_STAT_IO_SUCCESS:
> + case SNIC_STAT_IO_NOT_FOUND:
> + ret = SUCCESS;
> + break;
> +
> + default:
> + /* Firmware completed abort with error */
> + ret = FAILED;
> + break;
> + }
> +
> + CMD_SP(sc) = NULL;
> + SNIC_HOST_INFO(snic->shost,
> + "abt_fini: Tag %x, Cmpl Status %s flags 0x%llx\n",
> + tag, snic_io_status_to_str(CMD_ABTS_STATUS(sc)),
> + CMD_FLAGS(sc));
> +
> +abort_fail:
> + spin_unlock_irqrestore(io_lock, flags);
> + if (rqi)
> + snic_release_req_buf(snic, rqi, sc);
> +
> + return ret;
> +} /* end of snic_abort_finish */
> +
> +/*
> + * snic_send_abort_and_wait : Issues Abort, and Waits
> + */
> +static int
> +snic_send_abort_and_wait(struct snic *snic, struct scsi_cmnd *sc)
> +{
> + struct snic_req_info *rqi = NULL;
> + enum snic_ioreq_state sv_state;
> + struct snic_tgt *tgt = NULL;
> + spinlock_t *io_lock = NULL;
> + DECLARE_COMPLETION_ONSTACK(tm_done);
> + unsigned long flags;
> + int ret = 0, tmf = 0, tag = snic_cmd_tag(sc);
> +
> + tgt = starget_to_tgt(scsi_target(sc->device));
> + if ((snic_tgt_chkready(tgt) != 0) && (tgt->tdata.typ == SNIC_TGT_SAN))
> + tmf = SNIC_ITMF_ABTS_TASK_TERM;
> + else
> + tmf = SNIC_ITMF_ABTS_TASK;
> +
> + /* stats */
> +
> + io_lock = snic_io_lock_hash(snic, sc);
> +
> + /*
> + * Avoid a race between SCSI issuing the abort and the device
> + * completing the command.
> + *
> + * If the command is already completed by fw_cmpl code,
> + * we just return SUCCESS from here. This means that the abort
> + * succeeded. In the SCSI ML, since the timeout for command has
> + * happend, the completion wont actually complete the command
> + * and it will be considered as an aborted command
> + *
> + * The CMD_SP will not be cleared except while holding io_lock
> + */
> + spin_lock_irqsave(io_lock, flags);
> + rqi = (struct snic_req_info *) CMD_SP(sc);
> + if (!rqi) {
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + SNIC_HOST_ERR(snic->shost,
> + "abt_cmd: rqi is null. Tag %d flags 0x%llx\n",
> + tag, CMD_FLAGS(sc));
> +
> + ret = SUCCESS;
> +
> + goto send_abts_end;
> + }
> +
> + rqi->abts_done = &tm_done;
> + if (CMD_STATE(sc) == SNIC_IOREQ_ABTS_PENDING) {
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + ret = 0;
> + goto abts_pending;
> + }
> + SNIC_BUG_ON(!rqi->abts_done);
> +
> + /* Save Command State, should be restored on failed to Queue. */
> + sv_state = CMD_STATE(sc);
> +
> + /*
> + * Command is still pending, need to abort it
> + * If the fw completes the command after this point,
> + * the completion won't be done till mid-layer, since abot
> + * has already started.
> + */
> + CMD_STATE(sc) = SNIC_IOREQ_ABTS_PENDING;
> + CMD_ABTS_STATUS(sc) = SNIC_INVALID_CODE;
> +
> + SNIC_SCSI_DBG(snic->shost, "send_abt_cmd: TAG 0x%x\n", tag);
> +
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + /* Now Queue the abort command to firmware */
> + ret = snic_queue_abort_req(snic, rqi, sc, tmf);
> + if (ret) {
> + SNIC_HOST_ERR(snic->shost,
> + "send_abt_cmd: IO w/ Tag 0x%x fail w/ err %d flags 0x%llx\n",
> + tag, ret, CMD_FLAGS(sc));
> +
> + spin_lock_irqsave(io_lock, flags);
> + /* Restore Command's previous state */
> + CMD_STATE(sc) = sv_state;
> + rqi = (struct snic_req_info *) CMD_SP(sc);
> + if (rqi)
> + rqi->abts_done = NULL;
> + spin_unlock_irqrestore(io_lock, flags);
> + ret = FAILED;
> +
> + goto send_abts_end;
> + }
> +
> + spin_lock_irqsave(io_lock, flags);
> + if (tmf == SNIC_ITMF_ABTS_TASK) {
> + CMD_FLAGS(sc) |= SNIC_IO_ABTS_ISSUED;
> + atomic64_inc(&snic->s_stats.abts.num);
> + } else {
> + /* term stats */
> + CMD_FLAGS(sc) |= SNIC_IO_TERM_ISSUED;
> + }
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + SNIC_SCSI_DBG(snic->shost,
> + "send_abt_cmd: sc %p Tag %x flags 0x%llx\n",
> + sc, tag, CMD_FLAGS(sc));
> +
> +
> + ret = 0;
> +
> +abts_pending:
> + /*
> + * Queued an abort IO, wait for its completion.
> + * Once the fw completes the abort command, it will
> + * wakeup this thread.
> + */
> + wait_for_completion_timeout(&tm_done, SNIC_ABTS_TIMEOUT);
> +
> +send_abts_end:
> + return ret;
> +} /* end of snic_send_abort_and_wait */
> +
> +/*
> + * This function is exported to SCSI for sending abort cmnds.
> + * A SCSI IO is represent by snic_ioreq in the driver.
> + * The snic_ioreq is linked to the SCSI Cmd, thus a link with the ULP'S IO
> + */
> +int
> +snic_abort_cmd(struct scsi_cmnd *sc)
> +{
> + struct snic *snic = shost_priv(sc->device->host);
> + int ret = SUCCESS, tag = snic_cmd_tag(sc);
> + u32 start_time = jiffies;
> +
> + SNIC_SCSI_DBG(snic->shost, "abt_cmd:sc %p :0x%x :req = %p :tag = %d\n",
> + sc, sc->cmnd[0], sc->request, tag);
> +
> + if (unlikely(snic_get_state(snic) != SNIC_ONLINE)) {
> + SNIC_HOST_ERR(snic->shost,
> + "abt_cmd: tag %x Parent Devs are not rdy\n",
> + tag);
> + ret = FAST_IO_FAIL;
> +
> + goto abort_end;
> + }
> +
> +
> + ret = snic_send_abort_and_wait(snic, sc);
> + if (ret)
> + goto abort_end;
> +
> + ret = snic_abort_finish(snic, sc);
> +
> +abort_end:
> + SNIC_TRC(snic->shost->host_no, tag, sc,
> + jiffies_to_msecs(jiffies - start_time), 0,
> + SNIC_TRC_CMD(sc), SNIC_TRC_CMD_STATE_FLAGS(sc));
> +
> + SNIC_SCSI_DBG(snic->shost,
> + "abts: Abort Req Status = %s\n",
> + (ret == SUCCESS) ? "SUCCESS" :
> + ((ret == FAST_IO_FAIL) ? "FAST_IO_FAIL" : "FAILED"));
> +
> + return ret;
> +}
> +
> +
> +
> +static int
> +snic_is_abts_pending(struct snic *snic, struct scsi_cmnd *lr_sc)
> +{
> + struct snic_req_info *rqi = NULL;
> + struct scsi_cmnd *sc = NULL;
> + struct scsi_device *lr_sdev = NULL;
> + spinlock_t *io_lock = NULL;
> + u32 tag;
> + unsigned long flags;
> +
> + if (lr_sc)
> + lr_sdev = lr_sc->device;
> +
> + /* walk through the tag map, an dcheck if IOs are still pending in fw*/
> + for (tag = 0; tag < snic->max_tag_id; tag++) {
> + io_lock = snic_io_lock_tag(snic, tag);
> +
> + spin_lock_irqsave(io_lock, flags);
> + sc = scsi_host_find_tag(snic->shost, tag);
> +
> + if (!sc || (lr_sc && (sc->device != lr_sdev || sc == lr_sc))) {
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + continue;
> + }
> +
> + rqi = (struct snic_req_info *) CMD_SP(sc);
> + if (!rqi) {
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + continue;
> + }
> +
> + /*
> + * Found IO that is still pending w/ firmware and belongs to
> + * the LUN that is under reset, if lr_sc != NULL
> + */
> + SNIC_SCSI_DBG(snic->shost, "Found IO in %s on LUN\n",
> + snic_ioreq_state_to_str(CMD_STATE(sc)));
> +
> + if (CMD_STATE(sc) == SNIC_IOREQ_ABTS_PENDING) {
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + return 1;
> + }
> +
> + spin_unlock_irqrestore(io_lock, flags);
> + }
> +
> + return 0;
> +} /* end of snic_is_abts_pending */
> +
> +static int
> +snic_dr_clean_single_req(struct snic *snic,
> + u32 tag,
> + struct scsi_device *lr_sdev)
> +{
> + struct snic_req_info *rqi = NULL;
> + struct snic_tgt *tgt = NULL;
> + struct scsi_cmnd *sc = NULL;
> + spinlock_t *io_lock = NULL;
> + u32 sv_state = 0, tmf = 0;
> + DECLARE_COMPLETION_ONSTACK(tm_done);
> + unsigned long flags;
> + int ret = 0;
> +
> + io_lock = snic_io_lock_tag(snic, tag);
> + spin_lock_irqsave(io_lock, flags);
> + sc = scsi_host_find_tag(snic->shost, tag);
> +
> + /* Ignore Cmd that don't belong to Lun Reset device */
> + if (!sc || sc->device != lr_sdev)
> + goto skip_clean;
> +
> + rqi = (struct snic_req_info *) CMD_SP(sc);
> +
> + if (!rqi)
> + goto skip_clean;
> +
> +
> + if (CMD_STATE(sc) == SNIC_IOREQ_ABTS_PENDING)
> + goto skip_clean;
> +
> +
> + if ((CMD_FLAGS(sc) & SNIC_DEVICE_RESET) &&
> + (!(CMD_FLAGS(sc) & SNIC_DEV_RST_ISSUED))) {
> +
> + SNIC_SCSI_DBG(snic->shost,
> + "clean_single_req: devrst is not pending sc 0x%p\n",
> + sc);
> +
> + goto skip_clean;
> + }
> +
> + SNIC_SCSI_DBG(snic->shost,
> + "clean_single_req: Found IO in %s on lun\n",
> + snic_ioreq_state_to_str(CMD_STATE(sc)));
> +
> + /* Save Command State */
> + sv_state = CMD_STATE(sc);
> +
> + /*
> + * Any pending IO issued prior to reset is expected to be
> + * in abts pending state, if not we need to set SNIC_IOREQ_ABTS_PENDING
> + * to indicate the IO is abort pending.
> + * When IO is completed, the IO will be handed over and handled
> + * in this function.
> + */
> +
> + CMD_STATE(sc) = SNIC_IOREQ_ABTS_PENDING;
> + SNIC_BUG_ON(rqi->abts_done);
> +
> + if (CMD_FLAGS(sc) & SNIC_DEVICE_RESET) {
> + rqi->tm_tag = SNIC_TAG_DEV_RST;
> +
> + SNIC_SCSI_DBG(snic->shost,
> + "clean_single_req:devrst sc 0x%p\n", sc);
> + }
> +
> + CMD_ABTS_STATUS(sc) = SNIC_INVALID_CODE;
> + rqi->abts_done = &tm_done;
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + tgt = starget_to_tgt(scsi_target(sc->device));
> + if ((snic_tgt_chkready(tgt) != 0) && (tgt->tdata.typ == SNIC_TGT_SAN))
> + tmf = SNIC_ITMF_ABTS_TASK_TERM;
> + else
> + tmf = SNIC_ITMF_ABTS_TASK;
> +
> + /* Now queue the abort command to firmware */
> + ret = snic_queue_abort_req(snic, rqi, sc, tmf);
> + if (ret) {
> + SNIC_HOST_ERR(snic->shost,
> + "clean_single_req_err:sc %p, tag %d abt failed. tm_tag %d flags 0x%llx\n",
> + sc, tag, rqi->tm_tag, CMD_FLAGS(sc));
> +
> + spin_lock_irqsave(io_lock, flags);
> + rqi = (struct snic_req_info *) CMD_SP(sc);
> + if (rqi)
> + rqi->abts_done = NULL;
> +
> + /* Restore Command State */
> + if (CMD_STATE(sc) == SNIC_IOREQ_ABTS_PENDING)
> + CMD_STATE(sc) = sv_state;
> +
> + ret = 1;
> + goto skip_clean;
> + }
> +
> + spin_lock_irqsave(io_lock, flags);
> + if (CMD_FLAGS(sc) & SNIC_DEVICE_RESET)
> + CMD_FLAGS(sc) |= SNIC_DEV_RST_TERM_ISSUED;
> +
> + CMD_FLAGS(sc) |= SNIC_IO_INTERNAL_TERM_ISSUED;
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + wait_for_completion_timeout(&tm_done, SNIC_ABTS_TIMEOUT);
> +
> + /* Recheck cmd state to check if it now aborted. */
> + spin_lock_irqsave(io_lock, flags);
> + rqi = (struct snic_req_info *) CMD_SP(sc);
> + if (!rqi) {
> + CMD_FLAGS(sc) |= SNIC_IO_ABTS_TERM_REQ_NULL;
> + goto skip_clean;
> + }
> + rqi->abts_done = NULL;
> +
> + /* if abort is still pending w/ fw, fail */
> + if (CMD_ABTS_STATUS(sc) == SNIC_INVALID_CODE) {
> + SNIC_HOST_ERR(snic->shost,
> + "clean_single_req_err:sc %p tag %d abt still pending w/ fw, tm_tag %d flags 0x%llx\n",
> + sc, tag, rqi->tm_tag, CMD_FLAGS(sc));
> +
> + CMD_FLAGS(sc) |= SNIC_IO_ABTS_TERM_DONE;
> + ret = 1;
> +
> + goto skip_clean;
> + }
> +
> + CMD_STATE(sc) = SNIC_IOREQ_ABTS_COMPLETE;
> + CMD_SP(sc) = NULL;
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + snic_release_req_buf(snic, rqi, sc);
> +
> + ret = 0;
> +
> + return ret;
> +
> +skip_clean:
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + return ret;
> +} /* end of snic_dr_clean_single_req */
> +
> +static int
> +snic_dr_clean_pending_req(struct snic *snic, struct scsi_cmnd *lr_sc)
> +{
> + struct scsi_device *lr_sdev = lr_sc->device;
> + u32 tag = 0;
> + int ret = FAILED;
> +
> + for (tag = 0; tag < snic->max_tag_id; tag++) {
> + if (tag == snic_cmd_tag(lr_sc))
> + continue;
> +
> + ret = snic_dr_clean_single_req(snic, tag, lr_sdev);
> + if (ret) {
> + SNIC_HOST_ERR(snic->shost, "clean_err:tag = %d\n", tag);
> +
> + goto clean_err;
> + }
> + }
> +
> + schedule_timeout(msecs_to_jiffies(100));
> +
> + /* Walk through all the cmds and check abts status. */
> + if (snic_is_abts_pending(snic, lr_sc)) {
> + ret = FAILED;
> +
> + goto clean_err;
> + }
> +
> + ret = 0;
> + SNIC_SCSI_DBG(snic->shost, "clean_pending_req: Success.\n");
> +
> + return ret;
> +
> +clean_err:
> + ret = FAILED;
> + SNIC_HOST_ERR(snic->shost,
> + "Failed to Clean Pending IOs on %s device.\n",
> + dev_name(&lr_sdev->sdev_gendev));
> +
> + return ret;
> +
> +} /* end of snic_dr_clean_pending_req */
> +
> +/*
> + * snic_dr_finish : Called by snic_device_reset
> + */
> +static int
> +snic_dr_finish(struct snic *snic, struct scsi_cmnd *sc)
> +{
> + struct snic_req_info *rqi = NULL;
> + spinlock_t *io_lock = NULL;
> + unsigned long flags;
> + int lr_res = 0;
> + int ret = FAILED;
> +
> + io_lock = snic_io_lock_hash(snic, sc);
> + spin_lock_irqsave(io_lock, flags);
> + rqi = (struct snic_req_info *) CMD_SP(sc);
> + if (!rqi) {
> + spin_unlock_irqrestore(io_lock, flags);
> + SNIC_SCSI_DBG(snic->shost,
> + "dr_fini: rqi is null tag 0x%x sc 0x%p flags 0x%llx\n",
> + snic_cmd_tag(sc), sc, CMD_FLAGS(sc));
> +
> + ret = FAILED;
> + goto dr_fini_end;
> + }
> +
> + rqi->dr_done = NULL;
> +
> + lr_res = CMD_LR_STATUS(sc);
> +
> + switch (lr_res) {
> + case SNIC_INVALID_CODE:
> + /* stats */
> + SNIC_SCSI_DBG(snic->shost,
> + "dr_fini: Tag %x Dev Reset Timedout. flags 0x%llx\n",
> + snic_cmd_tag(sc), CMD_FLAGS(sc));
> +
> + CMD_FLAGS(sc) |= SNIC_DEV_RST_TIMEDOUT;
> + ret = FAILED;
> +
> + goto dr_failed;
> +
> + case SNIC_STAT_IO_SUCCESS:
> + SNIC_SCSI_DBG(snic->shost,
> + "dr_fini: Tag %x Dev Reset cmpl\n",
> + snic_cmd_tag(sc));
> + ret = 0;
> + break;
> +
> + default:
> + SNIC_HOST_ERR(snic->shost,
> + "dr_fini:Device Reset completed& failed.Tag = %x lr_status %s flags 0x%llx\n",
> + snic_cmd_tag(sc),
> + snic_io_status_to_str(lr_res), CMD_FLAGS(sc));
> + ret = FAILED;
> + goto dr_failed;
> + }
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + /*
> + * Cleanup any IOs on this LUN that have still not completed.
> + * If any of these fail, then LUN Reset fails.
> + * Cleanup cleans all commands on this LUN except
> + * the lun reset command. If all cmds get cleaned, the LUN Reset
> + * succeeds.
> + */
> +
> + ret = snic_dr_clean_pending_req(snic, sc);
> + if (ret) {
> + spin_lock_irqsave(io_lock, flags);
> + SNIC_SCSI_DBG(snic->shost,
> + "dr_fini: Device Reset Failed since could not abort all IOs. Tag = %x.\n",
> + snic_cmd_tag(sc));
> + rqi = (struct snic_req_info *) CMD_SP(sc);
> +
> + goto dr_failed;
> + } else {
> + /* Cleanup LUN Reset Command */
> + spin_lock_irqsave(io_lock, flags);
> + rqi = (struct snic_req_info *) CMD_SP(sc);
> + if (rqi)
> + ret = SUCCESS; /* Completed Successfully */
> + else
> + ret = FAILED;
> + }
> +
> +dr_failed:
> + SNIC_BUG_ON(!spin_is_locked(io_lock));
> + if (rqi)
> + CMD_SP(sc) = NULL;
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + if (rqi)
> + snic_release_req_buf(snic, rqi, sc);
> +
> +dr_fini_end:
> + return ret;
> +} /* end of snic_dr_finish */
> +
> +static int
> +snic_queue_dr_req(struct snic *snic,
> + struct snic_req_info *rqi,
> + struct scsi_cmnd *sc)
> +{
> + /* Add special tag for device reset */
> + rqi->tm_tag |= SNIC_TAG_DEV_RST;
> +
> + return snic_issue_tm_req(snic, rqi, sc, SNIC_ITMF_LUN_RESET);
> +}
> +
> +static int
> +snic_send_dr_and_wait(struct snic *snic, struct scsi_cmnd *sc)
> +{
> + struct snic_req_info *rqi = NULL;
> + enum snic_ioreq_state sv_state;
> + spinlock_t *io_lock = NULL;
> + unsigned long flags;
> + DECLARE_COMPLETION_ONSTACK(tm_done);
> + int ret = FAILED, tag = snic_cmd_tag(sc);
> +
> + io_lock = snic_io_lock_hash(snic, sc);
> + spin_lock_irqsave(io_lock, flags);
> + CMD_FLAGS(sc) |= SNIC_DEVICE_RESET;
> + rqi = (struct snic_req_info *) CMD_SP(sc);
> + if (!rqi) {
> + SNIC_HOST_ERR(snic->shost,
> + "send_dr: rqi is null, Tag 0x%x flags 0x%llx\n",
> + tag, CMD_FLAGS(sc));
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + ret = FAILED;
> + goto send_dr_end;
> + }
> +
> + /* Save Command state to restore in case Queuing failed. */
> + sv_state = CMD_STATE(sc);
> +
> + CMD_STATE(sc) = SNIC_IOREQ_LR_PENDING;
> + CMD_LR_STATUS(sc) = SNIC_INVALID_CODE;
> +
> + SNIC_SCSI_DBG(snic->shost, "dr: TAG = %x\n", tag);
> +
> + rqi->dr_done = &tm_done;
> + SNIC_BUG_ON(!rqi->dr_done);
> +
> + spin_unlock_irqrestore(io_lock, flags);
> + /*
> + * The Command state is changed to IOREQ_PENDING,
> + * in this case, if the command is completed, the icmnd_cmpl will
> + * mark the cmd as completed.
> + * This logic still makes LUN Reset is inevitable.
> + */
> +
> + ret = snic_queue_dr_req(snic, rqi, sc);
> + if (ret) {
> + SNIC_HOST_ERR(snic->shost,
> + "send_dr: IO w/ Tag 0x%x Failed err = %d. flags 0x%llx\n",
> + tag, ret, CMD_FLAGS(sc));
> +
> + spin_lock_irqsave(io_lock, flags);
> + /* Restore State */
> + CMD_STATE(sc) = sv_state;
> + rqi = (struct snic_req_info *) CMD_SP(sc);
> + if (rqi)
> + rqi->dr_done = NULL;
> + /* rqi is freed in caller. */
> + spin_unlock_irqrestore(io_lock, flags);
> + ret = FAILED;
> +
> + goto send_dr_end;
> + }
> +
> + spin_lock_irqsave(io_lock, flags);
> + CMD_FLAGS(sc) |= SNIC_DEV_RST_ISSUED;
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + ret = 0;
> +
> + wait_for_completion_timeout(&tm_done, SNIC_LUN_RESET_TIMEOUT);
> +
> +send_dr_end:
> + return ret;
> +}
> +
> +/*
> + * auxillary funciton to check lun reset op is supported or not
> + * Not supported if returns 0
> + */
> +static int
> +snic_dev_reset_supported(struct scsi_device *sdev)
> +{
> + struct snic_tgt *tgt = starget_to_tgt(scsi_target(sdev));
> +
> + if (tgt->tdata.typ == SNIC_TGT_DAS)
> + return 0;
> +
> + return 1;
> +}
> +
> +static void
> +snic_unlink_and_release_req(struct snic *snic, struct scsi_cmnd *sc, int flag)
> +{
> + struct snic_req_info *rqi = NULL;
> + spinlock_t *io_lock = NULL;
> + unsigned long flags;
> + u32 start_time = jiffies;
> +
> + io_lock = snic_io_lock_hash(snic, sc);
> + spin_lock_irqsave(io_lock, flags);
> + rqi = (struct snic_req_info *) CMD_SP(sc);
> + if (rqi) {
> + start_time = rqi->start_time;
> + CMD_SP(sc) = NULL;
> + }
> +
> + CMD_FLAGS(sc) |= flag;
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + if (rqi)
> + snic_release_req_buf(snic, rqi, sc);
> +
> + SNIC_TRC(snic->shost->host_no, snic_cmd_tag(sc), sc,
> + jiffies_to_msecs(jiffies - start_time),
> + rqi, SNIC_TRC_CMD(sc), SNIC_TRC_CMD_STATE_FLAGS(sc));
> +}
> +
> +/*
> + * SCSI Eh thread issues a LUN Reset when one or more commands on a LUN
> + * fail to get aborted. It calls driver's eh_device_reset with a SCSI
> + * command on the LUN.
> + */
> +int
> +snic_device_reset(struct scsi_cmnd *sc)
> +{
> + struct Scsi_Host *shost = sc->device->host;
> + struct snic *snic = shost_priv(shost);
> + struct snic_req_info *rqi = NULL;
> + int tag = snic_cmd_tag(sc);
> + int start_time = jiffies;
> + int ret = FAILED;
> + int dr_supp = 0;
> +
> + SNIC_SCSI_DBG(shost, "dev_reset:sc %p :0x%x :req = %p :tag = %d\n",
> + sc, sc->cmnd[0], sc->request,
> + snic_cmd_tag(sc));
> + dr_supp = snic_dev_reset_supported(sc->device);
> + if (!dr_supp) {
> + /* device reset op is not supported */
> + SNIC_HOST_INFO(shost, "LUN Reset Op not supported.\n");
> + snic_unlink_and_release_req(snic, sc, SNIC_DEV_RST_NOTSUP);
> +
> + goto dev_rst_end;
> + }
> +
> + if (unlikely(snic_get_state(snic) != SNIC_ONLINE)) {
> + snic_unlink_and_release_req(snic, sc, 0);
> + SNIC_HOST_ERR(shost, "Devrst: Parent Devs are not online.\n");
> +
> + goto dev_rst_end;
> + }
> +
> + /* There is no tag when lun reset is issue through ioctl. */
> + if (unlikely(tag <= SNIC_NO_TAG)) {
> + SNIC_HOST_INFO(snic->shost,
> + "Devrst: LUN Reset Recvd thru IOCTL.\n");
> +
> + rqi = snic_req_init(snic, 0);
> + if (!rqi)
> + goto dev_rst_end;
> +
> + memset(scsi_cmd_priv(sc), 0,
> + sizeof(struct snic_internal_io_state));
> + CMD_SP(sc) = (char *)rqi;
> + CMD_FLAGS(sc) = SNIC_NO_FLAGS;
> +
> + /* Add special tag for dr coming from user spc */
> + rqi->tm_tag = SNIC_TAG_IOCTL_DEV_RST;
> + rqi->sc = sc;
> + }
> +
> + ret = snic_send_dr_and_wait(snic, sc);
> + if (ret) {
> + SNIC_HOST_ERR(snic->shost,
> + "Devrst: IO w/ Tag %x Failed w/ err = %d\n",
> + tag, ret);
> +
> + snic_unlink_and_release_req(snic, sc, 0);
> +
> + goto dev_rst_end;
> + }
> +
> + ret = snic_dr_finish(snic, sc);
> +
> +dev_rst_end:
> + SNIC_TRC(snic->shost->host_no, tag, sc,
> + jiffies_to_msecs(jiffies - start_time),
> + 0, SNIC_TRC_CMD(sc), SNIC_TRC_CMD_STATE_FLAGS(sc));
> +
> + SNIC_SCSI_DBG(snic->shost,
> + "Devrst: Returning from Device Reset : %s\n",
> + (ret == SUCCESS) ? "SUCCESS" : "FAILED");
> +
> + return ret;
> +} /* end of snic_device_reset */
> +
> +/*
> + * SCSI Error handling calls driver's eh_host_reset if all prior
> + * error handling levels return FAILED.
> + *
> + * Host Reset is the highest level of error recovery. If this fails, then
> + * host is offlined by SCSI.
> + */
> +/*
> + * snic_issue_hba_reset : Queues FW Reset Request.
> + */
> +static int
> +snic_issue_hba_reset(struct snic *snic, struct scsi_cmnd *sc)
> +{
> + struct snic_req_info *rqi = NULL;
> + struct snic_host_req *req = NULL;
> + spinlock_t *io_lock = NULL;
> + DECLARE_COMPLETION_ONSTACK(wait);
> + unsigned long flags;
> + int ret = -ENOMEM;
> +
> + rqi = snic_req_init(snic, 0);
> + if (!rqi) {
> + ret = -ENOMEM;
> +
> + goto hba_rst_end;
> + }
> +
> + if (snic_cmd_tag(sc) == SCSI_NO_TAG) {
> + memset(scsi_cmd_priv(sc), 0,
> + sizeof(struct snic_internal_io_state));
> + SNIC_HOST_INFO(snic->shost, "issu_hr:Host reset thru ioctl.\n");
> + rqi->sc = sc;
> + }
> +
> + req = rqi_to_req(rqi);
> +
> + io_lock = snic_io_lock_hash(snic, sc);
> + spin_lock_irqsave(io_lock, flags);
> + SNIC_BUG_ON(CMD_SP(sc) != NULL);
> + CMD_STATE(sc) = SNIC_IOREQ_PENDING;
> + CMD_SP(sc) = (char *) rqi;
> + CMD_FLAGS(sc) |= SNIC_IO_INITIALIZED;
> + snic->remove_wait = &wait;
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + /* Initialize Request */
> + snic_io_hdr_enc(&req->hdr, SNIC_REQ_HBA_RESET, 0, snic_cmd_tag(sc),
> + snic->config.hid, 0, (u64)rqi);
> +
> + req->u.reset.flags = 0;
> +
> + ret = snic_queue_wq_desc(snic, req, sizeof(*req));
> + if (ret) {
> + SNIC_HOST_ERR(snic->shost,
> + "issu_hr:Queuing HBA Reset Failed. w err %d\n",
> + ret);
> +
> + goto hba_rst_err;
> + }
> +
> + spin_lock_irqsave(io_lock, flags);
> + CMD_FLAGS(sc) |= SNIC_HOST_RESET_ISSUED;
> + spin_unlock_irqrestore(io_lock, flags);
> + atomic64_inc(&snic->s_stats.reset.hba_resets);
> + SNIC_HOST_INFO(snic->shost, "Queued HBA Reset Successfully.\n");
> +
> + wait_for_completion_timeout(snic->remove_wait,
> + SNIC_HOST_RESET_TIMEOUT);
> +
> + if (snic_get_state(snic) == SNIC_FWRESET) {
> + SNIC_HOST_ERR(snic->shost, "reset_cmpl: Reset Timedout.\n");
> + ret = -ETIMEDOUT;
> +
> + goto hba_rst_err;
> + }
> +
> + spin_lock_irqsave(io_lock, flags);
> + snic->remove_wait = NULL;
> + rqi = (struct snic_req_info *) CMD_SP(sc);
> + CMD_SP(sc) = NULL;
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + if (rqi)
> + snic_req_free(snic, rqi);
> +
> + ret = 0;
> +
> + return ret;
> +
> +hba_rst_err:
> + spin_lock_irqsave(io_lock, flags);
> + snic->remove_wait = NULL;
> + rqi = (struct snic_req_info *) CMD_SP(sc);
> + CMD_SP(sc) = NULL;
> + spin_unlock_irqrestore(io_lock, flags);
> +
> + if (rqi)
> + snic_req_free(snic, rqi);
> +
> +hba_rst_end:
> + SNIC_HOST_ERR(snic->shost,
> + "reset:HBA Reset Failed w/ err = %d.\n",
> + ret);
> +
> + return ret;
> +} /* end of snic_issue_hba_reset */
> +
> +int
> +snic_reset(struct Scsi_Host *shost, struct scsi_cmnd *sc)
> +{
> + struct snic *snic = shost_priv(shost);
> + enum snic_state sv_state;
> + unsigned long flags;
> + int ret = FAILED;
> +
> + /* Set snic state as SNIC_FWRESET*/
> + sv_state = snic_get_state(snic);
> +
> + spin_lock_irqsave(&snic->snic_lock, flags);
> + if (snic_get_state(snic) == SNIC_FWRESET) {
> + spin_unlock_irqrestore(&snic->snic_lock, flags);
> + SNIC_HOST_INFO(shost, "reset:prev reset is in progres\n");
> +
> + msleep(SNIC_HOST_RESET_TIMEOUT);
> + ret = SUCCESS;
> +
> + goto reset_end;
> + }
> +
> + snic_set_state(snic, SNIC_FWRESET);
> + spin_unlock_irqrestore(&snic->snic_lock, flags);
> +
> +
> + /* Wait for all the IOs that are entered in Qcmd */
> + while (atomic_read(&snic->ios_inflight))
> + schedule_timeout(msecs_to_jiffies(1));
> +
> + ret = snic_issue_hba_reset(snic, sc);
> + if (ret) {
> + SNIC_HOST_ERR(shost,
> + "reset:Host Reset Failed w/ err %d.\n",
> + ret);
> + spin_lock_irqsave(&snic->snic_lock, flags);
> + snic_set_state(snic, sv_state);
> + spin_unlock_irqrestore(&snic->snic_lock, flags);
> + atomic64_inc(&snic->s_stats.reset.hba_reset_fail);
> + ret = FAILED;
> +
> + goto reset_end;
> + }
> +
> + ret = SUCCESS;
> +
> +reset_end:
> + return ret;
> +} /* end of snic_reset */
> +
> +/*
> + * SCSI Error handling calls driver's eh_host_reset if all prior
> + * error handling levels return FAILED.
> + *
> + * Host Reset is the highest level of error recovery. If this fails, then
> + * host is offlined by SCSI.
> + */
> +int
> +snic_host_reset(struct scsi_cmnd *sc)
> +{
> + struct Scsi_Host *shost = sc->device->host;
> + u32 start_time = jiffies;
> + int ret = FAILED;
> +
> + SNIC_SCSI_DBG(shost,
> + "host reset:sc %p sc_cmd 0x%x req %p tag %d flags 0x%llx\n",
> + sc, sc->cmnd[0], sc->request,
> + snic_cmd_tag(sc), CMD_FLAGS(sc));
> +
> + ret = snic_reset(shost, sc);
> +
> + SNIC_TRC(shost->host_no, snic_cmd_tag(sc), sc,
> + jiffies_to_msecs(jiffies - start_time),
> + 0, SNIC_TRC_CMD(sc), SNIC_TRC_CMD_STATE_FLAGS(sc));
> +
> + return ret;
> +} /* end of snic_host_reset */
> +
As mentioned above:
The send/receive queues are established during device initialization.
So if this information get garbled for whatever reason or if the
firmware becomes hosed you've got no way of resetting the HBA.
I would rather see if there wouldn't be an out-of-be mechanism to reset
the HBA (via PCI reset, say), and then send a new initialisation command.
This way you can be sure to reset the HBA correctly.
Cheers,
Hannes
--
Dr. Hannes Reinecke zSeries & Storage
hare@suse.de +49 911 74053 688
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg
GF: J. Hawn, J. Guild, F. Imendörffer, HRB 16746 (AG Nürnberg)
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH v4 6/9] snic:Add low level queuing interfaces
2015-04-09 11:49 ` [PATCH v4 6/9] snic:Add low level queuing interfaces Narsimhulu Musini
@ 2015-04-09 13:23 ` Hannes Reinecke
2015-04-13 5:40 ` Narsimhulu Musini (nmusini)
0 siblings, 1 reply; 24+ messages in thread
From: Hannes Reinecke @ 2015-04-09 13:23 UTC (permalink / raw)
To: Narsimhulu Musini, JBottomley, linux-scsi, hch; +Cc: Sesidhar Baddela
Hi Narsimhulu,
please find my comment at the bottom.
On 04/09/2015 01:49 PM, Narsimhulu Musini wrote:
> These files contain low level queueing interfaces includes
> hardware queues, and management of hardware features.
>
> Signed-off-by: Narsimhulu Musini <nmusini@cisco.com>
> Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
> ---
> * v3
> - Cleaned up unused functions.
>
> * v2
> - driver supports x86-64 arch, so removed cpu_to_XX API to maintain consistency.
>
> drivers/scsi/snic/cq_desc.h | 76 ++++
> drivers/scsi/snic/cq_enet_desc.h | 38 ++
> drivers/scsi/snic/vnic_cq.c | 86 +++++
> drivers/scsi/snic/vnic_cq.h | 120 ++++++
> drivers/scsi/snic/vnic_cq_fw.h | 62 ++++
> drivers/scsi/snic/vnic_dev.c | 749 ++++++++++++++++++++++++++++++++++++++
> drivers/scsi/snic/vnic_dev.h | 140 +++++++
> drivers/scsi/snic/vnic_devcmd.h | 270 ++++++++++++++
> drivers/scsi/snic/vnic_intr.c | 59 +++
> drivers/scsi/snic/vnic_intr.h | 119 ++++++
> drivers/scsi/snic/vnic_resource.h | 68 ++++
> drivers/scsi/snic/vnic_snic.h | 54 +++
> drivers/scsi/snic/vnic_stats.h | 68 ++++
> drivers/scsi/snic/vnic_wq.c | 236 ++++++++++++
> drivers/scsi/snic/vnic_wq.h | 187 ++++++++++
> drivers/scsi/snic/wq_enet_desc.h | 91 +++++
> 16 files changed, 2423 insertions(+)
> create mode 100644 drivers/scsi/snic/cq_desc.h
> create mode 100644 drivers/scsi/snic/cq_enet_desc.h
> create mode 100644 drivers/scsi/snic/vnic_cq.c
> create mode 100644 drivers/scsi/snic/vnic_cq.h
> create mode 100644 drivers/scsi/snic/vnic_cq_fw.h
> create mode 100644 drivers/scsi/snic/vnic_dev.c
> create mode 100644 drivers/scsi/snic/vnic_dev.h
> create mode 100644 drivers/scsi/snic/vnic_devcmd.h
> create mode 100644 drivers/scsi/snic/vnic_intr.c
> create mode 100644 drivers/scsi/snic/vnic_intr.h
> create mode 100644 drivers/scsi/snic/vnic_resource.h
> create mode 100644 drivers/scsi/snic/vnic_snic.h
> create mode 100644 drivers/scsi/snic/vnic_stats.h
> create mode 100644 drivers/scsi/snic/vnic_wq.c
> create mode 100644 drivers/scsi/snic/vnic_wq.h
> create mode 100644 drivers/scsi/snic/wq_enet_desc.h
>
> diff --git a/drivers/scsi/snic/cq_desc.h b/drivers/scsi/snic/cq_desc.h
> new file mode 100644
> index 0000000..630edfa
> --- /dev/null
> +++ b/drivers/scsi/snic/cq_desc.h
> @@ -0,0 +1,76 @@
> +/*
> + * Copyright 2014 Cisco Systems, Inc. All rights reserved.
> + *
> + * This program is free software; you may redistribute it and/or modify
> + * it under the terms of the GNU General Public License as published by
> + * the Free Software Foundation; version 2 of the License.
> + *
> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
> + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
> + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
> + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
> + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
> + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
> + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
> + * SOFTWARE.
> + */
> +
> +#ifndef _CQ_DESC_H_
> +#define _CQ_DESC_H_
> +
> +/*
> + * Completion queue descriptor types
> + */
> +enum cq_desc_types {
> + CQ_DESC_TYPE_WQ_ENET = 0,
> + CQ_DESC_TYPE_DESC_COPY = 1,
> + CQ_DESC_TYPE_WQ_EXCH = 2,
> + CQ_DESC_TYPE_RQ_ENET = 3,
> + CQ_DESC_TYPE_RQ_FCP = 4,
> +};
> +
> +/* Completion queue descriptor: 16B
> + *
> + * All completion queues have this basic layout. The
> + * type_specific area is unique for each completion
> + * queue type.
> + */
> +struct cq_desc {
> + u16 completed_index;
> + u16 q_number;
> + u8 type_specific[11];
> + u8 type_color;
> +};
> +
> +#define CQ_DESC_TYPE_BITS 4
> +#define CQ_DESC_TYPE_MASK ((1 << CQ_DESC_TYPE_BITS) - 1)
> +#define CQ_DESC_COLOR_MASK 1
> +#define CQ_DESC_COLOR_SHIFT 7
> +#define CQ_DESC_Q_NUM_BITS 10
> +#define CQ_DESC_Q_NUM_MASK ((1 << CQ_DESC_Q_NUM_BITS) - 1)
> +#define CQ_DESC_COMP_NDX_BITS 12
> +#define CQ_DESC_COMP_NDX_MASK ((1 << CQ_DESC_COMP_NDX_BITS) - 1)
> +
> +static inline void cq_desc_dec(const struct cq_desc *desc_arg,
> + u8 *type, u8 *color, u16 *q_number, u16 *completed_index)
> +{
> + const struct cq_desc *desc = desc_arg;
> + const u8 type_color = desc->type_color;
> +
> + *color = (type_color >> CQ_DESC_COLOR_SHIFT) & CQ_DESC_COLOR_MASK;
> +
> + /*
> + * Make sure color bit is read from desc *before* other fields
> + * are read from desc. Hardware guarantees color bit is last
> + * bit (byte) written. Adding the rmb() prevents the compiler
> + * and/or CPU from reordering the reads which would potentially
> + * result in reading stale values.
> + */
> + rmb();
> +
> + *type = type_color & CQ_DESC_TYPE_MASK;
> + *q_number = desc->q_number & CQ_DESC_Q_NUM_MASK;
> + *completed_index = desc->completed_index & CQ_DESC_COMP_NDX_MASK;
> +}
> +
> +#endif /* _CQ_DESC_H_ */
> diff --git a/drivers/scsi/snic/cq_enet_desc.h b/drivers/scsi/snic/cq_enet_desc.h
> new file mode 100644
> index 0000000..99ecd20
> --- /dev/null
> +++ b/drivers/scsi/snic/cq_enet_desc.h
> @@ -0,0 +1,38 @@
> +/*
> + * Copyright 2014 Cisco Systems, Inc. All rights reserved.
> + *
> + * This program is free software; you may redistribute it and/or modify
> + * it under the terms of the GNU General Public License as published by
> + * the Free Software Foundation; version 2 of the License.
> + *
> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
> + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
> + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
> + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
> + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
> + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
> + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
> + * SOFTWARE.
> + */
> +
> +#ifndef _CQ_ENET_DESC_H_
> +#define _CQ_ENET_DESC_H_
> +
> +#include "cq_desc.h"
> +
> +/* Ethernet completion queue descriptor: 16B */
> +struct cq_enet_wq_desc {
> + u16 completed_index;
> + u16 q_number;
> + u8 reserved[11];
> + u8 type_color;
> +};
> +
> +static inline void cq_enet_wq_desc_dec(struct cq_enet_wq_desc *desc,
> + u8 *type, u8 *color, u16 *q_number, u16 *completed_index)
> +{
> + cq_desc_dec((struct cq_desc *)desc, type,
> + color, q_number, completed_index);
> +}
> +
> +#endif /* _CQ_ENET_DESC_H_ */
> diff --git a/drivers/scsi/snic/vnic_cq.c b/drivers/scsi/snic/vnic_cq.c
> new file mode 100644
> index 0000000..88d4537
> --- /dev/null
> +++ b/drivers/scsi/snic/vnic_cq.c
> @@ -0,0 +1,86 @@
> +/*
> + * Copyright 2014 Cisco Systems, Inc. All rights reserved.
> + *
> + * This program is free software; you may redistribute it and/or modify
> + * it under the terms of the GNU General Public License as published by
> + * the Free Software Foundation; version 2 of the License.
> + *
> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
> + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
> + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
> + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
> + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
> + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
> + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
> + * SOFTWARE.
> + */
> +
> +#include <linux/errno.h>
> +#include <linux/types.h>
> +#include <linux/pci.h>
> +#include "vnic_dev.h"
> +#include "vnic_cq.h"
> +
> +void vnic_cq_free(struct vnic_cq *cq)
> +{
> + vnic_dev_free_desc_ring(cq->vdev, &cq->ring);
> +
> + cq->ctrl = NULL;
> +}
> +
> +int vnic_cq_alloc(struct vnic_dev *vdev, struct vnic_cq *cq, unsigned int index,
> + unsigned int desc_count, unsigned int desc_size)
> +{
> + int err;
> +
> + cq->index = index;
> + cq->vdev = vdev;
> +
> + cq->ctrl = vnic_dev_get_res(vdev, RES_TYPE_CQ, index);
> + if (!cq->ctrl) {
> + pr_err("Failed to hook CQ[%d] resource\n", index);
> +
> + return -EINVAL;
> + }
> +
> + err = vnic_dev_alloc_desc_ring(vdev, &cq->ring, desc_count, desc_size);
> + if (err)
> + return err;
> +
> + return 0;
> +}
> +
> +void vnic_cq_init(struct vnic_cq *cq, unsigned int flow_control_enable,
> + unsigned int color_enable, unsigned int cq_head, unsigned int cq_tail,
> + unsigned int cq_tail_color, unsigned int interrupt_enable,
> + unsigned int cq_entry_enable, unsigned int cq_message_enable,
> + unsigned int interrupt_offset, u64 cq_message_addr)
> +{
> + u64 paddr;
> +
> + paddr = (u64)cq->ring.base_addr | VNIC_PADDR_TARGET;
> + writeq(paddr, &cq->ctrl->ring_base);
> + iowrite32(cq->ring.desc_count, &cq->ctrl->ring_size);
> + iowrite32(flow_control_enable, &cq->ctrl->flow_control_enable);
> + iowrite32(color_enable, &cq->ctrl->color_enable);
> + iowrite32(cq_head, &cq->ctrl->cq_head);
> + iowrite32(cq_tail, &cq->ctrl->cq_tail);
> + iowrite32(cq_tail_color, &cq->ctrl->cq_tail_color);
> + iowrite32(interrupt_enable, &cq->ctrl->interrupt_enable);
> + iowrite32(cq_entry_enable, &cq->ctrl->cq_entry_enable);
> + iowrite32(cq_message_enable, &cq->ctrl->cq_message_enable);
> + iowrite32(interrupt_offset, &cq->ctrl->interrupt_offset);
> + writeq(cq_message_addr, &cq->ctrl->cq_message_addr);
> +}
> +
> +void vnic_cq_clean(struct vnic_cq *cq)
> +{
> + cq->to_clean = 0;
> + cq->last_color = 0;
> +
> + iowrite32(0, &cq->ctrl->cq_head);
> + iowrite32(0, &cq->ctrl->cq_tail);
> + iowrite32(1, &cq->ctrl->cq_tail_color);
> +
> + vnic_dev_clear_desc_ring(&cq->ring);
> +}
> diff --git a/drivers/scsi/snic/vnic_cq.h b/drivers/scsi/snic/vnic_cq.h
> new file mode 100644
> index 0000000..fb2dc61
> --- /dev/null
> +++ b/drivers/scsi/snic/vnic_cq.h
> @@ -0,0 +1,120 @@
> +/*
> + * Copyright 2014 Cisco Systems, Inc. All rights reserved.
> + *
> + * This program is free software; you may redistribute it and/or modify
> + * it under the terms of the GNU General Public License as published by
> + * the Free Software Foundation; version 2 of the License.
> + *
> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
> + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
> + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
> + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
> + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
> + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
> + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
> + * SOFTWARE.
> + */
> +
> +#ifndef _VNIC_CQ_H_
> +#define _VNIC_CQ_H_
> +
> +#include "cq_desc.h"
> +#include "vnic_dev.h"
> +
> +/*
> + * These defines avoid symbol clash between fnic and enic (Cisco 10G Eth
> + * Driver) when both are built with CONFIG options =y
> + */
> +#define vnic_cq_service snic_cq_service
> +#define vnic_cq_free snic_cq_free
> +#define vnic_cq_alloc snic_cq_alloc
> +#define vnic_cq_init snic_cq_init
> +#define vnic_cq_clean snic_cq_clean
> +
> +/* Completion queue control */
> +struct vnic_cq_ctrl {
> + u64 ring_base; /* 0x00 */
> + u32 ring_size; /* 0x08 */
> + u32 pad0;
> + u32 flow_control_enable; /* 0x10 */
> + u32 pad1;
> + u32 color_enable; /* 0x18 */
> + u32 pad2;
> + u32 cq_head; /* 0x20 */
> + u32 pad3;
> + u32 cq_tail; /* 0x28 */
> + u32 pad4;
> + u32 cq_tail_color; /* 0x30 */
> + u32 pad5;
> + u32 interrupt_enable; /* 0x38 */
> + u32 pad6;
> + u32 cq_entry_enable; /* 0x40 */
> + u32 pad7;
> + u32 cq_message_enable; /* 0x48 */
> + u32 pad8;
> + u32 interrupt_offset; /* 0x50 */
> + u32 pad9;
> + u64 cq_message_addr; /* 0x58 */
> + u32 pad10;
> +};
> +
> +struct vnic_cq {
> + unsigned int index;
> + struct vnic_dev *vdev;
> + struct vnic_cq_ctrl __iomem *ctrl; /* memory-mapped */
> + struct vnic_dev_ring ring;
> + unsigned int to_clean;
> + unsigned int last_color;
> +};
> +
> +static inline unsigned int vnic_cq_service(struct vnic_cq *cq,
> + unsigned int work_to_do,
> + int (*q_service)(struct vnic_dev *vdev, struct cq_desc *cq_desc,
> + u8 type, u16 q_number, u16 completed_index, void *opaque),
> + void *opaque)
> +{
> + struct cq_desc *cq_desc;
> + unsigned int work_done = 0;
> + u16 q_number, completed_index;
> + u8 type, color;
> +
> + cq_desc = (struct cq_desc *)((u8 *)cq->ring.descs +
> + cq->ring.desc_size * cq->to_clean);
> + cq_desc_dec(cq_desc, &type, &color,
> + &q_number, &completed_index);
> +
> + while (color != cq->last_color) {
> +
> + if ((*q_service)(cq->vdev, cq_desc, type,
> + q_number, completed_index, opaque))
> + break;
> +
> + cq->to_clean++;
> + if (cq->to_clean == cq->ring.desc_count) {
> + cq->to_clean = 0;
> + cq->last_color = cq->last_color ? 0 : 1;
> + }
> +
> + cq_desc = (struct cq_desc *)((u8 *)cq->ring.descs +
> + cq->ring.desc_size * cq->to_clean);
> + cq_desc_dec(cq_desc, &type, &color,
> + &q_number, &completed_index);
> +
> + work_done++;
> + if (work_done >= work_to_do)
> + break;
> + }
> +
> + return work_done;
> +}
> +
> +void vnic_cq_free(struct vnic_cq *cq);
> +int vnic_cq_alloc(struct vnic_dev *vdev, struct vnic_cq *cq, unsigned int index,
> + unsigned int desc_count, unsigned int desc_size);
> +void vnic_cq_init(struct vnic_cq *cq, unsigned int flow_control_enable,
> + unsigned int color_enable, unsigned int cq_head, unsigned int cq_tail,
> + unsigned int cq_tail_color, unsigned int interrupt_enable,
> + unsigned int cq_entry_enable, unsigned int message_enable,
> + unsigned int interrupt_offset, u64 message_addr);
> +void vnic_cq_clean(struct vnic_cq *cq);
> +#endif /* _VNIC_CQ_H_ */
> diff --git a/drivers/scsi/snic/vnic_cq_fw.h b/drivers/scsi/snic/vnic_cq_fw.h
> new file mode 100644
> index 0000000..c2d1bbd
> --- /dev/null
> +++ b/drivers/scsi/snic/vnic_cq_fw.h
> @@ -0,0 +1,62 @@
> +/*
> + * Copyright 2014 Cisco Systems, Inc. All rights reserved.
> + *
> + * This program is free software; you may redistribute it and/or modify
> + * it under the terms of the GNU General Public License as published by
> + * the Free Software Foundation; version 2 of the License.
> + *
> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
> + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
> + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
> + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
> + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
> + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
> + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
> + * SOFTWARE.
> + */
> +
> +#ifndef _VNIC_CQ_FW_H_
> +#define _VNIC_CQ_FW_H_
> +
> +#include "snic_fwint.h"
> +
> +static inline unsigned int
> +vnic_cq_fw_service(struct vnic_cq *cq,
> + int (*q_service)(struct vnic_dev *vdev,
> + unsigned int index,
> + struct snic_fw_req *desc),
> + unsigned int work_to_do)
> +
> +{
> + struct snic_fw_req *desc;
> + unsigned int work_done = 0;
> + u8 color;
> +
> + desc = (struct snic_fw_req *)((u8 *)cq->ring.descs +
> + cq->ring.desc_size * cq->to_clean);
> + snic_color_dec(desc, &color);
> +
> + while (color != cq->last_color) {
> +
> + if ((*q_service)(cq->vdev, cq->index, desc))
> + break;
> +
> + cq->to_clean++;
> + if (cq->to_clean == cq->ring.desc_count) {
> + cq->to_clean = 0;
> + cq->last_color = cq->last_color ? 0 : 1;
> + }
> +
> + desc = (struct snic_fw_req *)((u8 *)cq->ring.descs +
> + cq->ring.desc_size * cq->to_clean);
> + snic_color_dec(desc, &color);
> +
> + work_done++;
> + if (work_done >= work_to_do)
> + break;
> + }
> +
> + return work_done;
> +}
> +
> +#endif /* _VNIC_CQ_FW_H_ */
> diff --git a/drivers/scsi/snic/vnic_dev.c b/drivers/scsi/snic/vnic_dev.c
> new file mode 100644
> index 0000000..68c63d5
> --- /dev/null
> +++ b/drivers/scsi/snic/vnic_dev.c
> @@ -0,0 +1,749 @@
> +/*
> + * Copyright 2014 Cisco Systems, Inc. All rights reserved.
> + *
> + * This program is free software; you may redistribute it and/or modify
> + * it under the terms of the GNU General Public License as published by
> + * the Free Software Foundation; version 2 of the License.
> + *
> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
> + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
> + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
> + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
> + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
> + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
> + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
> + * SOFTWARE.
> + */
> +
> +#include <linux/kernel.h>
> +#include <linux/errno.h>
> +#include <linux/types.h>
> +#include <linux/pci.h>
> +#include <linux/delay.h>
> +#include <linux/if_ether.h>
> +#include <linux/slab.h>
> +#include "vnic_resource.h"
> +#include "vnic_devcmd.h"
> +#include "vnic_dev.h"
> +#include "vnic_stats.h"
> +#include "vnic_wq.h"
> +
> +#define VNIC_DVCMD_TMO 10000 /* Devcmd Timeout value */
> +#define VNIC_NOTIFY_INTR_MASK 0x0000ffff00000000ULL
> +
> +struct devcmd2_controller {
> + struct vnic_wq_ctrl __iomem *wq_ctrl;
> + struct vnic_dev_ring results_ring;
> + struct vnic_wq wq;
> + struct vnic_devcmd2 *cmd_ring;
> + struct devcmd2_result *result;
> + u16 next_result;
> + u16 result_size;
> + int color;
> +};
> +
> +struct vnic_res {
> + void __iomem *vaddr;
> + unsigned int count;
> +};
> +
> +struct vnic_dev {
> + void *priv;
> + struct pci_dev *pdev;
> + struct vnic_res res[RES_TYPE_MAX];
> + enum vnic_dev_intr_mode intr_mode;
> + struct vnic_devcmd __iomem *devcmd;
> + struct vnic_devcmd_notify *notify;
> + struct vnic_devcmd_notify notify_copy;
> + dma_addr_t notify_pa;
> + u32 *linkstatus;
> + dma_addr_t linkstatus_pa;
> + struct vnic_stats *stats;
> + dma_addr_t stats_pa;
> + struct vnic_devcmd_fw_info *fw_info;
> + dma_addr_t fw_info_pa;
> + u64 args[VNIC_DEVCMD_NARGS];
> + struct devcmd2_controller *devcmd2;
> +
> + int (*devcmd_rtn)(struct vnic_dev *vdev, enum vnic_devcmd_cmd cmd,
> + int wait);
> +};
> +
> +#define VNIC_MAX_RES_HDR_SIZE \
> + (sizeof(struct vnic_resource_header) + \
> + sizeof(struct vnic_resource) * RES_TYPE_MAX)
> +#define VNIC_RES_STRIDE 128
> +
> +void *vnic_dev_priv(struct vnic_dev *vdev)
> +{
> + return vdev->priv;
> +}
> +
> +static int vnic_dev_discover_res(struct vnic_dev *vdev,
> + struct vnic_dev_bar *bar, unsigned int num_bars)
> +{
> + struct vnic_resource_header __iomem *rh;
> + struct vnic_resource __iomem *r;
> + u8 type;
> +
> + if (num_bars == 0)
> + return -EINVAL;
> +
> + if (bar->len < VNIC_MAX_RES_HDR_SIZE) {
> + pr_err("vNIC BAR0 res hdr length error\n");
> +
> + return -EINVAL;
> + }
> +
> + rh = bar->vaddr;
> + if (!rh) {
> + pr_err("vNIC BAR0 res hdr not mem-mapped\n");
> +
> + return -EINVAL;
> + }
> +
> + if (ioread32(&rh->magic) != VNIC_RES_MAGIC ||
> + ioread32(&rh->version) != VNIC_RES_VERSION) {
> + pr_err("vNIC BAR0 res magic/version error exp (%lx/%lx) curr (%x/%x)\n",
> + VNIC_RES_MAGIC, VNIC_RES_VERSION,
> + ioread32(&rh->magic), ioread32(&rh->version));
> +
> + return -EINVAL;
> + }
> +
> + r = (struct vnic_resource __iomem *)(rh + 1);
> +
> + while ((type = ioread8(&r->type)) != RES_TYPE_EOL) {
> +
> + u8 bar_num = ioread8(&r->bar);
> + u32 bar_offset = ioread32(&r->bar_offset);
> + u32 count = ioread32(&r->count);
> + u32 len;
> +
> + r++;
> +
> + if (bar_num >= num_bars)
> + continue;
> +
> + if (!bar[bar_num].len || !bar[bar_num].vaddr)
> + continue;
> +
> + switch (type) {
> + case RES_TYPE_WQ:
> + case RES_TYPE_RQ:
> + case RES_TYPE_CQ:
> + case RES_TYPE_INTR_CTRL:
> + /* each count is stride bytes long */
> + len = count * VNIC_RES_STRIDE;
> + if (len + bar_offset > bar->len) {
> + pr_err("vNIC BAR0 resource %d out-of-bounds, offset 0x%x + size 0x%x > bar len 0x%lx\n",
> + type, bar_offset,
> + len,
> + bar->len);
> +
> + return -EINVAL;
> + }
> + break;
> +
> + case RES_TYPE_INTR_PBA_LEGACY:
> + case RES_TYPE_DEVCMD:
> + case RES_TYPE_DEVCMD2:
> + len = count;
> + break;
> +
> + default:
> + continue;
> + }
> +
> + vdev->res[type].count = count;
> + vdev->res[type].vaddr = (char __iomem *)bar->vaddr + bar_offset;
> + }
> +
> + return 0;
> +}
> +
> +unsigned int vnic_dev_get_res_count(struct vnic_dev *vdev,
> + enum vnic_res_type type)
> +{
> + return vdev->res[type].count;
> +}
> +
> +void __iomem *vnic_dev_get_res(struct vnic_dev *vdev, enum vnic_res_type type,
> + unsigned int index)
> +{
> + if (!vdev->res[type].vaddr)
> + return NULL;
> +
> + switch (type) {
> + case RES_TYPE_WQ:
> + case RES_TYPE_RQ:
> + case RES_TYPE_CQ:
> + case RES_TYPE_INTR_CTRL:
> + return (char __iomem *)vdev->res[type].vaddr +
> + index * VNIC_RES_STRIDE;
> +
> + default:
> + return (char __iomem *)vdev->res[type].vaddr;
> + }
> +}
> +
> +unsigned int vnic_dev_desc_ring_size(struct vnic_dev_ring *ring,
> + unsigned int desc_count,
> + unsigned int desc_size)
> +{
> + /* The base address of the desc rings must be 512 byte aligned.
> + * Descriptor count is aligned to groups of 32 descriptors. A
> + * count of 0 means the maximum 4096 descriptors. Descriptor
> + * size is aligned to 16 bytes.
> + */
> +
> + unsigned int count_align = 32;
> + unsigned int desc_align = 16;
> +
> + ring->base_align = 512;
> +
> + if (desc_count == 0)
> + desc_count = 4096;
> +
> + ring->desc_count = ALIGN(desc_count, count_align);
> +
> + ring->desc_size = ALIGN(desc_size, desc_align);
> +
> + ring->size = ring->desc_count * ring->desc_size;
> + ring->size_unaligned = ring->size + ring->base_align;
> +
> + return ring->size_unaligned;
> +}
> +
> +void vnic_dev_clear_desc_ring(struct vnic_dev_ring *ring)
> +{
> + memset(ring->descs, 0, ring->size);
> +}
> +
> +int vnic_dev_alloc_desc_ring(struct vnic_dev *vdev, struct vnic_dev_ring *ring,
> + unsigned int desc_count, unsigned int desc_size)
> +{
> + vnic_dev_desc_ring_size(ring, desc_count, desc_size);
> +
> + ring->descs_unaligned = pci_alloc_consistent(vdev->pdev,
> + ring->size_unaligned,
> + &ring->base_addr_unaligned);
> +
> + if (!ring->descs_unaligned) {
> + pr_err("Failed to allocate ring (size=%d), aborting\n",
> + (int)ring->size);
> +
> + return -ENOMEM;
> + }
> +
> + ring->base_addr = ALIGN(ring->base_addr_unaligned,
> + ring->base_align);
> + ring->descs = (u8 *)ring->descs_unaligned +
> + (ring->base_addr - ring->base_addr_unaligned);
> +
> + vnic_dev_clear_desc_ring(ring);
> +
> + ring->desc_avail = ring->desc_count - 1;
> +
> + return 0;
> +}
> +
> +void vnic_dev_free_desc_ring(struct vnic_dev *vdev, struct vnic_dev_ring *ring)
> +{
> + if (ring->descs) {
> + pci_free_consistent(vdev->pdev,
> + ring->size_unaligned,
> + ring->descs_unaligned,
> + ring->base_addr_unaligned);
> + ring->descs = NULL;
> + }
> +}
> +
> +static int _vnic_dev_cmd2(struct vnic_dev *vdev, enum vnic_devcmd_cmd cmd,
> + int wait)
> +{
> + struct devcmd2_controller *dc2c = vdev->devcmd2;
> + struct devcmd2_result *result = dc2c->result + dc2c->next_result;
> + unsigned int i;
> + int delay;
> + int err;
> + u32 posted;
> + u32 new_posted;
> +
> + posted = ioread32(&dc2c->wq_ctrl->posted_index);
> +
> + if (posted == 0xFFFFFFFF) { /* check for hardware gone */
> + /* Hardware surprise removal: return error */
> + return -ENODEV;
> + }
> +
> + new_posted = (posted + 1) % DEVCMD2_RING_SIZE;
> + dc2c->cmd_ring[posted].cmd = cmd;
> + dc2c->cmd_ring[posted].flags = 0;
> +
> + if ((_CMD_FLAGS(cmd) & _CMD_FLAGS_NOWAIT))
> + dc2c->cmd_ring[posted].flags |= DEVCMD2_FNORESULT;
> +
> + if (_CMD_DIR(cmd) & _CMD_DIR_WRITE) {
> + for (i = 0; i < VNIC_DEVCMD_NARGS; i++)
> + dc2c->cmd_ring[posted].args[i] = vdev->args[i];
> + }
> + /* Adding write memory barrier prevents compiler and/or CPU
> + * reordering, thus avoiding descriptor posting before
> + * descriptor is initialized. Otherwise, hardware can read
> + * stale descriptor fields.
> + */
> + wmb();
> + iowrite32(new_posted, &dc2c->wq_ctrl->posted_index);
> +
> + if (dc2c->cmd_ring[posted].flags & DEVCMD2_FNORESULT)
> + return 0;
> +
> + for (delay = 0; delay < wait; delay++) {
> + udelay(100);
> + if (result->color == dc2c->color) {
> + dc2c->next_result++;
> + if (dc2c->next_result == dc2c->result_size) {
> + dc2c->next_result = 0;
> + dc2c->color = dc2c->color ? 0 : 1;
> + }
> + if (result->error) {
> + err = (int) result->error;
> + if (err != ERR_ECMDUNKNOWN ||
> + cmd != CMD_CAPABILITY)
> + pr_err("Error %d devcmd %d\n",
> + err, _CMD_N(cmd));
> +
> + return err;
> + }
> + if (_CMD_DIR(cmd) & _CMD_DIR_READ) {
> + /*
> + * Adding the rmb() prevents the compiler
> + * and/or CPU from reordering the reads which
> + * would potentially result in reading stale
> + * values.
> + */
> + rmb();
> + for (i = 0; i < VNIC_DEVCMD_NARGS; i++)
> + vdev->args[i] = result->results[i];
> + }
> +
> + return 0;
> + }
> + }
> +
> + pr_err("Timed out devcmd %d\n", _CMD_N(cmd));
> +
> + return -ETIMEDOUT;
> +}
> +
> +static int vnic_dev_init_devcmd2(struct vnic_dev *vdev)
> +{
> + struct devcmd2_controller *dc2c = NULL;
> + unsigned int fetch_idx;
> + int ret;
> + void __iomem *p;
> +
> + if (vdev->devcmd2)
> + return 0;
> +
> + p = vnic_dev_get_res(vdev, RES_TYPE_DEVCMD2, 0);
> + if (!p)
> + return -ENODEV;
> +
> + dc2c = kzalloc(sizeof(*dc2c), GFP_ATOMIC);
> + if (!dc2c)
> + return -ENOMEM;
> +
> + vdev->devcmd2 = dc2c;
> +
> + dc2c->color = 1;
> + dc2c->result_size = DEVCMD2_RING_SIZE;
> +
> + ret = vnic_wq_devcmd2_alloc(vdev,
> + &dc2c->wq,
> + DEVCMD2_RING_SIZE,
> + DEVCMD2_DESC_SIZE);
> + if (ret)
> + goto err_free_devcmd2;
> +
> + fetch_idx = ioread32(&dc2c->wq.ctrl->fetch_index);
> + if (fetch_idx == 0xFFFFFFFF) { /* check for hardware gone */
> + /* Hardware surprise removal: reset fetch_index */
> + fetch_idx = 0;
> + }
> +
> + /*
> + * Don't change fetch_index ever and
> + * set posted_index same as fetch_index
> + * when setting up the WQ for devcmd2.
> + */
> + vnic_wq_init_start(&dc2c->wq, 0, fetch_idx, fetch_idx, 0, 0);
> + vnic_wq_enable(&dc2c->wq);
> + ret = vnic_dev_alloc_desc_ring(vdev,
> + &dc2c->results_ring,
> + DEVCMD2_RING_SIZE,
> + DEVCMD2_DESC_SIZE);
> + if (ret)
> + goto err_free_wq;
> +
> + dc2c->result = (struct devcmd2_result *) dc2c->results_ring.descs;
> + dc2c->cmd_ring = (struct vnic_devcmd2 *) dc2c->wq.ring.descs;
> + dc2c->wq_ctrl = dc2c->wq.ctrl;
> + vdev->args[0] = (u64) dc2c->results_ring.base_addr | VNIC_PADDR_TARGET;
> + vdev->args[1] = DEVCMD2_RING_SIZE;
> +
> + ret = _vnic_dev_cmd2(vdev, CMD_INITIALIZE_DEVCMD2, VNIC_DVCMD_TMO);
> + if (ret < 0)
> + goto err_free_desc_ring;
> +
> + vdev->devcmd_rtn = &_vnic_dev_cmd2;
> + pr_info("DEVCMD2 Initialized.\n");
> +
> + return ret;
> +
> +err_free_desc_ring:
> + vnic_dev_free_desc_ring(vdev, &dc2c->results_ring);
> +
> +err_free_wq:
> + vnic_wq_disable(&dc2c->wq);
> + vnic_wq_free(&dc2c->wq);
> +
> +err_free_devcmd2:
> + kfree(dc2c);
> + vdev->devcmd2 = NULL;
> +
> + return ret;
> +} /* end of vnic_dev_init_devcmd2 */
> +
> +static void vnic_dev_deinit_devcmd2(struct vnic_dev *vdev)
> +{
> + struct devcmd2_controller *dc2c = vdev->devcmd2;
> +
> + vdev->devcmd2 = NULL;
> + vdev->devcmd_rtn = NULL;
> +
> + vnic_dev_free_desc_ring(vdev, &dc2c->results_ring);
> + vnic_wq_disable(&dc2c->wq);
> + vnic_wq_free(&dc2c->wq);
> + kfree(dc2c);
> +}
> +
> +int vnic_dev_cmd(struct vnic_dev *vdev, enum vnic_devcmd_cmd cmd,
> + u64 *a0, u64 *a1, int wait)
> +{
> + int err;
> +
> + memset(vdev->args, 0, sizeof(vdev->args));
> + vdev->args[0] = *a0;
> + vdev->args[1] = *a1;
> +
> + err = (*vdev->devcmd_rtn)(vdev, cmd, wait);
> +
> + *a0 = vdev->args[0];
> + *a1 = vdev->args[1];
> +
> + return err;
> +}
> +
> +int vnic_dev_fw_info(struct vnic_dev *vdev,
> + struct vnic_devcmd_fw_info **fw_info)
> +{
> + u64 a0, a1 = 0;
> + int wait = VNIC_DVCMD_TMO;
> + int err = 0;
> +
> + if (!vdev->fw_info) {
> + vdev->fw_info = pci_alloc_consistent(vdev->pdev,
> + sizeof(struct vnic_devcmd_fw_info),
> + &vdev->fw_info_pa);
> + if (!vdev->fw_info)
> + return -ENOMEM;
> +
> + a0 = vdev->fw_info_pa;
> +
> + /* only get fw_info once and cache it */
> + err = vnic_dev_cmd(vdev, CMD_MCPU_FW_INFO, &a0, &a1, wait);
> + }
> +
> + *fw_info = vdev->fw_info;
> +
> + return err;
> +}
> +
> +int vnic_dev_spec(struct vnic_dev *vdev, unsigned int offset, unsigned int size,
> + void *value)
> +{
> + u64 a0, a1;
> + int wait = VNIC_DVCMD_TMO;
> + int err;
> +
> + a0 = offset;
> + a1 = size;
> +
> + err = vnic_dev_cmd(vdev, CMD_DEV_SPEC, &a0, &a1, wait);
> +
> + switch (size) {
> + case 1:
> + *(u8 *)value = (u8)a0;
> + break;
> + case 2:
> + *(u16 *)value = (u16)a0;
> + break;
> + case 4:
> + *(u32 *)value = (u32)a0;
> + break;
> + case 8:
> + *(u64 *)value = a0;
> + break;
> + default:
> + BUG();
> + break;
> + }
> +
> + return err;
> +}
> +
> +int vnic_dev_stats_clear(struct vnic_dev *vdev)
> +{
> + u64 a0 = 0, a1 = 0;
> + int wait = VNIC_DVCMD_TMO;
> +
> + return vnic_dev_cmd(vdev, CMD_STATS_CLEAR, &a0, &a1, wait);
> +}
> +
> +int vnic_dev_stats_dump(struct vnic_dev *vdev, struct vnic_stats **stats)
> +{
> + u64 a0, a1;
> + int wait = VNIC_DVCMD_TMO;
> +
> + if (!vdev->stats) {
> + vdev->stats = pci_alloc_consistent(vdev->pdev,
> + sizeof(struct vnic_stats), &vdev->stats_pa);
> + if (!vdev->stats)
> + return -ENOMEM;
> + }
> +
> + *stats = vdev->stats;
> + a0 = vdev->stats_pa;
> + a1 = sizeof(struct vnic_stats);
> +
> + return vnic_dev_cmd(vdev, CMD_STATS_DUMP, &a0, &a1, wait);
> +}
> +
> +int vnic_dev_close(struct vnic_dev *vdev)
> +{
> + u64 a0 = 0, a1 = 0;
> + int wait = VNIC_DVCMD_TMO;
> +
> + return vnic_dev_cmd(vdev, CMD_CLOSE, &a0, &a1, wait);
> +}
> +
> +int vnic_dev_enable_wait(struct vnic_dev *vdev)
> +{
> + u64 a0 = 0, a1 = 0;
> + int wait = VNIC_DVCMD_TMO;
> + int err = 0;
> +
> + err = vnic_dev_cmd(vdev, CMD_ENABLE_WAIT, &a0, &a1, wait);
> + if (err == ERR_ECMDUNKNOWN)
> + return vnic_dev_cmd(vdev, CMD_ENABLE, &a0, &a1, wait);
> +
> + return err;
> +}
> +
> +int vnic_dev_disable(struct vnic_dev *vdev)
> +{
> + u64 a0 = 0, a1 = 0;
> + int wait = VNIC_DVCMD_TMO;
> +
> + return vnic_dev_cmd(vdev, CMD_DISABLE, &a0, &a1, wait);
> +}
> +
> +int vnic_dev_open(struct vnic_dev *vdev, int arg)
> +{
> + u64 a0 = (u32)arg, a1 = 0;
> + int wait = VNIC_DVCMD_TMO;
> +
> + return vnic_dev_cmd(vdev, CMD_OPEN, &a0, &a1, wait);
> +}
> +
> +int vnic_dev_open_done(struct vnic_dev *vdev, int *done)
> +{
> + u64 a0 = 0, a1 = 0;
> + int wait = VNIC_DVCMD_TMO;
> + int err;
> +
> + *done = 0;
> +
> + err = vnic_dev_cmd(vdev, CMD_OPEN_STATUS, &a0, &a1, wait);
> + if (err)
> + return err;
> +
> + *done = (a0 == 0);
> +
> + return 0;
> +}
> +
> +int vnic_dev_notify_set(struct vnic_dev *vdev, u16 intr)
> +{
> + u64 a0, a1;
> + int wait = VNIC_DVCMD_TMO;
> +
> + if (!vdev->notify) {
> + vdev->notify = pci_alloc_consistent(vdev->pdev,
> + sizeof(struct vnic_devcmd_notify),
> + &vdev->notify_pa);
> + if (!vdev->notify)
> + return -ENOMEM;
> + }
> +
> + a0 = vdev->notify_pa;
> + a1 = ((u64)intr << 32) & VNIC_NOTIFY_INTR_MASK;
> + a1 += sizeof(struct vnic_devcmd_notify);
> +
> + return vnic_dev_cmd(vdev, CMD_NOTIFY, &a0, &a1, wait);
> +}
> +
> +void vnic_dev_notify_unset(struct vnic_dev *vdev)
> +{
> + u64 a0, a1;
> + int wait = VNIC_DVCMD_TMO;
> +
> + a0 = 0; /* paddr = 0 to unset notify buffer */
> + a1 = VNIC_NOTIFY_INTR_MASK; /* intr num = -1 to unreg for intr */
> + a1 += sizeof(struct vnic_devcmd_notify);
> +
> + vnic_dev_cmd(vdev, CMD_NOTIFY, &a0, &a1, wait);
> +}
> +
> +static int vnic_dev_notify_ready(struct vnic_dev *vdev)
> +{
> + u32 *words;
> + unsigned int nwords = sizeof(struct vnic_devcmd_notify) / 4;
> + unsigned int i;
> + u32 csum;
> +
> + if (!vdev->notify)
> + return 0;
> +
> + do {
> + csum = 0;
> + memcpy(&vdev->notify_copy, vdev->notify,
> + sizeof(struct vnic_devcmd_notify));
> + words = (u32 *)&vdev->notify_copy;
> + for (i = 1; i < nwords; i++)
> + csum += words[i];
> + } while (csum != words[0]);
> +
> + return 1;
> +}
> +
> +int vnic_dev_init(struct vnic_dev *vdev, int arg)
> +{
> + u64 a0 = (u32)arg, a1 = 0;
> + int wait = VNIC_DVCMD_TMO;
> +
> + return vnic_dev_cmd(vdev, CMD_INIT, &a0, &a1, wait);
> +}
> +
> +int vnic_dev_link_status(struct vnic_dev *vdev)
> +{
> + if (vdev->linkstatus)
> + return *vdev->linkstatus;
> +
> + if (!vnic_dev_notify_ready(vdev))
> + return 0;
> +
> + return vdev->notify_copy.link_state;
> +}
> +
> +u32 vnic_dev_link_down_cnt(struct vnic_dev *vdev)
> +{
> + if (!vnic_dev_notify_ready(vdev))
> + return 0;
> +
> + return vdev->notify_copy.link_down_cnt;
> +}
> +
> +void vnic_dev_set_intr_mode(struct vnic_dev *vdev,
> + enum vnic_dev_intr_mode intr_mode)
> +{
> + vdev->intr_mode = intr_mode;
> +}
> +
> +enum vnic_dev_intr_mode vnic_dev_get_intr_mode(
> + struct vnic_dev *vdev)
> +{
> + return vdev->intr_mode;
> +}
> +
> +void vnic_dev_unregister(struct vnic_dev *vdev)
> +{
> + if (vdev) {
> + if (vdev->notify)
> + pci_free_consistent(vdev->pdev,
> + sizeof(struct vnic_devcmd_notify),
> + vdev->notify,
> + vdev->notify_pa);
> + if (vdev->linkstatus)
> + pci_free_consistent(vdev->pdev,
> + sizeof(u32),
> + vdev->linkstatus,
> + vdev->linkstatus_pa);
> + if (vdev->stats)
> + pci_free_consistent(vdev->pdev,
> + sizeof(struct vnic_stats),
> + vdev->stats, vdev->stats_pa);
> + if (vdev->fw_info)
> + pci_free_consistent(vdev->pdev,
> + sizeof(struct vnic_devcmd_fw_info),
> + vdev->fw_info, vdev->fw_info_pa);
> + if (vdev->devcmd2)
> + vnic_dev_deinit_devcmd2(vdev);
> + kfree(vdev);
> + }
> +}
> +
> +struct vnic_dev *vnic_dev_alloc_discover(struct vnic_dev *vdev,
> + void *priv,
> + struct pci_dev *pdev,
> + struct vnic_dev_bar *bar,
> + unsigned int num_bars)
> +{
> + if (!vdev) {
> + vdev = kzalloc(sizeof(struct vnic_dev), GFP_ATOMIC);
> + if (!vdev)
> + return NULL;
> + }
> +
> + vdev->priv = priv;
> + vdev->pdev = pdev;
> +
> + if (vnic_dev_discover_res(vdev, bar, num_bars))
> + goto err_out;
> +
> + return vdev;
> +
> +err_out:
> + vnic_dev_unregister(vdev);
> +
> + return NULL;
> +} /* end of vnic_dev_alloc_discover */
> +
> +/*
> + * fallback option is left to keep the interface common for other vnics.
> + */
> +int vnic_dev_cmd_init(struct vnic_dev *vdev, int fallback)
> +{
> + int err = -ENODEV;
> + void __iomem *p;
> +
> + p = vnic_dev_get_res(vdev, RES_TYPE_DEVCMD2, 0);
> + if (p)
> + err = vnic_dev_init_devcmd2(vdev);
> + else
> + pr_err("DEVCMD2 resource not found.\n");
> +
> + return err;
> +} /* end of vnic_dev_cmd_init */
> diff --git a/drivers/scsi/snic/vnic_dev.h b/drivers/scsi/snic/vnic_dev.h
> new file mode 100644
> index 0000000..19f3e76
> --- /dev/null
> +++ b/drivers/scsi/snic/vnic_dev.h
> @@ -0,0 +1,140 @@
> +/*
> + * Copyright 2014 Cisco Systems, Inc. All rights reserved.
> + *
> + * This program is free software; you may redistribute it and/or modify
> + * it under the terms of the GNU General Public License as published by
> + * the Free Software Foundation; version 2 of the License.
> + *
> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
> + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
> + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
> + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
> + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
> + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
> + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
> + * SOFTWARE.
> + */
> +
> +#ifndef _VNIC_DEV_H_
> +#define _VNIC_DEV_H_
> +
> +#include "vnic_resource.h"
> +#include "vnic_devcmd.h"
> +
> +/*
> + * These defines avoid symbol clash between fnic and enic (Cisco 10G Eth
> + * Driver) when both are built with CONFIG options =y
> + */
So why did you choose the same name then?
Seeing that you rename them anyway, please use the correct names and do
away with this 'define'.
Cheers,
Hannes
--
Dr. Hannes Reinecke zSeries & Storage
hare@suse.de +49 911 74053 688
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg
GF: J. Hawn, J. Guild, F. Imendörffer, HRB 16746 (AG Nürnberg)
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH v4 7/9] snic:Add sysfs entries to list stats and trace data
2015-04-09 11:49 ` [PATCH v4 7/9] snic:Add sysfs entries to list stats and trace data Narsimhulu Musini
@ 2015-04-09 13:23 ` Hannes Reinecke
0 siblings, 0 replies; 24+ messages in thread
From: Hannes Reinecke @ 2015-04-09 13:23 UTC (permalink / raw)
To: Narsimhulu Musini, JBottomley, linux-scsi, hch; +Cc: Sesidhar Baddela
On 04/09/2015 01:49 PM, Narsimhulu Musini wrote:
> snic_stats.h contains stats structre definitions for various events
> in snic driver.
>
> snic_debugfs.c contains setup and cleanup of sysfs entries for listing stats,
> resetting stats, enabling/disabling trace, and listing trace data.
>
> Signed-off-by: Narsimhulu Musini <nmusini@cisco.com>
> Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
> ---
Reviewed-by: Hannes Reinecke <hare@suse.de>
Cheers,
Hannes
--
Dr. Hannes Reinecke zSeries & Storage
hare@suse.de +49 911 74053 688
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg
GF: J. Hawn, J. Guild, F. Imendörffer, HRB 16746 (AG Nürnberg)
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH v4 8/9] snic:Add event tracing to capture IO events.
2015-04-09 11:49 ` [PATCH v4 8/9] snic:Add event tracing to capture IO events Narsimhulu Musini
@ 2015-04-09 13:24 ` Hannes Reinecke
0 siblings, 0 replies; 24+ messages in thread
From: Hannes Reinecke @ 2015-04-09 13:24 UTC (permalink / raw)
To: Narsimhulu Musini, JBottomley, linux-scsi, hch; +Cc: Sesidhar Baddela
On 04/09/2015 01:49 PM, Narsimhulu Musini wrote:
> snic_trc.h contains global trace structure definitions for snic driver
>
> snic_trc.c adds tracing functionality to capture various IO events.
> It maintains global trace buffer to maintain recent history of IO events.
> It helps to understand the sequence of events prior to particular IO event,
> or hung, panic, etc,.
>
> Signed-off-by: Narsimhulu Musini <nmusini@cisco.com>
> Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
> ---
Reviewed-by: Hannes Reinecke <hare@suse.de>
Cheers,
Hannes
--
Dr. Hannes Reinecke zSeries & Storage
hare@suse.de +49 911 74053 688
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg
GF: J. Hawn, J. Guild, F. Imendörffer, HRB 16746 (AG Nürnberg)
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH v4 9/9] snic:Add Makefile, patch Kconfig, MAINTAINERS
2015-04-09 11:49 ` [PATCH v4 9/9] snic:Add Makefile, patch Kconfig, MAINTAINERS Narsimhulu Musini
@ 2015-04-09 13:25 ` Hannes Reinecke
0 siblings, 0 replies; 24+ messages in thread
From: Hannes Reinecke @ 2015-04-09 13:25 UTC (permalink / raw)
To: Narsimhulu Musini, JBottomley, linux-scsi, hch; +Cc: Sesidhar Baddela
On 04/09/2015 01:49 PM, Narsimhulu Musini wrote:
> Kconfig for kbuild
> Makefile to build snic module
>
> Updated MAINTAINERS file
>
> Signed-off-by: Narsimhulu Musini <nmusini@cisco.com>
> Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
> ---
Reviewed-by: Hannes Reinecke <hare@suse.de>
Cheers,
Hannes
--
Dr. Hannes Reinecke zSeries & Storage
hare@suse.de +49 911 74053 688
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg
GF: J. Hawn, J. Guild, F. Imendörffer, HRB 16746 (AG Nürnberg)
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH v4 4/9] snic:Add snic target discovery
2015-04-09 12:59 ` Hannes Reinecke
@ 2015-04-13 5:25 ` Narsimhulu Musini (nmusini)
2015-04-13 14:29 ` Hannes Reinecke
0 siblings, 1 reply; 24+ messages in thread
From: Narsimhulu Musini (nmusini) @ 2015-04-13 5:25 UTC (permalink / raw)
To: Hannes Reinecke, JBottomley, linux-scsi, hch; +Cc: Sesidhar Baddela (sebaddel)
Hi Hannes,
Thank you for reviewing patches. Please find responses inline.
On 09/04/15 6:29 pm, "Hannes Reinecke" <hare@suse.de> wrote:
>Hi Narsimhulu,
>
>please find some comments inline.
>
>On 04/09/2015 01:49 PM, Narsimhulu Musini wrote:
>> snic_disc.h contains snic target structure definition.
>>
>> snic_disc.c contains target discovery, setup, lookup, and cleanup
>>
>> snic_ctl.c contains retrieval of snic capabilities includes
>> max ios, size, SGs per request, and max concurrent requests.
>>
>> Signed-off-by: Narsimhulu Musini <nmusini@cisco.com>
>> Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
>> ---
>> * v3
>> - Cleaned up redundant comment.
>>
>> drivers/scsi/snic/snic_ctl.c | 276 +++++++++++++++++++
>> drivers/scsi/snic/snic_disc.c | 602
>>++++++++++++++++++++++++++++++++++++++++++
>> drivers/scsi/snic/snic_disc.h | 124 +++++++++
>> 3 files changed, 1002 insertions(+)
>> create mode 100644 drivers/scsi/snic/snic_ctl.c
>> create mode 100644 drivers/scsi/snic/snic_disc.c
>> create mode 100644 drivers/scsi/snic/snic_disc.h
>>
>> diff --git a/drivers/scsi/snic/snic_ctl.c b/drivers/scsi/snic/snic_ctl.c
>> new file mode 100644
>> index 0000000..4bc55be
>> --- /dev/null
>> +++ b/drivers/scsi/snic/snic_ctl.c
>> @@ -0,0 +1,276 @@
>> +/*
>> + * Copyright 2014 Cisco Systems, Inc. All rights reserved.
>> + *
>> + * This program is free software; you may redistribute it and/or modify
>> + * it under the terms of the GNU General Public License as published by
>> + * the Free Software Foundation; version 2 of the License.
>> + *
>> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
>> + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
>> + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
>> + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
>> + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
>> + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
>> + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
>> + * SOFTWARE.
>> + */
>> +
>> +#include <linux/errno.h>
>> +#include <linux/pci.h>
>> +#include <linux/slab.h>
>> +
>> +#include <linux/interrupt.h>
>> +#include <linux/workqueue.h>
>> +#include <linux/spinlock.h>
>> +#include <linux/mempool.h>
>> +#include <scsi/scsi_tcq.h>
>> +#include <linux/ctype.h>
>> +
>> +#include "snic_io.h"
>> +#include "snic.h"
>> +#include "cq_enet_desc.h"
>> +#include "snic_fwint.h"
>> +
>> +/*
>> + * snic_handle_link : Handles link flaps.
>> + */
>> +void
>> +snic_handle_link(struct work_struct *work)
>> +{
>> + struct snic *snic = container_of(work, struct snic, link_work);
>> +
>> + if (snic->config.xpt_type != SNIC_DAS) {
>> + SNIC_HOST_INFO(snic->shost, "Link Event Received.\n");
>> + SNIC_ASSERT_NOT_IMPL(1);
>> +
>> + return;
>> + }
>> +
>> + snic->link_status = vnic_dev_link_status(snic->vdev);
>> + snic->link_down_cnt = vnic_dev_link_down_cnt(snic->vdev);
>> + SNIC_HOST_INFO(snic->shost, "Link Event: Link %s.\n",
>> + ((snic->link_status) ? "Up" : "Down"));
>> +}
>> +
>> +
>> +/*
>> + * snic_ver_enc : Encodes version str to int
>> + * version string is similar to netmask string
>> + */
>> +static int
>> +snic_ver_enc(const char *s)
>> +{
>> + int v[4] = {0};
>> + int i = 0, x = 0;
>> + char c;
>> + const char *p = s;
>> +
>> + /* validate version string */
>> + if ((strlen(s) > 15) || (strlen(s) < 7))
>> + goto end;
>> +
>> + while ((c = *p++)) {
>> + if (c == '.') {
>> + i++;
>> + continue;
>> + }
>> +
>> + if (i > 4 || !isdigit(c))
>> + goto end;
>> +
>> + v[i] = v[i] * 10 + (c - '0');
>> + }
>> +
>> + /* validate sub version numbers */
>> + for (i = 3; i >= 0; i--)
>> + if (v[i] > 0xff)
>> + goto end;
>> +
>> + x |= (v[0] << 24) | v[1] << 16 | v[2] << 8 | v[3];
>> +
>> +end:
>> + if (x == 0) {
>> + SNIC_ERR("Invalid version string [%s].\n", s);
>> +
>> + return -1;
>> + }
>> +
>> + return x;
>> +} /* end of snic_ver_enc */
>> +
>> +/*
>> + * snic_qeueue_exch_ver_req :
>> + *
>> + * Queues Exchange Version Request, to communicate host information
>> + * in return, it gets firmware version details
>> + */
>> +int
>> +snic_queue_exch_ver_req(struct snic *snic)
>> +{
>> + struct snic_req_info *rqi = NULL;
>> + struct snic_host_req *req = NULL;
>> + int ret = 0;
>> +
>> + SNIC_HOST_INFO(snic->shost, "Exch Ver Req Preparing...\n");
>> +
>> + rqi = snic_req_init(snic, 0);
>> + if (!rqi) {
>> + SNIC_HOST_ERR(snic->shost,
>> + "Queuing Exch Ver Req failed, err = %d\n",
>> + ret);
>> +
>> + ret = -ENOMEM;
>> + goto error;
>> + }
>> +
>> + req = rqi_to_req(rqi);
>> +
>> + /* Initialize snic_host_req */
>> + snic_io_hdr_enc(&req->hdr, SNIC_REQ_EXCH_VER, 0, SCSI_NO_TAG,
>> + snic->config.hid, 0, (u64)rqi);
>> + req->u.exch_ver.drvr_ver = snic_ver_enc(SNIC_DRV_VERSION);
>> + req->u.exch_ver.os_type = SNIC_OS_LINUX;
>> +
>> + snic_handle_untagged_req(snic, rqi);
>> +
>> + ret = snic_queue_wq_desc(snic, req, sizeof(*req));
>> + if (ret) {
>> + snic_release_untagged_req(snic, rqi);
>> + SNIC_HOST_ERR(snic->shost,
>> + "Queuing Exch Ver Req failed, err = %d\n",
>> + ret);
>> + goto error;
>> + }
>> +
>> + SNIC_HOST_INFO(snic->shost, "Exch Ver Req is issued. ret = %d\n",
>>ret);
>> +
>> +error:
>> + return ret;
>> +} /* end of snic_queue_exch_ver_req */
>> +
>> +/*
>> + * snic_io_exch_ver_cmpl_handler
>> + */
>> +int
>> +snic_io_exch_ver_cmpl_handler(struct snic *snic, struct snic_fw_req
>>*fwreq)
>> +{
>> + struct snic_req_info *rqi = NULL;
>> + u8 typ, hdr_stat;
>> + u32 cmnd_id, hid, max_sgs;
>> + u64 ctx = 0;
>> + unsigned long flags;
>> + int ret = 0;
>> +
>> + SNIC_HOST_INFO(snic->shost, "Exch Ver Compl Received.\n");
>> + snic_io_hdr_dec(&fwreq->hdr, &typ, &hdr_stat, &cmnd_id, &hid, &ctx);
>> + SNIC_BUG_ON(snic->config.hid != hid);
>> + rqi = (struct snic_req_info *) ctx;
>> +
>> + if (hdr_stat) {
>> + SNIC_HOST_ERR(snic->shost,
>> + "Exch Ver Completed w/ err status %d\n",
>> + hdr_stat);
>> +
>> + goto exch_cmpl_end;
>> + }
>> +
>> + spin_lock_irqsave(&snic->snic_lock, flags);
>> + snic->fwinfo.fw_ver = fwreq->u.exch_ver_cmpl.version;
>> + snic->fwinfo.hid = fwreq->u.exch_ver_cmpl.hid;
>> + snic->fwinfo.max_concur_ios = fwreq->u.exch_ver_cmpl.max_concur_ios;
>> + snic->fwinfo.max_sgs_per_cmd = fwreq->u.exch_ver_cmpl.max_sgs_per_cmd;
>> + snic->fwinfo.max_io_sz = fwreq->u.exch_ver_cmpl.max_io_sz;
>> + snic->fwinfo.max_tgts = fwreq->u.exch_ver_cmpl.max_tgts;
>> + snic->fwinfo.io_tmo = fwreq->u.exch_ver_cmpl.io_timeout;
>> +
>> + SNIC_HOST_INFO(snic->shost,
>> + "vers %u hid %u max_concur_ios %u max_sgs_per_cmd %u
>>max_io_sz %u max_tgts %u fw tmo %u\n",
>> + fwreq->u.exch_ver_cmpl.version,
>> + fwreq->u.exch_ver_cmpl.hid,
>> + fwreq->u.exch_ver_cmpl.max_concur_ios,
>> + fwreq->u.exch_ver_cmpl.max_sgs_per_cmd,
>> + fwreq->u.exch_ver_cmpl.max_io_sz,
>> + fwreq->u.exch_ver_cmpl.max_tgts,
>> + fwreq->u.exch_ver_cmpl.io_timeout);
>> +
>> + SNIC_HOST_INFO(snic->shost,
>> + "HBA Capabilities = 0x%x\n",
>> + fwreq->u.exch_ver_cmpl.hba_cap);
>> +
>> + /* Updating SGList size */
>> + max_sgs = snic->fwinfo.max_sgs_per_cmd;
>> + if (max_sgs && max_sgs < SNIC_MAX_SG_DESC_CNT) {
>> + snic->shost->sg_tablesize = max_sgs;
>> + SNIC_HOST_INFO(snic->shost, "Max SGs set to %d\n",
>> + snic->shost->sg_tablesize);
>> + } else if (max_sgs > snic->shost->sg_tablesize) {
>> + SNIC_HOST_INFO(snic->shost,
>> + "Target type %d Supports Larger Max SGList %d than driver's
>>Max SG List %d.\n",
>> + snic->config.xpt_type, max_sgs,
>> + snic->shost->sg_tablesize);
>> + }
>> +
>> + if (snic->shost->can_queue > snic->fwinfo.max_concur_ios)
>> + snic->shost->can_queue = snic->fwinfo.max_concur_ios;
>> +
>> + snic->shost->max_sectors = snic->fwinfo.max_io_sz >> 9;
>> + if (snic->fwinfo.wait)
>> + complete(snic->fwinfo.wait);
>> +
>> + spin_unlock_irqrestore(&snic->snic_lock, flags);
>> +
>> +exch_cmpl_end:
>> + snic_release_untagged_req(snic, rqi);
>> +
>> + SNIC_HOST_INFO(snic->shost, "Exch_cmpl Done, hdr_stat %d.\n",
>>hdr_stat);
>> +
>> + return ret;
>> +} /* end of snic_io_exch_ver_cmpl_handler */
>> +
>> +/*
>> + * snic_get_conf
>> + *
>> + * Synchronous call, and Retrieves snic params.
>> + */
>> +int
>> +snic_get_conf(struct snic *snic)
>> +{
>> + DECLARE_COMPLETION_ONSTACK(wait);
>> + unsigned long flags;
>> + int ret;
>> + int nr_retries = 3;
>> +
>> + SNIC_HOST_INFO(snic->shost, "Retrieving snic params.\n");
>> + spin_lock_irqsave(&snic->snic_lock, flags);
>> + memset(&snic->fwinfo, 0, sizeof(snic->fwinfo));
>> + snic->fwinfo.wait = &wait;
>> + spin_unlock_irqrestore(&snic->snic_lock, flags);
>> +
>> + /* Additional delay to handle HW Resource initialization. */
>> + msleep(50);
>> +
>> + /*
>> + * Exch ver req can be ignored by FW, if HW Resource initialization
>> + * is in progress, Hence retry.
>> + */
>> + do {
>> + ret = snic_queue_exch_ver_req(snic);
>> + if (ret)
>> + return ret;
>> +
>> + wait_for_completion_timeout(&wait, msecs_to_jiffies(2000));
>> + spin_lock_irqsave(&snic->snic_lock, flags);
>> + ret = (snic->fwinfo.fw_ver != 0) ? 0 : -ETIMEDOUT;
>> + if (ret)
>> + SNIC_HOST_ERR(snic->shost,
>> + "Failed to retrieve snic params,\n");
>> +
>> + /* Unset fwinfo.wait, on success or on last retry */
>> + if (ret == 0 || nr_retries == 1)
>> + snic->fwinfo.wait = NULL;
>> +
>> + spin_unlock_irqrestore(&snic->snic_lock, flags);
>> + } while (ret && --nr_retries);
>> +
>> + return ret;
>> +} /* end of snic_get_info */
>> diff --git a/drivers/scsi/snic/snic_disc.c
>>b/drivers/scsi/snic/snic_disc.c
>> new file mode 100644
>> index 0000000..b6b6cca
>> --- /dev/null
>> +++ b/drivers/scsi/snic/snic_disc.c
>> @@ -0,0 +1,602 @@
>> +/*
>> + * Copyright 2014 Cisco Systems, Inc. All rights reserved.
>> + *
>> + * This program is free software; you may redistribute it and/or modify
>> + * it under the terms of the GNU General Public License as published by
>> + * the Free Software Foundation; version 2 of the License.
>> + *
>> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
>> + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
>> + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
>> + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
>> + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
>> + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
>> + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
>> + * SOFTWARE.
>> + */
>> +
>> +#include <linux/errno.h>
>> +#include <linux/mempool.h>
>> +
>> +#include <scsi/scsi_tcq.h>
>> +
>> +#include "snic_disc.h"
>> +#include "snic.h"
>> +#include "snic_io.h"
>> +
>> +
>> +/* snic target types */
>> +static const char * const snic_tgt_type_str[] = {
>> + [SNIC_TGT_DAS] = "DAS",
>> + [SNIC_TGT_SAN] = "SAN",
>> +};
>> +
>> +static inline const char *
>> +snic_tgt_type_to_str(int typ)
>> +{
>> + return ((typ > SNIC_TGT_NONE && typ <= SNIC_TGT_SAN) ?
>> + snic_tgt_type_str[typ] : "Unknown");
>> +}
>> +
>> +static const char * const snic_tgt_state_str[] = {
>> + [SNIC_TGT_STAT_INIT] = "INIT",
>> + [SNIC_TGT_STAT_ONLINE] = "ONLINE",
>> + [SNIC_TGT_STAT_OFFLINE] = "OFFLINE",
>> + [SNIC_TGT_STAT_DEL] = "DELETION IN PROGRESS",
>> +};
>> +
>> +const char *
>> +snic_tgt_state_to_str(int state)
>> +{
>> + return ((state >= SNIC_TGT_STAT_INIT && state <= SNIC_TGT_STAT_DEL) ?
>> + snic_tgt_state_str[state] : "UNKNOWN");
>> +}
>> +
>> +/*
>> + * Initiate report_tgt req desc
>> + */
>> +static void
>> +snic_report_tgt_init(struct snic_host_req *req, u32 hid, u8 *buf, u32
>>len,
>> + u64 rsp_buf_pa, u64 ctx)
>> +{
>> + struct snic_sg_desc *sgd = NULL;
>> +
>> +
>> + snic_io_hdr_enc(&req->hdr, SNIC_REQ_REPORT_TGTS, 0, SCSI_NO_TAG, hid,
>> + 1, ctx);
>> +
>> + req->u.rpt_tgts.sg_cnt = 1;
>> + sgd = req_to_sgl(req);
>> + sgd[0].addr = cpu_to_le64(rsp_buf_pa);
>> + sgd[0].len = cpu_to_le32(len);
>> + sgd[0]._resvd = 0;
>> + req->u.rpt_tgts.sg_addr = (u64) sgd;
>> +}
>> +
>> +/*
>> + * snic_queue_report_tgt_req: Queues report target request.
>> + */
>> +static int
>> +snic_queue_report_tgt_req(struct snic *snic)
>> +{
>> + struct snic_req_info *rqi = NULL;
>> + u32 ntgts, buf_len = 0;
>> + u8 *buf = NULL;
>> + dma_addr_t pa = 0;
>> + int ret = 0;
>> +
>> + rqi = snic_req_init(snic, 1);
>> + if (!rqi) {
>> + ret = -ENOMEM;
>> + goto error;
>> + }
>> +
>> + if (snic->fwinfo.max_tgts)
>> + ntgts = min_t(u32, snic->fwinfo.max_tgts, snic->shost->max_id);
>> + else
>> + ntgts = snic->shost->max_id;
>> +
>> + /* Allocate Response Buffer */
>> + SNIC_BUG_ON(ntgts == 0);
>> + buf_len = ntgts * sizeof(struct snic_tgt_id) + SNIC_SG_DESC_ALIGN;
>> +
>> + buf = kzalloc(buf_len, GFP_KERNEL|GFP_DMA);
>> + if (!buf) {
>> + snic_req_free(snic, rqi);
>> + SNIC_HOST_ERR(snic->shost, "Resp Buf Alloc Failed.\n");
>> +
>> + ret = -ENOMEM;
>> + goto error;
>> + }
>> +
>> + SNIC_BUG_ON((((unsigned long)buf) % SNIC_SG_DESC_ALIGN) != 0);
>> +
>> + pa = pci_map_single(snic->pdev, buf, buf_len, PCI_DMA_FROMDEVICE);
>> + if (pci_dma_mapping_error(snic->pdev, pa)) {
>> + kfree(buf);
>> + snic_req_free(snic, rqi);
>> + SNIC_HOST_ERR(snic->shost,
>> + "Rpt-tgt rspbuf %p: PCI DMA Mapping Failed\n",
>> + buf);
>> + ret = -EINVAL;
>> +
>> + goto error;
>> + }
>> +
>> +
>> + SNIC_BUG_ON(pa == 0);
>> + rqi->sge_va = (u64) buf;
>> +
>> + snic_report_tgt_init(rqi->req,
>> + snic->config.hid,
>> + buf,
>> + buf_len,
>> + (u64)pa,
>> + (u64)rqi);
>> +
>> + snic_handle_untagged_req(snic, rqi);
>> +
>> + ret = snic_queue_wq_desc(snic, rqi->req, rqi->req_len);
>> + if (ret) {
>> + pci_unmap_single(snic->pdev, pa, buf_len, PCI_DMA_FROMDEVICE);
>> + kfree(buf);
>> + rqi->sge_va = 0;
>> + snic_release_untagged_req(snic, rqi);
>> + SNIC_HOST_ERR(snic->shost, "Queuing Report Tgts Failed.\n");
>> +
>> + goto error;
>> + }
>> +
>> + SNIC_DISC_DBG(snic->shost, "Report Targets Issued.\n");
>> +
>> + return ret;
>> +
>> +error:
>> + SNIC_HOST_ERR(snic->shost,
>> + "Queuing Report Targets Failed, err = %d\n",
>> + ret);
>> + return ret;
>> +} /* end of snic_queue_report_tgt_req */
>> +
>> +/* call into SML */
>> +static void
>> +snic_scsi_scan_tgt(struct work_struct *work)
>> +{
>> + struct snic_tgt *tgt = container_of(work, struct snic_tgt, scan_work);
>> + struct Scsi_Host *shost = dev_to_shost(&tgt->dev);
>> + unsigned long flags;
>> +
>> + SNIC_HOST_INFO(shost, "Scanning Target id 0x%x\n", tgt->id);
>> + scsi_scan_target(&tgt->dev,
>> + tgt->channel,
>> + tgt->scsi_tgt_id,
>> + SCAN_WILD_CARD,
>> + 1);
>> +
>> + spin_lock_irqsave(shost->host_lock, flags);
>> + tgt->flags &= ~SNIC_TGT_SCAN_PENDING;
>> + spin_unlock_irqrestore(shost->host_lock, flags);
>> +} /* end of snic_scsi_scan_tgt */
>> +
>> +/*
>> + * snic_tgt_lookup :
>> + */
>> +static struct snic_tgt *
>> +snic_tgt_lookup(struct snic *snic, struct snic_tgt_id *tgtid)
>> +{
>> + struct list_head *cur, *nxt;
>> + struct snic_tgt *tgt = NULL;
>> +
>> + list_for_each_safe(cur, nxt, &snic->disc.tgt_list) {
>> + tgt = list_entry(cur, struct snic_tgt, list);
>> + if (tgt->id == tgtid->tgt_id)
>> + return tgt;
>> + tgt = NULL;
>> + }
>> +
>> + return tgt;
>> +} /* end of snic_tgt_lookup */
>> +
>> +/*
>> + * snic_tgt_dev_release : Called on dropping last ref for snic_tgt
>>object
>> + */
>> +void
>> +snic_tgt_dev_release(struct device *dev)
>> +{
>> + struct snic_tgt *tgt = dev_to_tgt(dev);
>> +
>> + SNIC_HOST_INFO(snic_tgt_to_shost(tgt),
>> + "Target Device ID %d (%s) Permanently Deleted.\n",
>> + tgt->id,
>> + dev_name(dev));
>> +
>> + SNIC_BUG_ON(!list_empty(&tgt->list));
>> + kfree(tgt);
>> +}
>> +
>> +/*
>> + * snic_tgt_del : work function to delete snic_tgt
>> + */
>> +static void
>> +snic_tgt_del(struct work_struct *work)
>> +{
>> + struct snic_tgt *tgt = container_of(work, struct snic_tgt, del_work);
>> + struct Scsi_Host *shost = snic_tgt_to_shost(tgt);
>> +
>> + if (tgt->flags & SNIC_TGT_SCAN_PENDING)
>> + scsi_flush_work(shost);
>> +
>> + /* Block IOs on child devices, stops new IOs */
>> + scsi_target_block(&tgt->dev);
>> +
>> + /* Cleanup IOs */
>> + snic_tgt_scsi_abort_io(tgt);
>> +
>> + /* Unblock IOs now, to flush if there are any. */
>> + scsi_target_unblock(&tgt->dev, SDEV_TRANSPORT_OFFLINE);
>> +
>> + /* Delete SCSI Target and sdevs */
>> + scsi_remove_target(&tgt->dev); /* ?? */
>> + device_del(&tgt->dev);
>> + put_device(&tgt->dev);
>> +} /* end of snic_tgt_del */
>> +
>> +/* snic_tgt_create: checks for existence of snic_tgt, if it doesn't
>> + * it creates one.
>> + */
>> +static struct snic_tgt *
>> +snic_tgt_create(struct snic *snic, struct snic_tgt_id *tgtid)
>> +{
>> + struct snic_tgt *tgt = NULL;
>> + unsigned long flags;
>> + int ret;
>> +
>> + tgt = snic_tgt_lookup(snic, tgtid);
>> + if (tgt) {
>> + /* update the information if required */
>> + return tgt;
>> + }
>> +
>> + tgt = kzalloc(sizeof(*tgt), GFP_KERNEL);
>> + if (!tgt) {
>> + SNIC_HOST_ERR(snic->shost, "Failure to allocate snic_tgt.\n");
>> + ret = -ENOMEM;
>> +
>> + return tgt;
>> + }
>> +
>> + INIT_LIST_HEAD(&tgt->list);
>> + tgt->id = tgtid->tgt_id;
>> + tgt->channel = 0;
>> +
>> + SNIC_BUG_ON(tgtid->tgt_type > SNIC_TGT_SAN);
>> + tgt->tdata.typ = tgtid->tgt_type;
>> +
>> + /*
>> + * Plugging into SML Device Tree
>> + */
>> + tgt->tdata.disc_id = 0;
>> + tgt->state = SNIC_TGT_STAT_INIT;
>> + device_initialize(&tgt->dev);
>> + tgt->dev.parent = get_device(&snic->shost->shost_gendev);
>> + tgt->dev.release = snic_tgt_dev_release;
>Why do you use your own scsi target instantiation here?
>If it's equivalent to the scsi target than you should rather use the
>'scsi_target' structure here and attach driver-specific information
>to either hostdata or starget_data.
I got your idea, we followed an approach similar to fc (fc_rport_create()
in scsi_transport_fc.c).
Both are valid approaches. We opted for current approach, and gone through
good amount of testing.
I would like to keep the way it is. Please share your thoughts.
>
>Cheers,
>
>Hannes
>--
>Dr. Hannes Reinecke zSeries & Storage
>hare@suse.de +49 911 74053 688
>SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg
>GF: J. Hawn, J. Guild, F. Imendörffer, HRB 16746 (AG Nürnberg)
Thanks
Narsimhulu
>
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH v4 5/9] snic:add SCSI handling, AEN, and fwreset handling
2015-04-09 13:16 ` Hannes Reinecke
@ 2015-04-13 5:37 ` Narsimhulu Musini (nmusini)
0 siblings, 0 replies; 24+ messages in thread
From: Narsimhulu Musini (nmusini) @ 2015-04-13 5:37 UTC (permalink / raw)
To: Hannes Reinecke, JBottomley, linux-scsi, hch; +Cc: Sesidhar Baddela (sebaddel)
Hi Hannes,
Thank you for reviewing patches. Please find responses inline.
I will incorporate the comments and suggestions in next patch submittal.
On 09/04/15 6:46 pm, "Hannes Reinecke" <hare@suse.de> wrote:
>Hi Narsimhulu,
>
>please find some comments inline.
>
>On 04/09/2015 01:49 PM, Narsimhulu Musini wrote:
>> snic_scsi.c contains scsi handling, includes queuing io, abort, lun
>>reset,
>> and host reset. Also it handles asynchronous event notifications from
>>FW.
>>
>> Signed-off-by: Narsimhulu Musini <nmusini@cisco.com>
>> Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
>> ---
>> * v2
>> - Changed queuecommand to lock-free version.
>> - Converted custom error codes to standard error codes.
>>
>> drivers/scsi/snic/snic_scsi.c | 2638
>>+++++++++++++++++++++++++++++++++++++++++
>> 1 file changed, 2638 insertions(+)
>> create mode 100644 drivers/scsi/snic/snic_scsi.c
>>
>> diff --git a/drivers/scsi/snic/snic_scsi.c
>>b/drivers/scsi/snic/snic_scsi.c
>> new file mode 100644
>> index 0000000..5d877f0
>> --- /dev/null
>> +++ b/drivers/scsi/snic/snic_scsi.c
>> @@ -0,0 +1,2638 @@
>> +/*
>> + * Copyright 2014 Cisco Systems, Inc. All rights reserved.
>> + *
>> + * This program is free software; you may redistribute it and/or modify
>> + * it under the terms of the GNU General Public License as published by
>> + * the Free Software Foundation; version 2 of the License.
>> + *
>> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
>> + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
>> + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
>> + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
>> + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
>> + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
>> + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
>> + * SOFTWARE.
>> + */
>> +
>> +#include <linux/mempool.h>
>> +#include <linux/errno.h>
>> +#include <linux/init.h>
>> +#include <linux/workqueue.h>
>> +#include <linux/pci.h
>> +#include <linux/spinlock.h>
>> +#include <linux/delay.h>
>> +#include <linux/gfp.h>
>> +#include <scsi/scsi.h>
>> +#include <scsi/scsi_host.h>
>> +#include <scsi/scsi_device.h>
>> +#include <scsi/scsi_cmnd.h>
>> +#include <scsi/scsi_tcq.h>
>> +#include <scsi/scsi_dbg.h>
>> +
>> +#include "snic_io.h"
>> +#include "snic.h"
>> +
>> +#define snic_cmd_tag(sc) (((struct scsi_cmnd *) sc)->request->tag)
>> +
>> +const char *snic_state_str[] = {
>> + [SNIC_INIT] = "SNIC_INIT",
>> + [SNIC_ERROR] = "SNIC_ERROR",
>> + [SNIC_ONLINE] = "SNIC_ONLINE",
>> + [SNIC_OFFLINE] = "SNIC_OFFLINE",
>> + [SNIC_FWRESET] = "SNIC_FWRESET",
>> +};
>> +
>> +static const char * const snic_req_state_str[] = {
>> + [SNIC_IOREQ_NOT_INITED] = "SNIC_IOREQ_NOT_INITED",
>> + [SNIC_IOREQ_PENDING] = "SNIC_IOREQ_PENDING",
>> + [SNIC_IOREQ_ABTS_PENDING] = "SNIC_IOREQ_ABTS_PENDING",
>> + [SNIC_IOREQ_ABTS_COMPLETE] = "SNIC_IOREQ_ABTS_COMPELTE",
>> + [SNIC_IOREQ_LR_PENDING] = "SNIC_IOREQ_LR_PENDING",
>> + [SNIC_IOREQ_LR_COMPLETE] = "SNIC_IOREQ_LR_COMPELTE",
>> + [SNIC_IOREQ_COMPLETE] = "SNIC_IOREQ_CMD_COMPELTE",
>> +};
>> +
>> +/* snic cmd status strings */
>> +static const char * const snic_io_status_str[] = {
>> + [SNIC_STAT_IO_SUCCESS] = "SNIC_STAT_IO_SUCCESS", /* 0x0 */
>> + [SNIC_STAT_INVALID_HDR] = "SNIC_STAT_INVALID_HDR",
>> + [SNIC_STAT_OUT_OF_RES] = "SNIC_STAT_OUT_OF_RES",
>> + [SNIC_STAT_INVALID_PARM] = "SNIC_STAT_INVALID_PARM",
>> + [SNIC_STAT_REQ_NOT_SUP] = "SNIC_STAT_REQ_NOT_SUP",
>> + [SNIC_STAT_IO_NOT_FOUND] = "SNIC_STAT_IO_NOT_FOUND",
>> + [SNIC_STAT_ABORTED] = "SNIC_STAT_ABORTED",
>> + [SNIC_STAT_TIMEOUT] = "SNIC_STAT_TIMEOUT",
>> + [SNIC_STAT_SGL_INVALID] = "SNIC_STAT_SGL_INVALID",
>> + [SNIC_STAT_DATA_CNT_MISMATCH] = "SNIC_STAT_DATA_CNT_MISMATCH",
>> + [SNIC_STAT_FW_ERR] = "SNIC_STAT_FW_ERR",
>> + [SNIC_STAT_ITMF_REJECT] = "SNIC_STAT_ITMF_REJECT",
>> + [SNIC_STAT_ITMF_FAIL] = "SNIC_STAT_ITMF_FAIL",
>> + [SNIC_STAT_ITMF_INCORRECT_LUN] = "SNIC_STAT_ITMF_INCORRECT_LUN",
>> + [SNIC_STAT_CMND_REJECT] = "SNIC_STAT_CMND_REJECT",
>> + [SNIC_STAT_DEV_OFFLINE] = "SNIC_STAT_DEV_OFFLINE",
>> + [SNIC_STAT_NO_BOOTLUN] = "SNIC_STAT_NO_BOOTLUN",
>> + [SNIC_STAT_SCSI_ERR] = "SNIC_STAT_SCSI_ERR",
>> + [SNIC_STAT_NOT_READY] = "SNIC_STAT_NOT_READY",
>> + [SNIC_STAT_FATAL_ERROR] = "SNIC_STAT_FATAL_ERROR",
>> +};
>> +
>> +static void snic_scsi_cleanup(struct snic *, int);
>> +
>> +const char *
>> +snic_state_to_str(unsigned int state)
>> +{
>> + if (state >= ARRAY_SIZE(snic_state_str) || !snic_state_str[state])
>> + return "Unknown";
>> +
>> + return snic_state_str[state];
>> +}
>> +
>> +static const char *
>> +snic_io_status_to_str(unsigned int state)
>> +{
>> + if ((state >= ARRAY_SIZE(snic_io_status_str)) ||
>> + (!snic_io_status_str[state]))
>> + return "Unknown";
>> +
>> + return snic_io_status_str[state];
>> +}
>> +
>> +static const char *
>> +snic_ioreq_state_to_str(unsigned int state)
>> +{
>> + if (state >= ARRAY_SIZE(snic_req_state_str) ||
>> + !snic_req_state_str[state])
>> + return "Unknown";
>> +
>> + return snic_req_state_str[state];
>> +}
>> +
>> +static inline spinlock_t *
>> +snic_io_lock_hash(struct snic *snic, struct scsi_cmnd *sc)
>> +{
>> + u32 hash = snic_cmd_tag(sc) & (SNIC_IO_LOCKS - 1);
>> +
>> + return &snic->io_req_lock[hash];
>> +}
>> +
>> +static inline spinlock_t *
>> +snic_io_lock_tag(struct snic *snic, int tag)
>> +{
>> + return &snic->io_req_lock[tag & (SNIC_IO_LOCKS - 1)];
>> +}
>> +
>> +/* snic_release_req_buf : Releases snic_req_info */
>> +static void
>> +snic_release_req_buf(struct snic *snic,
>> + struct snic_req_info *rqi,
>> + struct scsi_cmnd *sc)
>> +{
>> + struct snic_host_req *req = rqi_to_req(rqi);
>> +
>> + /* Freeing cmd without marking completion, not okay */
>> + SNIC_BUG_ON(!((CMD_STATE(sc) == SNIC_IOREQ_COMPLETE) ||
>> + (CMD_STATE(sc) == SNIC_IOREQ_ABTS_COMPLETE) ||
>> + (CMD_FLAGS(sc) & SNIC_DEV_RST_NOTSUP) ||
>> + (CMD_FLAGS(sc) & SNIC_IO_INTERNAL_TERM_ISSUED) ||
>> + (CMD_FLAGS(sc) & SNIC_DEV_RST_TERM_ISSUED) ||
>> + (CMD_FLAGS(sc) & SNIC_SCSI_CLEANUP) ||
>> + (CMD_STATE(sc) == SNIC_IOREQ_LR_COMPLETE)));
>> +
>> + SNIC_SCSI_DBG(snic->shost,
>> + "Rel_req:sc %p:tag %x:rqi %p:ioreq %p:abt %p:dr %p: state
>>%s:flags 0x%llx\n",
>> + sc, snic_cmd_tag(sc), rqi, rqi->req, rqi->abort_req,
>> + qi->dr_req, snic_ioreq_state_to_str(CMD_STATE(sc)),
>> + CMD_FLAGS(sc));
>> +
>> + if (req->u.icmnd.sense_addr)
>> + pci_unmap_single(snic->pdev,
>> + req->u.icmnd.sense_addr,
>> + SCSI_SENSE_BUFFERSIZE,
>> + PCI_DMA_FROMDEVICE);
>> +
>> + scsi_dma_unmap(sc);
>> +
>> + snic_req_free(snic, rqi);
>> +} /* end of snic_release_req_buf */
>> +
>> +/*
>> + * snic_queue_icmnd_req : Queues snic_icmnd request
>> + */
>> +static int
>> +snic_queue_icmnd_req(struct snic *snic,
>> + struct snic_req_info *rqi,
>> + struct scsi_cmnd *sc,
>> + int sg_cnt)
>> +{
>> + struct scatterlist *sg;
>> + struct snic_sg_desc *sgd;
>> + dma_addr_t pa = 0;
>> + struct scsi_lun lun;
>> + int flags = 0;
>> + int ret = 0;
>> + unsigned int i;
>> +
>> + if (sg_cnt) {
>> + flags = SNIC_ICMND_ESGL;
>> + sgd = (struct snic_sg_desc *) req_to_sgl(rqi->req);
>> +
>> + for_each_sg(scsi_sglist(sc), sg, sg_cnt, i) {
>> + sgd->addr = cpu_to_le64(sg_dma_address(sg));
>> + sgd->len = cpu_to_le32(sg_dma_len(sg));
>> + sgd->_resvd = 0;
>> + sgd++;
>> + }
>> + }
>> +
>> + pa = pci_map_single(snic->pdev,
>> + sc->sense_buffer,
>> + SCSI_SENSE_BUFFERSIZE,
>> + PCI_DMA_FROMDEVICE);
>> +
>> + if (pci_dma_mapping_error(snic->pdev, pa)) {
>> + SNIC_HOST_ERR(snic->shost,
>> + "QIcmnd:PCI Map Failed for sns buf %p tag %x\n",
>> + sc->sense_buffer, snic_cmd_tag(sc));
>> + ret = -ENOMEM;
>> +
>> + return ret;
>> + }
>> +
>> + int_to_scsilun(sc->device->lun, &lun);
>> + if (sc->sc_data_direction == DMA_FROM_DEVICE)
>> + flags |= SNIC_ICMND_RD;
>> + if (sc->sc_data_direction == DMA_TO_DEVICE)
>> + flags |= SNIC_ICMND_WR;
>> +
>> + /* Initialize icmnd */
>> + snic_icmnd_init(rqi->req,
>> + snic_cmd_tag(sc),
>> + snic->config.hid, /* hid */
>> + (u64)rqi,
>> + flags, /* command flags */
>> + rqi->tgt_id,
>> + lun.scsi_lun,
>> + sc->cmnd,
>> + sc->cmd_len,
>> + scsi_bufflen(sc),
>> + sg_cnt,
>> + (u64)req_to_sgl(rqi->req),
>> + pa, /* sense buffer pa */
>> + SCSI_SENSE_BUFFERSIZE);
>> +
>> + ret = snic_queue_wq_desc(snic, rqi->req, rqi->req_len);
>> + if (ret)
>> + SNIC_HOST_ERR(snic->shost,
>> + "QIcmnd: Queuing Icmnd Failed. ret = %d\n",
>> + ret);
>> +
>> + return ret;
>> +} /* end of snic_queue_icmnd_req */
>> +
>> +/*
>> + * snic_issue_scsi_req : Prepares IO request and Issues to FW.
>> + */
>> +static int
>> +snic_issue_scsi_req(struct snic *snic,
>> + struct snic_tgt *tgt,
>> + struct scsi_cmnd *sc)
>> +{
>> + struct snic_req_info *rqi = NULL;
>> + int sg_cnt = 0;
>> + int ret = 0;
>> + u32 tag = snic_cmd_tag(sc);
>> + u64 cmd_trc = 0, cmd_st_flags = 0;
>> + spinlock_t *io_lock = NULL;
>> + unsigned long flags;
>> +
>> + CMD_STATE(sc) = SNIC_IOREQ_NOT_INITED;
>> + CMD_FLAGS(sc) = SNIC_NO_FLAGS;
>> + sg_cnt = scsi_dma_map(sc);
>> + if (sg_cnt < 0) {
>> + SNIC_TRC((u16)snic->shost->host_no, tag, (u64)sc, 0,
>> + sc->cmnd[0], sg_cnt, CMD_STATE(sc));
>> +
>> + SNIC_HOST_ERR(snic->shost, "issue_sc:Failed to map SG List.\n");
>> + ret = -ENOMEM;
>> +
>> + goto issue_sc_end;
>> + }
>> +
>> + rqi = snic_req_init(snic, sg_cnt);
>> + if (!rqi) {
>> + scsi_dma_unmap(sc);
>> + ret = -ENOMEM;
>> +
>> + goto issue_sc_end;
>> + }
>> +
>> + rqi->tgt_id = tgt->id;
>> + rqi->sc = sc;
>> +
>> + CMD_STATE(sc) = SNIC_IOREQ_PENDING;
>> + CMD_SP(sc) = (char *) rqi;
>> + cmd_trc = SNIC_TRC_CMD(sc);
>> + CMD_FLAGS(sc) |= (SNIC_IO_INITIALIZED | SNIC_IO_ISSUED);
>> + cmd_st_flags = SNIC_TRC_CMD_STATE_FLAGS(sc);
>> + io_lock = snic_io_lock_hash(snic, sc);
>> +
>> + /* create wq desc and enqueue it */
>> + ret = snic_queue_icmnd_req(snic, rqi, sc, sg_cnt);
>> + if (ret) {
>> + SNIC_HOST_ERR(snic->shost,
>> + "issue_sc: icmnd qing Failed for sc %p, err %d\n",
>> + sc, ret);
>> +
>> + spin_lock_irqsave(io_lock, flags);
>> + rqi = (struct snic_req_info *) CMD_SP(sc);
>> + CMD_SP(sc) = NULL;
>> + CMD_STATE(sc) = SNIC_IOREQ_COMPLETE;
>> + CMD_FLAGS(sc) &= ~SNIC_IO_ISSUED; /* turn off the flag */
>> + spin_unlock_irqrestore(io_lock, flags);
>> +
>> + if (rqi)
>> + snic_release_req_buf(snic, rqi, sc);
>> +
>> + SNIC_TRC(snic->shost->host_no, tag, (u64)sc, 0, 0, 0,
>> + SNIC_TRC_CMD_STATE_FLAGS(sc));
>> + } else {
>> + u32 io_sz = scsi_bufflen(sc) >> 9;
>> + u32 qtime = jiffies - rqi->start_time;
>> + struct snic_io_stats *iostats = &snic->s_stats.io;
>> +
>> + if (io_sz > atomic64_read(&iostats->max_io_sz))
>> + atomic64_set(&iostats->max_io_sz, io_sz);
>> +
>> + if (qtime > atomic64_read(&iostats->max_qtime))
>> + atomic64_set(&iostats->max_qtime, qtime);
>> +
>> + SNIC_SCSI_DBG(snic->shost,
>> + "issue_sc:sc %p, tag %d queued to WQ.\n",
>> + sc, tag);
>> +
>> + SNIC_TRC(snic->shost->host_no, tag, (u64)sc, (u64)rqi, sg_cnt,
>> + cmd_trc, cmd_st_flags);
>> + }
>> +
>> +issue_sc_end:
>> +
>> + return ret;
>> +} /* end of snic_issue_scsi_req */
>> +
>> +
>> +/*
>> + * snic_queuecommand
>> + * Routine to send a scsi cdb to LLD
>> + * Called with host_lock held and interrupts disabled
>> + */
>> +int
>> +snic_queuecommand(struct Scsi_Host *shost, struct scsi_cmnd *sc)
>> +{
>> + struct snic_tgt *tgt = NULL;
>> + struct snic *snic = shost_priv(shost);
>> + int ret;
>> +
>> + tgt = starget_to_tgt(scsi_target(sc->device));
>> + ret = snic_tgt_chkready(tgt);
>> + if (ret) {
>> + SNIC_HOST_ERR(shost, "Tgt %p id %d Not Ready.\n", tgt, tgt->id);
>> + atomic64_inc(&snic->s_stats.misc.tgt_not_rdy);
>> + sc->result = ret;
>> + sc->scsi_done(sc);
>> +
>> + return 0;
>> + }
>> +
>> + if (snic_get_state(snic) != SNIC_ONLINE) {
>> + SNIC_HOST_ERR(shost, "snic state is %s\n",
>> + snic_state_str[snic_get_state(snic)]);
>> +
>> + return SCSI_MLQUEUE_HOST_BUSY;
>> + }
>> + atomic_inc(&snic->ios_inflight);
>> +
>> + SNIC_SCSI_DBG(shost, "sc %p Tag %d (sc %0x) lun %lld in snic_qcmd\n",
>> + sc, snic_cmd_tag(sc), sc->cmnd[0], sc->device->lun);
>> +
>> + memset(scsi_cmd_priv(sc), 0, sizeof(struct snic_internal_io_state));
>> +
>> +#ifdef SNIC_DEBUG
>> + scsi_print_command(sc);
>> +#endif
>> +
>> + ret = snic_issue_scsi_req(snic, tgt, sc);
>> + if (ret) {
>> + SNIC_HOST_ERR(shost, "Failed to Q, Scsi Req w/ err %d.\n", ret);
>> + ret = SCSI_MLQUEUE_HOST_BUSY;
>> + } else
>> + snic_stats_update_active_ios(&snic->s_stats);
>> +
>> + atomic_dec(&snic->ios_inflight);
>> +
>> + return ret;
>> +} /* end of snic_queuecommand */
>> +
>> +/*
>> + * snic_process_abts_pending_state:
>> + * caller should hold IO lock
>> + */
>> +static void
>> +snic_proc_tmreq_pending_state(struct snic *snic,
>> + struct scsi_cmnd *sc,
>> + u8 cmpl_status)
>> +{
>> + int state = CMD_STATE(sc);
>> +
>> + if (state == SNIC_IOREQ_ABTS_PENDING)
>> + CMD_FLAGS(sc) |= SNIC_IO_ABTS_PENDING;
>> + else if (state == SNIC_IOREQ_LR_PENDING)
>> + CMD_FLAGS(sc) |= SNIC_DEV_RST_PENDING;
>> + else
>> + SNIC_BUG_ON(1);
>> +
>> + switch (cmpl_status) {
>> + case SNIC_STAT_IO_SUCCESS:
>> + CMD_FLAGS(sc) |= SNIC_IO_DONE;
>> + break;
>> +
>> + case SNIC_STAT_ABORTED:
>> + CMD_FLAGS(sc) |= SNIC_IO_ABORTED;
>> + break;
>> +
>> + default:
>> + SNIC_BUG_ON(1);
>> + }
>> +}
>> +
>> +/*
>> + * snic_process_io_failed_state:
>> + * Processes IO's error states
>> + */
>> +static void
>> +snic_process_io_failed_state(struct snic *snic,
>> + struct snic_icmnd_cmpl *icmnd_cmpl,
>> + struct scsi_cmnd *sc,
>> + u8 cmpl_stat)
>> +{
>> + int res = 0;
>> +
>> + switch (cmpl_stat) {
>> + case SNIC_STAT_TIMEOUT: /* Req was timedout */
>> + atomic64_inc(&snic->s_stats.misc.io_tmo);
>> + res = DID_TIME_OUT;
>> + break;
>> +
>> + case SNIC_STAT_ABORTED: /* Req was aborted */
>> + atomic64_inc(&snic->s_stats.misc.io_aborted);
>> + res = DID_ERROR;
>> + break;
>> +
>DID_ABORTED?
Sure, I will change it to DID_ABORT.
>
>> + case SNIC_STAT_DATA_CNT_MISMATCH:/* Recv/Sent more/less data than exp
>>*/
>> + atomic64_inc(&snic->s_stats.misc.data_cnt_mismat);
>> + scsi_set_resid(sc, icmnd_cmpl->resid);
>> + res = DID_ERROR;
>> + break;
>> +
>> + case SNIC_STAT_OUT_OF_RES: /* Out of resources to complete request */
>> + atomic64_inc(&snic->s_stats.fw.out_of_res);
>> + res = DID_REQUEUE;
>> + break;
>> +
>> + case SNIC_STAT_IO_NOT_FOUND: /* Requested I/O was not found */
>> + atomic64_inc(&snic->s_stats.io.io_not_found);
>> + res = DID_ERROR;
>> + break;
>> +
>> + case SNIC_STAT_SGL_INVALID: /* Req was aborted to due to sgl error*/
>> + atomic64_inc(&snic->s_stats.misc.sgl_inval);
>> + res = DID_ERROR;
>> + break;
>> +
>> + case SNIC_STAT_FW_ERR: /* Req terminated due to FW Error */
>> + atomic64_inc(&snic->s_stats.fw.io_errs);
>> + res = DID_ERROR;
>> + break;
>> +
>> + case SNIC_STAT_SCSI_ERR: /* FW hits SCSI Error */
>> + atomic64_inc(&snic->s_stats.fw.scsi_errs);
>> + res = DID_ERROR;
>> + break;
>> +
>SCSI error? What should stat be? If it's the normal SAM_STAT_XXX values
>it should be decoded correctly.
Thanks for pointing this, FW sends CHECKCONDITION status, sense
information to midlyaer, and expect mid layer to act on it. In this case,
DID_ERROR causes to take no action on check condition status. So passing
scsi status without DID_ERROR will make mid layer to act on CHECKCONDITION.
>
>> + case SNIC_STAT_INVALID_HDR: /* Hdr contains invalid data */
>> + case SNIC_STAT_INVALID_PARM: /* Some param in req is invalid */
>> + case SNIC_STAT_REQ_NOT_SUP: /* Req type is not supported */
>> + case SNIC_STAT_CMND_REJECT: /* Req rejected */
>> + case SNIC_STAT_DEV_OFFLINE: /* Device offline */
>> + case SNIC_STAT_NOT_READY: /* XPT yet to initialize */
>> + case SNIC_STAT_FATAL_ERROR: /* XPT Error */
>> + default:
>> + SNIC_SCSI_DBG(snic->shost,
>> + "Invalid Hdr/Param or Req Not Supported or Cmnd Rejected or
>>Device Offline. or Unknown\n");
>> + res = DID_ERROR;
>> + break;
>> + }
>> +
>You should be returning 'DID_NO_CONNECT' for DEV_OFFLINE. If NOT_READY
>is temporary you should be returning DID_RETRY or similar.
>Otherwise DID_NO_CONNECT is preferred, too.
Sure, I will change it to DID_NO_CONNECT for DEV OFFLINE, And DID_RETRY
for NOT_READY.
>
>> + SNIC_HOST_ERR(snic->shost, "fw returns failed status %s flags
>>0x%llx\n",
>> + snic_io_status_to_str(cmpl_stat), CMD_FLAGS(sc));
>> +
>> + /* Set sc->result */
>> + sc->result = (res << 16) | icmnd_cmpl->scsi_status;
>> +} /* end of snic_process_io_failed_state */
>> +
>> +/*
>> + * snic_tmreq_pending : is task management in progress.
>> + */
>> +static int
>> +snic_tmreq_pending(struct scsi_cmnd *sc)
>> +{
>> + int state = CMD_STATE(sc);
>> +
>> + return ((state == SNIC_IOREQ_ABTS_PENDING) ||
>> + (state == SNIC_IOREQ_LR_PENDING));
>> +}
>> +
>> +/*
>> + * snic_process_icmnd_cmpl_status:
>> + * Caller should hold io_lock
>> + */
>> +static int
>> +snic_process_icmnd_cmpl_status(struct snic *snic,
>> + struct snic_icmnd_cmpl *icmnd_cmpl,
>> + u8 cmpl_stat,
>> + struct scsi_cmnd *sc)
>> +{
>> + u8 scsi_stat = icmnd_cmpl->scsi_status;
>> + u64 xfer_len = 0;
>> + int ret = 0;
>> +
>> + /* Mark the IO as complete */
>> + CMD_STATE(sc) = SNIC_IOREQ_COMPLETE;
>> +
>> + if (likely(cmpl_stat == SNIC_STAT_IO_SUCCESS)) {
>> + sc->result = (DID_OK << 16) | scsi_stat;
>> +
>> + xfer_len = scsi_bufflen(sc);
>> +
>> + /* Update SCSI Cmd with resid value */
>> + scsi_set_resid(sc, icmnd_cmpl->resid);
>> +
>> + if (icmnd_cmpl->flags & SNIC_ICMND_CMPL_UNDR_RUN) {
>> + xfer_len -= icmnd_cmpl->resid;
>> + atomic64_inc(&snic->s_stats.misc.io_under_run);
>> + }
>> +
>> + if (icmnd_cmpl->scsi_status == SAM_STAT_TASK_SET_FULL)
>> + atomic64_inc(&snic->s_stats.misc.qfull);
>> +
>> + ret = 0;
>> + } else {
>> + snic_process_io_failed_state(snic, icmnd_cmpl, sc, cmpl_stat);
>> + atomic64_inc(&snic->s_stats.io.fail);
>> + SNIC_HOST_ERR(snic->shost,
>> + "icmnd_cmpl: IO Failed : Hdr Status %s flags 0x%llx\n",
>> + snic_io_status_to_str(cmpl_stat), CMD_FLAGS(sc));
>> + ret = 1;
>> + }
>> +
>> + return ret;
>> +} /* end of snic_process_icmnd_cmpl_status */
>> +
>> +
>> +/*
>> + * snic_icmnd_cmpl_handler
>> + * Routine to handle icmnd completions
>> + */
>> +static void
>> +snic_icmnd_cmpl_handler(struct snic *snic, struct snic_fw_req *fwreq)
>> +{
>> + u8 typ, hdr_stat;
>> + u32 cmnd_id, hid;
>> + u64 ctx;
>> + struct scsi_cmnd *sc = NULL;
>> + struct snic_icmnd_cmpl *icmnd_cmpl = NULL;
>> + struct snic_host_req *req = NULL;
>> + struct snic_req_info *rqi = NULL;
>> + unsigned long flags, start_time;
>> + spinlock_t *io_lock;
>> + u8 sc_stat = 0;
>> +
>> + snic_io_hdr_dec(&fwreq->hdr, &typ, &hdr_stat, &cmnd_id, &hid, &ctx);
>> + icmnd_cmpl = &fwreq->u.icmnd_cmpl;
>> + sc_stat = icmnd_cmpl->scsi_status;
>> +
>> + SNIC_SCSI_DBG(snic->shost,
>> + "Icmnd_cmpl: type = %x, hdr_stat = %x, cmnd_id = %x, hid =
>>%x,i ctx = %llx\n",
>> + typ, hdr_stat, cmnd_id, hid, ctx);
>> +
>> + if (cmnd_id >= snic->max_tag_id) {
>> + SNIC_HOST_ERR(snic->shost,
>> + "Icmnd_cmpl:Tag Error:Out of Range Tag %d, hdr status = %s\n",
>> + cmnd_id, snic_io_status_to_str(hdr_stat));
>> + return;
>> + }
>> +
>> + sc = scsi_host_find_tag(snic->shost, cmnd_id);
>> + WARN_ON_ONCE(!sc);
>> +
>> + if (!sc) {
>> + atomic64_inc(&snic->s_stats.io.sc_null);
>> + SNIC_HOST_ERR(snic->shost,
>> + "Icmnd_cmpl: Scsi Cmnd Not found, sc = NULL Hdr Status = %s
>>tag = 0x%x fwreq = 0x%p\n",
>> + snic_io_status_to_str(hdr_stat),
>> + cmnd_id,
>> + fwreq);
>> +
>> + SNIC_TRC(snic->shost->host_no, cmnd_id, 0,
>> + ((u64)hdr_stat << 16 |
>> + (u64)sc_stat << 8 | (u64)icmnd_cmpl->flags),
>> + fwreq, icmnd_cmpl->resid, ctx);
>> +
>> + return;
>> + }
>> +
>> +
>> +#ifdef SNIC_DEBUG
>> + scsi_print_result(sc);
>> +#endif
>> +
>> + io_lock = snic_io_lock_hash(snic, sc);
>> +
>> + spin_lock_irqsave(io_lock, flags);
>> + rqi = (struct snic_req_info *) CMD_SP(sc);
>> + SNIC_SCSI_DBG(snic->shost,
>> + "Icmnd_cmpl:lun %lld sc %p cmd %xtag %d flags 0x%llx rqi %p\n",
>> + sc->device->lun, sc, sc->cmnd[0], snic_cmd_tag(sc),
>> + CMD_FLAGS(sc), rqi);
>> +
>> + SNIC_BUG_ON(rqi != (struct snic_req_info *)ctx);
>> + WARN_ON_ONCE(req);
>> + if (!rqi) {
>> + atomic64_inc(&snic->s_stats.io.req_null);
>> + CMD_FLAGS(sc) |= SNIC_IO_REQ_NULL;
>> + spin_unlock_irqrestore(io_lock, flags);
>> +
>> + SNIC_HOST_ERR(snic->shost,
>> + "Icmnd_cmpl:Host Req Not Found(null), Hdr Status %s, Tag
>>0x%x, sc 0x%p flags 0x%llx\n",
>> + snic_io_status_to_str(hdr_stat),
>> + cmnd_id, sc, CMD_FLAGS(sc));
>> + return;
>> + }
>> +
>> + rqi = (struct snic_req_info *) ctx;
>> + start_time = rqi->start_time;
>> +
>> + /* firmware completed the io */
>> + rqi->io_cmpl = 1;
>> +
>> + /*
>> + * if SCSI-ML has already issued abort on this command,
>> + * ignore completion of the IO. The abts path will clean it up
>> + */
>> + if (unlikely(snic_tmreq_pending(sc))) {
>> + snic_proc_tmreq_pending_state(snic, sc, hdr_stat);
>> + spin_unlock_irqrestore(io_lock, flags);
>> +
>> + snic_stats_update_io_cmpl(&snic->s_stats);
>> +
>> + /* Expected value is SNIC_STAT_ABORTED */
>> + if (likely(hdr_stat == SNIC_STAT_ABORTED))
>> + return;
>> +
>> + SNIC_SCSI_DBG(snic->shost,
>> + "icmnd_cmpl:TM Req Pending(%s), Hdr Status %s sc 0x%p scsi
>>status %x resid %d flags 0x%llx\n",
>> + snic_ioreq_state_to_str(CMD_STATE(sc)),
>> + snic_io_status_to_str(hdr_stat),
>> + sc, sc_stat, icmnd_cmpl->resid, CMD_FLAGS(sc));
>> +
>> + SNIC_TRC(snic->shost->host_no, cmnd_id, sc,
>> + jiffies_to_msecs(jiffies - start_time), fwreq,
>> + SNIC_TRC_CMD(sc), SNIC_TRC_CMD_STATE_FLAGS(sc));
>> +
>> + return;
>> + }
>> +
>> + if (snic_process_icmnd_cmpl_status(snic, icmnd_cmpl, hdr_stat, sc)) {
>> +#ifdef SNIC_DEBUG
>> + scsi_print_command(sc);
>> +#endif
>> + SNIC_HOST_ERR(snic->shost,
>> + "icmnd_cmpl:IO Failed, sc 0x%p Tag %d Cmd %x Hdr Status %s
>>flags 0x%llx\n",
>> + sc, sc->cmnd[0], cmnd_id,
>> + snic_io_status_to_str(hdr_stat), CMD_FLAGS(sc));
>> + }
>> +
>> + /* Break link with the SCSI Command */
>> + CMD_SP(sc) = NULL;
>> + CMD_FLAGS(sc) |= SNIC_IO_DONE;
>> +
>> + spin_unlock_irqrestore(io_lock, flags);
>> +
>> + /* For now, consider only successful IO. */
>> + snic_calc_io_process_time(snic, rqi);
>> +
>> + snic_release_req_buf(snic, rqi, sc);
>> +
>> + SNIC_TRC(snic->shost->host_no, cmnd_id, (u64)sc,
>> + jiffies_to_msecs(jiffies - start_time), (u64)fwreq,
>> + SNIC_TRC_CMD(sc), SNIC_TRC_CMD_STATE_FLAGS(sc));
>> +
>> +
>> + if (sc->scsi_done)
>> + sc->scsi_done(sc);
>> +
>> + snic_stats_update_io_cmpl(&snic->s_stats);
>> +} /* end of snic_icmnd_cmpl_handler */
>> +
>> +static void
>> +snic_proc_dr_cmpl_locked(struct snic *snic,
>> + struct snic_fw_req *fwreq,
>> + u8 cmpl_stat,
>> + u32 cmnd_id,
>> + struct scsi_cmnd *sc)
>> +{
>> + struct snic_req_info *rqi = (struct snic_req_info *) CMD_SP(sc);
>> + u32 start_time = rqi->start_time;
>> +
>> + CMD_LR_STATUS(sc) = cmpl_stat;
>> +
>> + SNIC_SCSI_DBG(snic->shost, "itmf_cmpl: Cmd State = %s\n",
>> + snic_ioreq_state_to_str(CMD_STATE(sc)));
>> +
>> + if (CMD_STATE(sc) == SNIC_IOREQ_ABTS_PENDING) {
>> + CMD_FLAGS(sc) |= SNIC_DEV_RST_ABTS_PENDING;
>> +
>> + SNIC_TRC(snic->shost->host_no, cmnd_id, sc,
>> + jiffies_to_msecs(jiffies - start_time),
>> + fwreq, 0, SNIC_TRC_CMD_STATE_FLAGS(sc));
>> +
>> + SNIC_SCSI_DBG(snic->shost,
>> + "itmf_cmpl: Terminate Pending Dev Reset Cmpl Recvd.id %x,
>>status %s flags 0x%llx\n",
>> + (int)(cmnd_id & SNIC_TAG_MASK),
>> + snic_io_status_to_str(cmpl_stat),
>> + CMD_FLAGS(sc));
>> +
>> + return;
>> + }
>> +
>> +
>> + if (CMD_FLAGS(sc) & SNIC_DEV_RST_TIMEDOUT) {
>> + SNIC_TRC(snic->shost->host_no, cmnd_id, sc,
>> + jiffies_to_msecs(jiffies - start_time),
>> + fwreq, 0, SNIC_TRC_CMD_STATE_FLAGS(sc));
>> +
>> + SNIC_SCSI_DBG(snic->shost,
>> + "itmf_cmpl:Dev Reset Completion Received after timeout. id %d
>>cmpl status %s flags 0x%llx\n",
>> + (int)(cmnd_id & SNIC_TAG_MASK),
>> + snic_io_status_to_str(cmpl_stat),
>> + CMD_FLAGS(sc));
>> +
>> + return;
>> + }
>> +
>> + CMD_STATE(sc) = SNIC_IOREQ_LR_COMPLETE;
>> + CMD_FLAGS(sc) |= SNIC_DEV_RST_DONE;
>> +
>> + SNIC_SCSI_DBG(snic->shost,
>> + "itmf_cmpl:Dev Reset Cmpl Recvd id %d cmpl status %s flags
>>0x%llx\n",
>> + (int)(cmnd_id & SNIC_TAG_MASK),
>> + snic_io_status_to_str(cmpl_stat),
>> + CMD_FLAGS(sc));
>> +
>> + if (rqi->dr_done)
>> + complete(rqi->dr_done);
>> +} /* end of snic_proc_dr_cmpl_locked */
>> +
>> +/*
>> + * snic_update_abort_stats : Updates abort stats based on completion
>>status.
>> + */
>> +static void
>> +snic_update_abort_stats(struct snic *snic, u8 cmpl_stat)
>> +{
>> + struct snic_abort_stats *abt_stats = &snic->s_stats.abts;
>> +
>> + SNIC_SCSI_DBG(snic->shost, "Updating Abort stats.\n");
>> +
>> + switch (cmpl_stat) {
>> + case SNIC_STAT_IO_SUCCESS:
>> + break;
>> +
>> + case SNIC_STAT_TIMEOUT:
>> + atomic64_inc(&abt_stats->fw_tmo);
>> + break;
>> +
>> + case SNIC_STAT_IO_NOT_FOUND:
>> + atomic64_inc(&abt_stats->io_not_found);
>> + break;
>> +
>> + default:
>> + atomic64_inc(&abt_stats->fail);
>> + break;
>> + }
>> +}
>> +
>> +static int
>> +snic_process_itmf_cmpl(struct snic *snic,
>> + struct snic_fw_req *fwreq,
>> + u32 cmnd_id,
>> + u8 cmpl_stat,
>> + struct scsi_cmnd *sc)
>> +{
>> + struct snic_req_info *rqi = NULL;
>> + u32 tm_tags = 0;
>> + spinlock_t *io_lock = NULL;
>> + unsigned long flags;
>> + u32 start_time = 0;
>> + int ret = 0;
>> +
>> + io_lock = snic_io_lock_hash(snic, sc);
>> + spin_lock_irqsave(io_lock, flags);
>> + rqi = (struct snic_req_info *) CMD_SP(sc);
>> + WARN_ON_ONCE(!rqi);
>> +
>> + if (!rqi) {
>> + atomic64_inc(&snic->s_stats.io.req_null);
>> + spin_unlock_irqrestore(io_lock, flags);
>> + CMD_FLAGS(sc) |= SNIC_IO_ABTS_TERM_REQ_NULL;
>> + SNIC_HOST_ERR(snic->shost,
>> + "itmf_cmpl: rqi is null,Hdr stat = %s Tag = 0x%x sc = 0x%p
>>flags 0x%llx\n",
>> + snic_io_status_to_str(cmpl_stat), cmnd_id, sc,
>> + CMD_FLAGS(sc));
>> +
>> + return ret;
>> + }
>> +
>> + /* Extract task management flags */
>> + tm_tags = cmnd_id & ~(SNIC_TAG_MASK);
>> +
>> + start_time = rqi->start_time;
>> + cmnd_id &= (SNIC_TAG_MASK);
>> +
>> + switch (tm_tags) {
>> + case SNIC_TAG_ABORT:
>> + /* Abort only issued on cmd */
>> + snic_update_abort_stats(snic, cmpl_stat);
>> +
>> + if (CMD_STATE(sc) != SNIC_IOREQ_ABTS_PENDING) {
>> + /* This is a late completion. Ignore it. */
>> + ret = -1;
>> + spin_unlock_irqrestore(io_lock, flags);
>> + break;
>> + }
>> +
>> + CMD_STATE(sc) = SNIC_IOREQ_ABTS_COMPLETE;
>> + CMD_ABTS_STATUS(sc) = cmpl_stat;
>> + CMD_FLAGS(sc) |= SNIC_IO_ABTS_TERM_DONE;
>> +
>> + SNIC_SCSI_DBG(snic->shost,
>> + "itmf_cmpl:Abort Cmpl Recvd.Tag 0x%x Status %s flags
>>0x%llx\n",
>> + cmnd_id,
>> + snic_io_status_to_str(cmpl_stat),
>> + CMD_FLAGS(sc));
>> +
>> + /*
>> + * If scsi_eh thread is blocked waiting for abts complete,
>> + * signal completion to it. IO will be cleaned in the thread,
>> + * else clean it in this context.
>> + */
>> + if (rqi->abts_done) {
>> + complete(rqi->abts_done);
>> + spin_unlock_irqrestore(io_lock, flags);
>> +
>> + break; /* jump out */
>> + }
>> +
>> + CMD_SP(sc) = NULL;
>> + sc->result = (DID_ERROR << 16);
>> + SNIC_SCSI_DBG(snic->shost,
>> + "itmf_cmpl: Completing IO. sc %p flags 0x%llx\n",
>> + sc, CMD_FLAGS(sc));
>> +
>> + spin_unlock_irqrestore(io_lock, flags);
>> +
>> + snic_release_req_buf(snic, rqi, sc);
>> +
>> + if (sc->scsi_done) {
>> + SNIC_TRC(snic->shost->host_no, cmnd_id, sc,
>> + jiffies_to_msecs(jiffies - start_time),
>> + fwreq,
>> + SNIC_TRC_CMD(sc),
>> + SNIC_TRC_CMD_STATE_FLAGS(sc));
>> +
>> + sc->scsi_done(sc);
>> + }
>> +
>> + break;
>> +
>> + case SNIC_TAG_DEV_RST:
>> + case SNIC_TAG_DEV_RST | SNIC_TAG_IOCTL_DEV_RST:
>> + snic_proc_dr_cmpl_locked(snic, fwreq, cmpl_stat, cmnd_id, sc);
>> + spin_unlock_irqrestore(io_lock, flags);
>> + ret = 0;
>> +
>> + break;
>> +
>> + case SNIC_TAG_ABORT | SNIC_TAG_DEV_RST:
>> + /* Abort and terminate completion of device reset req */
>> +
>> + CMD_STATE(sc) = SNIC_IOREQ_ABTS_COMPLETE;
>> + CMD_ABTS_STATUS(sc) = cmpl_stat;
>> + CMD_FLAGS(sc) |= SNIC_DEV_RST_DONE;
>> +
>> + SNIC_SCSI_DBG(snic->shost,
>> + "itmf_cmpl:dev reset abts cmpl recvd. id %d status %s flags
>>0x%llx\n",
>> + cmnd_id, snic_io_status_to_str(cmpl_stat),
>> + CMD_FLAGS(sc));
>> +
>> + if (rqi->abts_done)
>> + complete(rqi->abts_done);
>> +
>> + spin_unlock_irqrestore(io_lock, flags);
>> +
>> + break;
>> +
>> + default:
>> + spin_unlock_irqrestore(io_lock, flags);
>> + SNIC_HOST_ERR(snic->shost,
>> + "itmf_cmpl: Unknown TM tag bit 0x%x\n", tm_tags);
>> +
>> + SNIC_HOST_ERR(snic->shost,
>> + "itmf_cmpl:Unexpected itmf io stat %s Tag = 0x%x flags
>>0x%llx\n",
>> + snic_ioreq_state_to_str(CMD_STATE(sc)),
>> + cmnd_id,
>> + CMD_FLAGS(sc));
>> + ret = -1;
>> + SNIC_BUG_ON(1);
>> +
>> + break;
>> + }
>> +
>> + return ret;
>> +} /* end of snic_process_itmf_cmpl_status */
>> +
>> +/*
>> + * snic_itmf_cmpl_handler.
>> + * Routine to handle itmf completions.
>> + */
>> +static void
>> +snic_itmf_cmpl_handler(struct snic *snic, struct snic_fw_req *fwreq)
>> +{
>> + struct scsi_cmnd *sc = NULL;
>> + struct snic_req_info *rqi = NULL;
>> + struct snic_itmf_cmpl *itmf_cmpl = NULL;
>> + u64 ctx;
>> + u32 cmnd_id;
>> + u32 hid;
>> + u8 typ;
>> + u8 hdr_stat;
>> +
>> + snic_io_hdr_dec(&fwreq->hdr, &typ, &hdr_stat, &cmnd_id, &hid, &ctx);
>> + SNIC_SCSI_DBG(snic->shost,
>> + "Itmf_cmpl: %s: type = %x, hdr_stat = %x, cmnd_id = %x, hid =
>>%x,ctx = %llx\n",
>> + __func__, typ, hdr_stat, cmnd_id, hid, ctx);
>> +
>> + itmf_cmpl = &fwreq->u.itmf_cmpl;
>> + SNIC_SCSI_DBG(snic->shost,
>> + "Itmf_cmpl: nterm %u , flags 0x%x\n",
>> + itmf_cmpl->nterminated, itmf_cmpl->flags);
>> +
>> + /* spl case, dev reset issued through ioctl */
>> + if (cmnd_id & SNIC_TAG_IOCTL_DEV_RST) {
>> + rqi = (struct snic_req_info *) ctx;
>> + sc = rqi->sc;
>> +
>> + goto ioctl_dev_rst;
>> + }
>> +
>> + if ((cmnd_id & SNIC_TAG_MASK) >= snic->max_tag_id) {
>> + SNIC_HOST_ERR(snic->shost,
>> + "Itmf_cmpl: Tag 0x%x out of Range,HdrStat %s\n",
>> + cmnd_id, snic_io_status_to_str(hdr_stat));
>> + SNIC_BUG_ON(1);
>> +
>> + return;
>> + }
>> +
>> + sc = scsi_host_find_tag(snic->shost, cmnd_id & SNIC_TAG_MASK);
>> + WARN_ON_ONCE(!sc);
>> +
>> +ioctl_dev_rst:
>> + if (!sc) {
>> + atomic64_inc(&snic->s_stats.io.sc_null);
>> + SNIC_HOST_ERR(snic->shost,
>> + "Itmf_cmpl: sc is NULL - Hdr Stat %s Tag 0x%x\n",
>> + snic_io_status_to_str(hdr_stat), cmnd_id);
>> +
>> + return;
>> + }
>> +
>> + snic_process_itmf_cmpl(snic, fwreq, cmnd_id, hdr_stat, sc);
>> +} /* end of snic_itmf_cmpl_handler */
>> +
>> +
>> +
>> +static void
>> +snic_hba_reset_scsi_cleanup(struct snic *snic, struct scsi_cmnd *sc)
>> +{
>> + struct snic_stats *st = &snic->s_stats;
>> + long act_ios = 0, act_fwreqs = 0;
>> +
>> + SNIC_SCSI_DBG(snic->shost, "HBA Reset scsi cleanup.\n");
>> + snic_scsi_cleanup(snic, snic_cmd_tag(sc));
>> +
>> + /* Update stats on pending IOs */
>> + act_ios = atomic64_read(&st->io.active);
>> + atomic64_add(act_ios, &st->io.compl);
>> + atomic64_sub(act_ios, &st->io.active);
>> +
>> + act_fwreqs = atomic64_read(&st->fw.actv_reqs);
>> + atomic64_sub(act_fwreqs, &st->fw.actv_reqs);
>> +}
>> +
>> +/*
>> + * snic_hba_reset_cmpl_handler :
>> + *
>> + * Notes :
>> + * 1. Cleanup all the scsi cmds, release all snic specific cmds
>> + * 2. Issue Report Targets in case of SAN targets
>> + */
>> +static int
>> +snic_hba_reset_cmpl_handler(struct snic *snic, struct snic_fw_req
>>*fwreq)
>> +{
>> + u64 ctx;
>> + u32 cmnd_id;
>> + u32 hid;
>> + u8 typ;
>> + u8 hdr_stat;
>> + struct scsi_cmnd *sc = NULL;
>> + struct snic_req_info *rqi = NULL;
>> + spinlock_t *io_lock = NULL;
>> + unsigned long flags, gflags;
>> + int ret = 0;
>> +
>> + SNIC_HOST_INFO(snic->shost,
>> + "reset_cmpl:HBA Reset Completion received.\n");
>> +
>> + snic_io_hdr_dec(&fwreq->hdr, &typ, &hdr_stat, &cmnd_id, &hid, &ctx);
>> + SNIC_SCSI_DBG(snic->shost,
>> + "reset_cmpl: type = %x, hdr_stat = %x, cmnd_id = %x, hid = %x,
>>ctx = %llx\n",
>> + typ, hdr_stat, cmnd_id, hid, ctx);
>> +
>> + /* spl case, host reset issued through ioctl */
>> + if (cmnd_id == SCSI_NO_TAG) {
>> + rqi = (struct snic_req_info *) ctx;
>> + sc = rqi->sc;
>> +
>> + goto ioctl_hba_rst;
>> + }
>> +
>> + if (cmnd_id >= snic->max_tag_id) {
>> + SNIC_HOST_ERR(snic->shost,
>> + "reset_cmpl: Tag 0x%x out of Range,HdrStat %s\n",
>> + cmnd_id, snic_io_status_to_str(hdr_stat));
>> + SNIC_BUG_ON(1);
>> +
>> + return 1;
>> + }
>> +
>> + sc = scsi_host_find_tag(snic->shost, cmnd_id);
>> +ioctl_hba_rst:
>> + if (!sc) {
>> + atomic64_inc(&snic->s_stats.io.sc_null);
>> + SNIC_HOST_ERR(snic->shost,
>> + "reset_cmpl: sc is NULL - Hdr Stat %s Tag 0x%x\n",
>> + snic_io_status_to_str(hdr_stat), cmnd_id);
>> + ret = 1;
>> +
>> + return ret;
>> + }
>> +
>> + io_lock = snic_io_lock_hash(snic, sc);
>> + spin_lock_irqsave(io_lock, flags);
>> +
>> + if (!snic->remove_wait) {
>> + spin_unlock_irqrestore(io_lock, flags);
>> + SNIC_HOST_ERR(snic->shost,
>> + "reset_cmpl:host reset completed after timout\n");
>> + ret = 1;
>> +
>> + return ret;
>> + }
>> +
>> + rqi = (struct snic_req_info *) CMD_SP(sc);
>> + WARN_ON_ONCE(!rqi);
>> +
>> + if (!rqi) {
>> + atomic64_inc(&snic->s_stats.io.req_null);
>> + spin_unlock_irqrestore(io_lock, flags);
>> + CMD_FLAGS(sc) |= SNIC_IO_ABTS_TERM_REQ_NULL;
>> + SNIC_HOST_ERR(snic->shost,
>> + "reset_cmpl: rqi is null,Hdr stat %s Tag 0x%x sc 0x%p flags
>>0x%llx\n",
>> + snic_io_status_to_str(hdr_stat), cmnd_id, sc,
>> + CMD_FLAGS(sc));
>> +
>> + ret = 1;
>> +
>> + return ret;
>> + }
>> + /* stats */
>> + spin_unlock_irqrestore(io_lock, flags);
>> +
>> + /* scsi cleanup */
>> + snic_hba_reset_scsi_cleanup(snic, sc);
>> +
>> + SNIC_BUG_ON(snic_get_state(snic) != SNIC_OFFLINE &&
>> + snic_get_state(snic) != SNIC_FWRESET);
>> +
>> + /* Careful locking between snic_lock and io lock */
>> + spin_lock_irqsave(io_lock, flags);
>> + spin_lock_irqsave(&snic->snic_lock, gflags);
>> + if (snic_get_state(snic) == SNIC_FWRESET)
>> + snic_set_state(snic, SNIC_ONLINE);
>> + spin_unlock_irqrestore(&snic->snic_lock, gflags);
>> +
>> + if (snic->remove_wait)
>> + complete(snic->remove_wait);
>> +
>> + spin_unlock_irqrestore(io_lock, flags);
>> + atomic64_inc(&snic->s_stats.reset.hba_reset_cmpl);
>> +
>> + ret = 0;
>> + /* Rediscovery is for SAN */
>> + if (snic->config.xpt_type == SNIC_DAS)
>> + return ret;
>> +
>> + SNIC_SCSI_DBG(snic->shost, "reset_cmpl: Queuing discovery work.\n");
>> + queue_work(snic_glob->event_q, &snic->disc_work);
>> +
>> + return ret;
>> +}
>So the hba reset relies on the firmware to process the queues?
>Hmm. But if the firmware is hosed it wouldn't process commands, right?
>I would have really, _really_, designed an out-of-band mechanism
>for doing an HBA reset.
>Having the HBA reset to rely on functionality which only is established
>during startup is ... dangerous, to say the least.
>We've been through this with the various RAID controller.
>Eventually everyone of them implemented an out-of-band firmware reset...
We will use the out-of-band mechanism in a future-patch
>
>> +
>> +static void
>> +snic_msg_ack_handler(struct snic *snic, struct snic_fw_req *fwreq)
>> +{
>> + SNIC_HOST_INFO(snic->shost, "Message Ack Received.\n");
>> +
>> + SNIC_ASSERT_NOT_IMPL(1);
>> +}
>> +
>> +static void
>> +snic_aen_handler(struct snic *snic, struct snic_fw_req *fwreq)
>> +{
>> + u8 typ, hdr_stat;
>> + u32 cmnd_id, hid;
>> + u64 ctx;
>> + struct snic_async_evnotify *aen = &fwreq->u.async_ev;
>> +
>> + snic_io_hdr_dec(&fwreq->hdr, &typ, &hdr_stat, &cmnd_id, &hid, &ctx);
>> + SNIC_SCSI_DBG(snic->shost,
>> + "aen: type = %x, hdr_stat = %x, cmnd_id = %x, hid = %x, ctx =
>>%llx\n",
>> + typ, hdr_stat, cmnd_id, hid, ctx);
>> +
>> + switch (aen->ev_id) {
>> + case SNIC_EV_TGT_OFFLINE:
>> + SNIC_HOST_INFO(snic->shost, "aen:TGT_OFFLINE Event Recvd.\n");
>> + break;
>> +
>> + case SNIC_EV_TGT_ONLINE:
>> + SNIC_HOST_INFO(snic->shost, "aen:TGT_ONLINE Event Recvd.\n");
>> + break;
>> +
>> + case SNIC_EV_LUN_OFFLINE:
>> + SNIC_HOST_INFO(snic->shost, "aen:LUN_OFFLINE Event Recvd.\n");
>> + break;
>> +
>> + case SNIC_EV_LUN_ONLINE:
>> + SNIC_HOST_INFO(snic->shost, "aen:LUN_ONLINE Event Recvd.\n");
>> + break;
>> +
>> + case SNIC_EV_CONF_CHG:
>> + SNIC_HOST_INFO(snic->shost, "aen:Config Change Event Recvd.\n");
>> + break;
>> +
>> + case SNIC_EV_TGT_ADDED:
>> + SNIC_HOST_INFO(snic->shost, "aen:TGT_ADD Event Recvd.\n");
>> + break;
>> +
>> + case SNIC_EV_TGT_DELTD:
>> + SNIC_HOST_INFO(snic->shost, "aen:TGT_DEL Event Recvd.\n");
>> + break;
>> +
>> + case SNIC_EV_LUN_ADDED:
>> + SNIC_HOST_INFO(snic->shost, "aen:LUN_ADD Event Recvd.\n");
>> + break;
>> +
>> + case SNIC_EV_LUN_DELTD:
>> + SNIC_HOST_INFO(snic->shost, "aen:LUN_DEL Event Recvd.\n");
>> + break;
>> +
>> + case SNIC_EV_DISC_CMPL:
>> + SNIC_HOST_INFO(snic->shost, "aen:DISC_CMPL Event Recvd.\n");
>> + break;
>> +
>> + default:
>> + SNIC_HOST_INFO(snic->shost, "aen:Unknown Event Recvd.\n");
>> + SNIC_BUG_ON(1);
>> + break;
>> + }
>> +
>> + SNIC_ASSERT_NOT_IMPL(1);
>> +} /* end of snic_aen_handler */
>> +
>> +/*
>> + * snic_io_cmpl_handler
>> + * Routine to process CQ entries(IO Completions) posted by fw.
>> + */
>> +static int
>> +snic_io_cmpl_handler(struct vnic_dev *vdev,
>> + unsigned int cq_idx,
>> + struct snic_fw_req *fwreq)
>> +{
>> + struct snic *snic = vnic_dev_priv(vdev);
>> + u64 start = jiffies, cmpl_time;
>> +
>> + snic_print_desc(__func__, (char *)fwreq, sizeof(*fwreq));
>> +
>> + /* Update FW Stats */
>> + if ((fwreq->hdr.type >= SNIC_RSP_REPORT_TGTS_CMPL) &&
>> + (fwreq->hdr.type <= SNIC_RSP_BOOT_LUNS_CMPL))
>> + atomic64_dec(&snic->s_stats.fw.actv_reqs);
>> +
>> + SNIC_BUG_ON((fwreq->hdr.type > SNIC_RSP_BOOT_LUNS_CMPL) &&
>> + (fwreq->hdr.type < SNIC_MSG_ASYNC_EVNOTIFY));
>> +
>> + /* Check for snic subsys errors */
>> + switch (fwreq->hdr.status) {
>> + case SNIC_STAT_NOT_READY: /* XPT yet to initialize */
>> + SNIC_HOST_ERR(snic->shost,
>> + "sNIC SubSystem is NOT Ready.\n");
>> + break;
>> +
>> + case SNIC_STAT_FATAL_ERROR: /* XPT Error */
>> + SNIC_HOST_ERR(snic->shost,
>> + "sNIC SubSystem in Unrecoverable State.\n");
>> + break;
>> + }
>> +
>> + switch (fwreq->hdr.type) {
>> + case SNIC_RSP_EXCH_VER_CMPL:
>> + snic_io_exch_ver_cmpl_handler(snic, fwreq);
>> + break;
>> +
>> + case SNIC_RSP_REPORT_TGTS_CMPL:
>> + snic_report_tgt_cmpl_handler(snic, fwreq);
>> + break;
>> +
>> + case SNIC_RSP_ICMND_CMPL:
>> + snic_icmnd_cmpl_handler(snic, fwreq);
>> + break;
>> +
>> + case SNIC_RSP_ITMF_CMPL:
>> + snic_itmf_cmpl_handler(snic, fwreq);
>> + break;
>> +
>> + case SNIC_RSP_HBA_RESET_CMPL:
>> + snic_hba_reset_cmpl_handler(snic, fwreq);
>> + break;
>> +
>> + case SNIC_MSG_ACK:
>> + snic_msg_ack_handler(snic, fwreq);
>> + break;
>> +
>> + case SNIC_MSG_ASYNC_EVNOTIFY:
>> + snic_aen_handler(snic, fwreq);
>> + break;
>> +
>> + default:
>> + SNIC_BUG_ON(1);
>> + SNIC_SCSI_DBG(snic->shost,
>> + "Unknown Firmwqre completion request type %d\n",
>> + fwreq->hdr.type);
>> + break;
>> + }
>> +
>> + /* Update Stats */
>> + cmpl_time = jiffies - start;
>> + if (cmpl_time > atomic64_read(&snic->s_stats.io.max_cmpl_time))
>> + atomic64_set(&snic->s_stats.io.max_cmpl_time, cmpl_time);
>> +
>> + return 0;
>> +} /* end of snic_io_cmpl_handler */
>> +
>> +/*
>> + * snic_fwcq_cmpl_handler
>> + * Routine to process fwCQ
>> + * This CQ is independent, and not associated with wq/rq/wq_copy queues
>> + */
>> +int
>> +snic_fwcq_cmpl_handler(struct snic *snic, int io_cmpl_work)
>> +{
>> + unsigned int num_ent = 0; /* number cq entries processed */
>> + unsigned int cq_idx;
>> + unsigned int nent_per_cq;
>> + struct snic_misc_stats *misc_stats = &snic->s_stats.misc;
>> +
>> + for (cq_idx = snic->wq_count; cq_idx < snic->cq_count; cq_idx++) {
>> + nent_per_cq = vnic_cq_fw_service(&snic->cq[cq_idx],
>> + snic_io_cmpl_handler,
>> + io_cmpl_work);
>> + num_ent += nent_per_cq;
>> +
>> + if (nent_per_cq > atomic64_read(&misc_stats->max_cq_ents))
>> + atomic64_set(&misc_stats->max_cq_ents, nent_per_cq);
>> + }
>> +
>> + return num_ent;
>> +} /* end of snic_fwcq_cmpl_handler */
>> +
>> +/*
>> + * snic_queue_itmf_req: Common API to queue Task Management requests.
>> + * Use rqi->tm_tag for passing special tags.
>> + * @req_id : aborted request's tag, -1 for lun reset.
>> + */
>> +static int
>> +snic_queue_itmf_req(struct snic *snic,
>> + struct snic_host_req *tmreq,
>> + struct scsi_cmnd *sc,
>> + u32 tmf,
>> + u32 req_id)
>> +{
>> + struct snic_req_info *rqi = req_to_rqi(tmreq);
>> + struct scsi_lun lun;
>> + int tm_tag = snic_cmd_tag(sc) | rqi->tm_tag;
>> + int ret = 0;
>> +
>> + SNIC_BUG_ON(!rqi);
>> + SNIC_BUG_ON(!rqi->tm_tag);
>> +
>> + /* fill in lun info */
>> + int_to_scsilun(sc->device->lun, &lun);
>> +
>> + /* Initialize snic_host_req: itmf */
>> + snic_itmf_init(tmreq,
>> + tm_tag,
>> + snic->config.hid,
>> + (u64) rqi,
>> + 0 /* flags */,
>> + req_id, /* Command to be aborted. */
>> + rqi->tgt_id,
>> + lun.scsi_lun,
>> + tmf);
>> +
>> + /*
>> + * In case of multiple aborts on same cmd,
>> + * use try_wait_for_completion and completion_done() to check
>> + * whether it queues aborts even after completion of abort issued
>> + * prior.SNIC_BUG_ON(completion_done(&rqi->done));
>> + */
>> +
>> + ret = snic_queue_wq_desc(snic, tmreq, sizeof(*tmreq));
>> + if (ret)
>> + SNIC_HOST_ERR(snic->shost,
>> + "qitmf:Queuing ITMF(%d) Req sc %p, rqi %p, req_id %d tag %d
>>Failed, ret = %d\n",
>> + tmf, sc, rqi, req_id, snic_cmd_tag(sc), ret);
>> + else
>> + SNIC_SCSI_DBG(snic->shost,
>> + "qitmf:Queuing ITMF(%d) Req sc %p, rqi %p, req_id %d, tag %d
>>(req_id)- Success.",
>> + tmf, sc, rqi, req_id, snic_cmd_tag(sc));
>> +
>> + return ret;
>> +} /* end of snic_queue_itmf_req */
>> +
>> +static int
>> +snic_issue_tm_req(struct snic *snic,
>> + struct snic_req_info *rqi,
>> + struct scsi_cmnd *sc,
>> + int tmf)
>> +{
>> + struct snic_host_req *tmreq = NULL;
>> + int req_id = 0, tag = snic_cmd_tag(sc);
>> + int ret = 0;
>> +
>> + if (snic_get_state(snic) == SNIC_FWRESET)
>> + return -EBUSY;
>> +
>> + atomic_inc(&snic->ios_inflight);
>> +
>> + SNIC_SCSI_DBG(snic->shost,
>> + "issu_tmreq: Task mgmt req %d. rqi %p w/ tag %x\n",
>> + tmf, rqi, tag);
>> +
>> +
>> + if (tmf == SNIC_ITMF_LUN_RESET) {
>> + tmreq = snic_dr_req_init(snic, rqi);
>> + req_id = SCSI_NO_TAG;
>> + } else {
>> + tmreq = snic_abort_req_init(snic, rqi);
>> + req_id = tag;
>> + }
>> +
>> + if (!tmreq) {
>> + ret = -ENOMEM;
>> +
>> + goto tmreq_err;
>> + }
>> +
>> + ret = snic_queue_itmf_req(snic, tmreq, sc, tmf, req_id);
>> + if (ret)
>> + goto tmreq_err;
>> +
>> + ret = 0;
>> +
>> +tmreq_err:
>> + if (ret) {
>> + SNIC_HOST_ERR(snic->shost,
>> + "issu_tmreq: Queing ITMF(%d) Req, sc %p rqi %p req_id %d tag
>>%x fails err = %d\n",
>> + tmf, sc, rqi, req_id, tag, ret);
>> + } else {
>> + SNIC_SCSI_DBG(snic->shost,
>> + "issu_tmreq: Queuing ITMF(%d) Req, sc %p, rqi %p, req_id %d
>>tag %x - Success.\n",
>> + tmf, sc, rqi, req_id, tag);
>> + }
>> +
>> + atomic_dec(&snic->ios_inflight);
>> +
>> + return ret;
>> +}
>> +
>> +/*
>> + * snic_queue_abort_req : Queues abort req to WQ
>> + */
>> +static int
>> +snic_queue_abort_req(struct snic *snic,
>> + struct snic_req_info *rqi,
>> + struct scsi_cmnd *sc,
>> + int tmf)
>> +{
>> + SNIC_SCSI_DBG(snic->shost, "q_abtreq: sc %p, rqi %p, tag %x, tmf
>>%d\n",
>> + sc, rqi, snic_cmd_tag(sc), tmf);
>> +
>> + /* Add special tag for abort */
>> + rqi->tm_tag |= SNIC_TAG_ABORT;
>> +
>> + return snic_issue_tm_req(snic, rqi, sc, tmf);
>> +}
>> +
>> +/*
>> + * snic_abort_finish : called by snic_abort_cmd on queuing abort
>>successfully.
>> + */
>> +static int
>> +snic_abort_finish(struct snic *snic, struct scsi_cmnd *sc)
>> +{
>> + struct snic_req_info *rqi = NULL;
>> + spinlock_t *io_lock = NULL;
>> + unsigned long flags;
>> + int ret = 0, tag = snic_cmd_tag(sc);
>> +
>> + io_lock = snic_io_lock_hash(snic, sc);
>> + spin_lock_irqsave(io_lock, flags);
>> + rqi = (struct snic_req_info *) CMD_SP(sc);
>> + if (!rqi) {
>> + atomic64_inc(&snic->s_stats.io.req_null);
>> + CMD_FLAGS(sc) |= SNIC_IO_ABTS_TERM_REQ_NULL;
>> +
>> + SNIC_SCSI_DBG(snic->shost,
>> + "abt_fini:req info is null tag 0x%x, sc 0x%p flags 0x%llx\n",
>> + tag, sc, CMD_FLAGS(sc));
>> + ret = FAILED;
>> +
>> + goto abort_fail;
>> + }
>> +
>> + rqi->abts_done = NULL;
>> +
>> + ret = FAILED;
>> +
>> + /* Check the abort status. */
>> + switch (CMD_ABTS_STATUS(sc)) {
>> + case SNIC_INVALID_CODE:
>> + /* Firmware didn't complete abort req, timedout */
>> + CMD_FLAGS(sc) |= SNIC_IO_ABTS_TIMEDOUT;
>> + atomic64_inc(&snic->s_stats.abts.drv_tmo);
>> + SNIC_SCSI_DBG(snic->shost,
>> + "abt_fini:sc %p Tag %x Driver Timeout.flags 0x%llx\n",
>> + sc, snic_cmd_tag(sc), CMD_FLAGS(sc));
>> + /* do not release snic request in timedout case */
>> + rqi = NULL;
>> +
>> + goto abort_fail;
>> +
>> + case SNIC_STAT_IO_SUCCESS:
>> + case SNIC_STAT_IO_NOT_FOUND:
>> + ret = SUCCESS;
>> + break;
>> +
>> + default:
>> + /* Firmware completed abort with error */
>> + ret = FAILED;
>> + break;
>> + }
>> +
>> + CMD_SP(sc) = NULL;
>> + SNIC_HOST_INFO(snic->shost,
>> + "abt_fini: Tag %x, Cmpl Status %s flags 0x%llx\n",
>> + tag, snic_io_status_to_str(CMD_ABTS_STATUS(sc)),
>> + CMD_FLAGS(sc));
>> +
>> +abort_fail:
>> + spin_unlock_irqrestore(io_lock, flags);
>> + if (rqi)
>> + snic_release_req_buf(snic, rqi, sc);
>> +
>> + return ret;
>> +} /* end of snic_abort_finish */
>> +
>> +/*
>> + * snic_send_abort_and_wait : Issues Abort, and Waits
>> + */
>> +static int
>> +snic_send_abort_and_wait(struct snic *snic, struct scsi_cmnd *sc)
>> +{
>> + struct snic_req_info *rqi = NULL;
>> + enum snic_ioreq_state sv_state;
>> + struct snic_tgt *tgt = NULL;
>> + spinlock_t *io_lock = NULL;
>> + DECLARE_COMPLETION_ONSTACK(tm_done);
>> + unsigned long flags;
>> + int ret = 0, tmf = 0, tag = snic_cmd_tag(sc);
>> +
>> + tgt = starget_to_tgt(scsi_target(sc->device));
>> + if ((snic_tgt_chkready(tgt) != 0) && (tgt->tdata.typ == SNIC_TGT_SAN))
>> + tmf = SNIC_ITMF_ABTS_TASK_TERM;
>> + else
>> + tmf = SNIC_ITMF_ABTS_TASK;
>> +
>> + /* stats */
>> +
>> + io_lock = snic_io_lock_hash(snic, sc);
>> +
>> + /*
>> + * Avoid a race between SCSI issuing the abort and the device
>> + * completing the command.
>> + *
>> + * If the command is already completed by fw_cmpl code,
>> + * we just return SUCCESS from here. This means that the abort
>> + * succeeded. In the SCSI ML, since the timeout for command has
>> + * happend, the completion wont actually complete the command
>> + * and it will be considered as an aborted command
>> + *
>> + * The CMD_SP will not be cleared except while holding io_lock
>> + */
>> + spin_lock_irqsave(io_lock, flags);
>> + rqi = (struct snic_req_info *) CMD_SP(sc);
>> + if (!rqi) {
>> + spin_unlock_irqrestore(io_lock, flags);
>> +
>> + SNIC_HOST_ERR(snic->shost,
>> + "abt_cmd: rqi is null. Tag %d flags 0x%llx\n",
>> + tag, CMD_FLAGS(sc));
>> +
>> + ret = SUCCESS;
>> +
>> + goto send_abts_end;
>> + }
>> +
>> + rqi->abts_done = &tm_done;
>> + if (CMD_STATE(sc) == SNIC_IOREQ_ABTS_PENDING) {
>> + spin_unlock_irqrestore(io_lock, flags);
>> +
>> + ret = 0;
>> + goto abts_pending;
>> + }
>> + SNIC_BUG_ON(!rqi->abts_done);
>> +
>> + /* Save Command State, should be restored on failed to Queue. */
>> + sv_state = CMD_STATE(sc);
>> +
>> + /*
>> + * Command is still pending, need to abort it
>> + * If the fw completes the command after this point,
>> + * the completion won't be done till mid-layer, since abot
>> + * has already started.
>> + */
>> + CMD_STATE(sc) = SNIC_IOREQ_ABTS_PENDING;
>> + CMD_ABTS_STATUS(sc) = SNIC_INVALID_CODE;
>> +
>> + SNIC_SCSI_DBG(snic->shost, "send_abt_cmd: TAG 0x%x\n", tag);
>> +
>> + spin_unlock_irqrestore(io_lock, flags);
>> +
>> + /* Now Queue the abort command to firmware */
>> + ret = snic_queue_abort_req(snic, rqi, sc, tmf);
>> + if (ret) {
>> + SNIC_HOST_ERR(snic->shost,
>> + "send_abt_cmd: IO w/ Tag 0x%x fail w/ err %d flags 0x%llx\n",
>> + tag, ret, CMD_FLAGS(sc));
>> +
>> + spin_lock_irqsave(io_lock, flags);
>> + /* Restore Command's previous state */
>> + CMD_STATE(sc) = sv_state;
>> + rqi = (struct snic_req_info *) CMD_SP(sc);
>> + if (rqi)
>> + rqi->abts_done = NULL;
>> + spin_unlock_irqrestore(io_lock, flags);
>> + ret = FAILED;
>> +
>> + goto send_abts_end;
>> + }
>> +
>> + spin_lock_irqsave(io_lock, flags);
>> + if (tmf == SNIC_ITMF_ABTS_TASK) {
>> + CMD_FLAGS(sc) |= SNIC_IO_ABTS_ISSUED;
>> + atomic64_inc(&snic->s_stats.abts.num);
>> + } else {
>> + /* term stats */
>> + CMD_FLAGS(sc) |= SNIC_IO_TERM_ISSUED;
>> + }
>> + spin_unlock_irqrestore(io_lock, flags);
>> +
>> + SNIC_SCSI_DBG(snic->shost,
>> + "send_abt_cmd: sc %p Tag %x flags 0x%llx\n",
>> + sc, tag, CMD_FLAGS(sc));
>> +
>> +
>> + ret = 0;
>> +
>> +abts_pending:
>> + /*
>> + * Queued an abort IO, wait for its completion.
>> + * Once the fw completes the abort command, it will
>> + * wakeup this thread.
>> + */
>> + wait_for_completion_timeout(&tm_done, SNIC_ABTS_TIMEOUT);
>> +
>> +send_abts_end:
>> + return ret;
>> +} /* end of snic_send_abort_and_wait */
>> +
>> +/*
>> + * This function is exported to SCSI for sending abort cmnds.
>> + * A SCSI IO is represent by snic_ioreq in the driver.
>> + * The snic_ioreq is linked to the SCSI Cmd, thus a link with the
>>ULP'S IO
>> + */
>> +int
>> +snic_abort_cmd(struct scsi_cmnd *sc)
>> +{
>> + struct snic *snic = shost_priv(sc->device->host);
>> + int ret = SUCCESS, tag = snic_cmd_tag(sc);
>> + u32 start_time = jiffies;
>> +
>> + SNIC_SCSI_DBG(snic->shost, "abt_cmd:sc %p :0x%x :req = %p :tag =
>>%d\n",
>> + sc, sc->cmnd[0], sc->request, tag);
>> +
>> + if (unlikely(snic_get_state(snic) != SNIC_ONLINE)) {
>> + SNIC_HOST_ERR(snic->shost,
>> + "abt_cmd: tag %x Parent Devs are not rdy\n",
>> + tag);
>> + ret = FAST_IO_FAIL;
>> +
>> + goto abort_end;
>> + }
>> +
>> +
>> + ret = snic_send_abort_and_wait(snic, sc);
>> + if (ret)
>> + goto abort_end;
>> +
>> + ret = snic_abort_finish(snic, sc);
>> +
>> +abort_end:
>> + SNIC_TRC(snic->shost->host_no, tag, sc,
>> + jiffies_to_msecs(jiffies - start_time), 0,
>> + SNIC_TRC_CMD(sc), SNIC_TRC_CMD_STATE_FLAGS(sc));
>> +
>> + SNIC_SCSI_DBG(snic->shost,
>> + "abts: Abort Req Status = %s\n",
>> + (ret == SUCCESS) ? "SUCCESS" :
>> + ((ret == FAST_IO_FAIL) ? "FAST_IO_FAIL" : "FAILED"));
>> +
>> + return ret;
>> +}
>> +
>> +
>> +
>> +static int
>> +snic_is_abts_pending(struct snic *snic, struct scsi_cmnd *lr_sc)
>> +{
>> + struct snic_req_info *rqi = NULL;
>> + struct scsi_cmnd *sc = NULL;
>> + struct scsi_device *lr_sdev = NULL;
>> + spinlock_t *io_lock = NULL;
>> + u32 tag;
>> + unsigned long flags;
>> +
>> + if (lr_sc)
>> + lr_sdev = lr_sc->device;
>> +
>> + /* walk through the tag map, an dcheck if IOs are still pending in
>>fw*/
>> + for (tag = 0; tag < snic->max_tag_id; tag++) {
>> + io_lock = snic_io_lock_tag(snic, tag);
>> +
>> + spin_lock_irqsave(io_lock, flags);
>> + sc = scsi_host_find_tag(snic->shost, tag);
>> +
>> + if (!sc || (lr_sc && (sc->device != lr_sdev || sc == lr_sc))) {
>> + spin_unlock_irqrestore(io_lock, flags);
>> +
>> + continue;
>> + }
>> +
>> + rqi = (struct snic_req_info *) CMD_SP(sc);
>> + if (!rqi) {
>> + spin_unlock_irqrestore(io_lock, flags);
>> +
>> + continue;
>> + }
>> +
>> + /*
>> + * Found IO that is still pending w/ firmware and belongs to
>> + * the LUN that is under reset, if lr_sc != NULL
>> + */
>> + SNIC_SCSI_DBG(snic->shost, "Found IO in %s on LUN\n",
>> + snic_ioreq_state_to_str(CMD_STATE(sc)));
>> +
>> + if (CMD_STATE(sc) == SNIC_IOREQ_ABTS_PENDING) {
>> + spin_unlock_irqrestore(io_lock, flags);
>> +
>> + return 1;
>> + }
>> +
>> + spin_unlock_irqrestore(io_lock, flags);
>> + }
>> +
>> + return 0;
>> +} /* end of snic_is_abts_pending */
>> +
>> +static int
>> +snic_dr_clean_single_req(struct snic *snic,
>> + u32 tag,
>> + struct scsi_device *lr_sdev)
>> +{
>> + struct snic_req_info *rqi = NULL;
>> + struct snic_tgt *tgt = NULL;
>> + struct scsi_cmnd *sc = NULL;
>> + spinlock_t *io_lock = NULL;
>> + u32 sv_state = 0, tmf = 0;
>> + DECLARE_COMPLETION_ONSTACK(tm_done);
>> + unsigned long flags;
>> + int ret = 0;
>> +
>> + io_lock = snic_io_lock_tag(snic, tag);
>> + spin_lock_irqsave(io_lock, flags);
>> + sc = scsi_host_find_tag(snic->shost, tag);
>> +
>> + /* Ignore Cmd that don't belong to Lun Reset device */
>> + if (!sc || sc->device != lr_sdev)
>> + goto skip_clean;
>> +
>> + rqi = (struct snic_req_info *) CMD_SP(sc);
>> +
>> + if (!rqi)
>> + goto skip_clean;
>> +
>> +
>> + if (CMD_STATE(sc) == SNIC_IOREQ_ABTS_PENDING)
>> + goto skip_clean;
>> +
>> +
>> + if ((CMD_FLAGS(sc) & SNIC_DEVICE_RESET) &&
>> + (!(CMD_FLAGS(sc) & SNIC_DEV_RST_ISSUED))) {
>> +
>> + SNIC_SCSI_DBG(snic->shost,
>> + "clean_single_req: devrst is not pending sc 0x%p\n",
>> + sc);
>> +
>> + goto skip_clean;
>> + }
>> +
>> + SNIC_SCSI_DBG(snic->shost,
>> + "clean_single_req: Found IO in %s on lun\n",
>> + snic_ioreq_state_to_str(CMD_STATE(sc)));
>> +
>> + /* Save Command State */
>> + sv_state = CMD_STATE(sc);
>> +
>> + /*
>> + * Any pending IO issued prior to reset is expected to be
>> + * in abts pending state, if not we need to set
>>SNIC_IOREQ_ABTS_PENDING
>> + * to indicate the IO is abort pending.
>> + * When IO is completed, the IO will be handed over and handled
>> + * in this function.
>> + */
>> +
>> + CMD_STATE(sc) = SNIC_IOREQ_ABTS_PENDING;
>> + SNIC_BUG_ON(rqi->abts_done);
>> +
>> + if (CMD_FLAGS(sc) & SNIC_DEVICE_RESET) {
>> + rqi->tm_tag = SNIC_TAG_DEV_RST;
>> +
>> + SNIC_SCSI_DBG(snic->shost,
>> + "clean_single_req:devrst sc 0x%p\n", sc);
>> + }
>> +
>> + CMD_ABTS_STATUS(sc) = SNIC_INVALID_CODE;
>> + rqi->abts_done = &tm_done;
>> + spin_unlock_irqrestore(io_lock, flags);
>> +
>> + tgt = starget_to_tgt(scsi_target(sc->device));
>> + if ((snic_tgt_chkready(tgt) != 0) && (tgt->tdata.typ == SNIC_TGT_SAN))
>> + tmf = SNIC_ITMF_ABTS_TASK_TERM;
>> + else
>> + tmf = SNIC_ITMF_ABTS_TASK;
>> +
>> + /* Now queue the abort command to firmware */
>> + ret = snic_queue_abort_req(snic, rqi, sc, tmf);
>> + if (ret) {
>> + SNIC_HOST_ERR(snic->shost,
>> + "clean_single_req_err:sc %p, tag %d abt failed. tm_tag %d
>>flags 0x%llx\n",
>> + sc, tag, rqi->tm_tag, CMD_FLAGS(sc));
>> +
>> + spin_lock_irqsave(io_lock, flags);
>> + rqi = (struct snic_req_info *) CMD_SP(sc);
>> + if (rqi)
>> + rqi->abts_done = NULL;
>> +
>> + /* Restore Command State */
>> + if (CMD_STATE(sc) == SNIC_IOREQ_ABTS_PENDING)
>> + CMD_STATE(sc) = sv_state;
>> +
>> + ret = 1;
>> + goto skip_clean;
>> + }
>> +
>> + spin_lock_irqsave(io_lock, flags);
>> + if (CMD_FLAGS(sc) & SNIC_DEVICE_RESET)
>> + CMD_FLAGS(sc) |= SNIC_DEV_RST_TERM_ISSUED;
>> +
>> + CMD_FLAGS(sc) |= SNIC_IO_INTERNAL_TERM_ISSUED;
>> + spin_unlock_irqrestore(io_lock, flags);
>> +
>> + wait_for_completion_timeout(&tm_done, SNIC_ABTS_TIMEOUT);
>> +
>> + /* Recheck cmd state to check if it now aborted. */
>> + spin_lock_irqsave(io_lock, flags);
>> + rqi = (struct snic_req_info *) CMD_SP(sc);
>> + if (!rqi) {
>> + CMD_FLAGS(sc) |= SNIC_IO_ABTS_TERM_REQ_NULL;
>> + goto skip_clean;
>> + }
>> + rqi->abts_done = NULL;
>> +
>> + /* if abort is still pending w/ fw, fail */
>> + if (CMD_ABTS_STATUS(sc) == SNIC_INVALID_CODE) {
>> + SNIC_HOST_ERR(snic->shost,
>> + "clean_single_req_err:sc %p tag %d abt still pending w/ fw,
>>tm_tag %d flags 0x%llx\n",
>> + sc, tag, rqi->tm_tag, CMD_FLAGS(sc));
>> +
>> + CMD_FLAGS(sc) |= SNIC_IO_ABTS_TERM_DONE;
>> + ret = 1;
>> +
>> + goto skip_clean;
>> + }
>> +
>> + CMD_STATE(sc) = SNIC_IOREQ_ABTS_COMPLETE;
>> + CMD_SP(sc) = NULL;
>> + spin_unlock_irqrestore(io_lock, flags);
>> +
>> + snic_release_req_buf(snic, rqi, sc);
>> +
>> + ret = 0;
>> +
>> + return ret;
>> +
>> +skip_clean:
>> + spin_unlock_irqrestore(io_lock, flags);
>> +
>> + return ret;
>> +} /* end of snic_dr_clean_single_req */
>> +
>> +static int
>> +snic_dr_clean_pending_req(struct snic *snic, struct scsi_cmnd *lr_sc)
>> +{
>> + struct scsi_device *lr_sdev = lr_sc->device;
>> + u32 tag = 0;
>> + int ret = FAILED;
>> +
>> + for (tag = 0; tag < snic->max_tag_id; tag++) {
>> + if (tag == snic_cmd_tag(lr_sc))
>> + continue;
>> +
>> + ret = snic_dr_clean_single_req(snic, tag, lr_sdev);
>> + if (ret) {
>> + SNIC_HOST_ERR(snic->shost, "clean_err:tag = %d\n", tag);
>> +
>> + goto clean_err;
>> + }
>> + }
>> +
>> + schedule_timeout(msecs_to_jiffies(100));
>> +
>> + /* Walk through all the cmds and check abts status. */
>> + if (snic_is_abts_pending(snic, lr_sc)) {
>> + ret = FAILED;
>> +
>> + goto clean_err;
>> + }
>> +
>> + ret = 0;
>> + SNIC_SCSI_DBG(snic->shost, "clean_pending_req: Success.\n");
>> +
>> + return ret;
>> +
>> +clean_err:
>> + ret = FAILED;
>> + SNIC_HOST_ERR(snic->shost,
>> + "Failed to Clean Pending IOs on %s device.\n",
>> + dev_name(&lr_sdev->sdev_gendev));
>> +
>> + return ret;
>> +
>> +} /* end of snic_dr_clean_pending_req */
>> +
>> +/*
>> + * snic_dr_finish : Called by snic_device_reset
>> + */
>> +static int
>> +snic_dr_finish(struct snic *snic, struct scsi_cmnd *sc)
>> +{
>> + struct snic_req_info *rqi = NULL;
>> + spinlock_t *io_lock = NULL;
>> + unsigned long flags;
>> + int lr_res = 0;
>> + int ret = FAILED;
>> +
>> + io_lock = snic_io_lock_hash(snic, sc);
>> + spin_lock_irqsave(io_lock, flags);
>> + rqi = (struct snic_req_info *) CMD_SP(sc);
>> + if (!rqi) {
>> + spin_unlock_irqrestore(io_lock, flags);
>> + SNIC_SCSI_DBG(snic->shost,
>> + "dr_fini: rqi is null tag 0x%x sc 0x%p flags 0x%llx\n",
>> + snic_cmd_tag(sc), sc, CMD_FLAGS(sc));
>> +
>> + ret = FAILED;
>> + goto dr_fini_end;
>> + }
>> +
>> + rqi->dr_done = NULL;
>> +
>> + lr_res = CMD_LR_STATUS(sc);
>> +
>> + switch (lr_res) {
>> + case SNIC_INVALID_CODE:
>> + /* stats */
>> + SNIC_SCSI_DBG(snic->shost,
>> + "dr_fini: Tag %x Dev Reset Timedout. flags 0x%llx\n",
>> + snic_cmd_tag(sc), CMD_FLAGS(sc));
>> +
>> + CMD_FLAGS(sc) |= SNIC_DEV_RST_TIMEDOUT;
>> + ret = FAILED;
>> +
>> + goto dr_failed;
>> +
>> + case SNIC_STAT_IO_SUCCESS:
>> + SNIC_SCSI_DBG(snic->shost,
>> + "dr_fini: Tag %x Dev Reset cmpl\n",
>> + snic_cmd_tag(sc));
>> + ret = 0;
>> + break;
>> +
>> + default:
>> + SNIC_HOST_ERR(snic->shost,
>> + "dr_fini:Device Reset completed& failed.Tag = %x lr_status %s
>>flags 0x%llx\n",
>> + snic_cmd_tag(sc),
>> + snic_io_status_to_str(lr_res), CMD_FLAGS(sc));
>> + ret = FAILED;
>> + goto dr_failed;
>> + }
>> + spin_unlock_irqrestore(io_lock, flags);
>> +
>> + /*
>> + * Cleanup any IOs on this LUN that have still not completed.
>> + * If any of these fail, then LUN Reset fails.
>> + * Cleanup cleans all commands on this LUN except
>> + * the lun reset command. If all cmds get cleaned, the LUN Reset
>> + * succeeds.
>> + */
>> +
>> + ret = snic_dr_clean_pending_req(snic, sc);
>> + if (ret) {
>> + spin_lock_irqsave(io_lock, flags);
>> + SNIC_SCSI_DBG(snic->shost,
>> + "dr_fini: Device Reset Failed since could not abort all IOs.
>>Tag = %x.\n",
>> + snic_cmd_tag(sc));
>> + rqi = (struct snic_req_info *) CMD_SP(sc);
>> +
>> + goto dr_failed;
>> + } else {
>> + /* Cleanup LUN Reset Command */
>> + spin_lock_irqsave(io_lock, flags);
>> + rqi = (struct snic_req_info *) CMD_SP(sc);
>> + if (rqi)
>> + ret = SUCCESS; /* Completed Successfully */
>> + else
>> + ret = FAILED;
>> + }
>> +
>> +dr_failed:
>> + SNIC_BUG_ON(!spin_is_locked(io_lock));
>> + if (rqi)
>> + CMD_SP(sc) = NULL;
>> + spin_unlock_irqrestore(io_lock, flags);
>> +
>> + if (rqi)
>> + snic_release_req_buf(snic, rqi, sc);
>> +
>> +dr_fini_end:
>> + return ret;
>> +} /* end of snic_dr_finish */
>> +
>> +static int
>> +snic_queue_dr_req(struct snic *snic,
>> + struct snic_req_info *rqi,
>> + struct scsi_cmnd *sc)
>> +{
>> + /* Add special tag for device reset */
>> + rqi->tm_tag |= SNIC_TAG_DEV_RST;
>> +
>> + return snic_issue_tm_req(snic, rqi, sc, SNIC_ITMF_LUN_RESET);
>> +}
>> +
>> +static int
>> +snic_send_dr_and_wait(struct snic *snic, struct scsi_cmnd *sc)
>> +{
>> + struct snic_req_info *rqi = NULL;
>> + enum snic_ioreq_state sv_state;
>> + spinlock_t *io_lock = NULL;
>> + unsigned long flags;
>> + DECLARE_COMPLETION_ONSTACK(tm_done);
>> + int ret = FAILED, tag = snic_cmd_tag(sc);
>> +
>> + io_lock = snic_io_lock_hash(snic, sc);
>> + spin_lock_irqsave(io_lock, flags);
>> + CMD_FLAGS(sc) |= SNIC_DEVICE_RESET;
>> + rqi = (struct snic_req_info *) CMD_SP(sc);
>> + if (!rqi) {
>> + SNIC_HOST_ERR(snic->shost,
>> + "send_dr: rqi is null, Tag 0x%x flags 0x%llx\n",
>> + tag, CMD_FLAGS(sc));
>> + spin_unlock_irqrestore(io_lock, flags);
>> +
>> + ret = FAILED;
>> + goto send_dr_end;
>> + }
>> +
>> + /* Save Command state to restore in case Queuing failed. */
>> + sv_state = CMD_STATE(sc);
>> +
>> + CMD_STATE(sc) = SNIC_IOREQ_LR_PENDING;
>> + CMD_LR_STATUS(sc) = SNIC_INVALID_CODE;
>> +
>> + SNIC_SCSI_DBG(snic->shost, "dr: TAG = %x\n", tag);
>> +
>> + rqi->dr_done = &tm_done;
>> + SNIC_BUG_ON(!rqi->dr_done);
>> +
>> + spin_unlock_irqrestore(io_lock, flags);
>> + /*
>> + * The Command state is changed to IOREQ_PENDING,
>> + * in this case, if the command is completed, the icmnd_cmpl will
>> + * mark the cmd as completed.
>> + * This logic still makes LUN Reset is inevitable.
>> + */
>> +
>> + ret = snic_queue_dr_req(snic, rqi, sc);
>> + if (ret) {
>> + SNIC_HOST_ERR(snic->shost,
>> + "send_dr: IO w/ Tag 0x%x Failed err = %d. flags 0x%llx\n",
>> + tag, ret, CMD_FLAGS(sc));
>> +
>> + spin_lock_irqsave(io_lock, flags);
>> + /* Restore State */
>> + CMD_STATE(sc) = sv_state;
>> + rqi = (struct snic_req_info *) CMD_SP(sc);
>> + if (rqi)
>> + rqi->dr_done = NULL;
>> + /* rqi is freed in caller. */
>> + spin_unlock_irqrestore(io_lock, flags);
>> + ret = FAILED;
>> +
>> + goto send_dr_end;
>> + }
>> +
>> + spin_lock_irqsave(io_lock, flags);
>> + CMD_FLAGS(sc) |= SNIC_DEV_RST_ISSUED;
>> + spin_unlock_irqrestore(io_lock, flags);
>> +
>> + ret = 0;
>> +
>> + wait_for_completion_timeout(&tm_done, SNIC_LUN_RESET_TIMEOUT);
>> +
>> +send_dr_end:
>> + return ret;
>> +}
>> +
>> +/*
>> + * auxillary funciton to check lun reset op is supported or not
>> + * Not supported if returns 0
>> + */
>> +static int
>> +snic_dev_reset_supported(struct scsi_device *sdev)
>> +{
>> + struct snic_tgt *tgt = starget_to_tgt(scsi_target(sdev));
>> +
>> + if (tgt->tdata.typ == SNIC_TGT_DAS)
>> + return 0;
>> +
>> + return 1;
>> +}
>> +
>> +static void
>> +snic_unlink_and_release_req(struct snic *snic, struct scsi_cmnd *sc,
>>int flag)
>> +{
>> + struct snic_req_info *rqi = NULL;
>> + spinlock_t *io_lock = NULL;
>> + unsigned long flags;
>> + u32 start_time = jiffies;
>> +
>> + io_lock = snic_io_lock_hash(snic, sc);
>> + spin_lock_irqsave(io_lock, flags);
>> + rqi = (struct snic_req_info *) CMD_SP(sc);
>> + if (rqi) {
>> + start_time = rqi->start_time;
>> + CMD_SP(sc) = NULL;
>> + }
>> +
>> + CMD_FLAGS(sc) |= flag;
>> + spin_unlock_irqrestore(io_lock, flags);
>> +
>> + if (rqi)
>> + snic_release_req_buf(snic, rqi, sc);
>> +
>> + SNIC_TRC(snic->shost->host_no, snic_cmd_tag(sc), sc,
>> + jiffies_to_msecs(jiffies - start_time),
>> + rqi, SNIC_TRC_CMD(sc), SNIC_TRC_CMD_STATE_FLAGS(sc));
>> +}
>> +
>> +/*
>> + * SCSI Eh thread issues a LUN Reset when one or more commands on a LUN
>> + * fail to get aborted. It calls driver's eh_device_reset with a SCSI
>> + * command on the LUN.
>> + */
>> +int
>> +snic_device_reset(struct scsi_cmnd *sc)
>> +{
>> + struct Scsi_Host *shost = sc->device->host;
>> + struct snic *snic = shost_priv(shost);
>> + struct snic_req_info *rqi = NULL;
>> + int tag = snic_cmd_tag(sc);
>> + int start_time = jiffies;
>> + int ret = FAILED;
>> + int dr_supp = 0;
>> +
>> + SNIC_SCSI_DBG(shost, "dev_reset:sc %p :0x%x :req = %p :tag = %d\n",
>> + sc, sc->cmnd[0], sc->request,
>> + snic_cmd_tag(sc));
>> + dr_supp = snic_dev_reset_supported(sc->device);
>> + if (!dr_supp) {
>> + /* device reset op is not supported */
>> + SNIC_HOST_INFO(shost, "LUN Reset Op not supported.\n");
>> + snic_unlink_and_release_req(snic, sc, SNIC_DEV_RST_NOTSUP);
>> +
>> + goto dev_rst_end;
>> + }
>> +
>> + if (unlikely(snic_get_state(snic) != SNIC_ONLINE)) {
>> + snic_unlink_and_release_req(snic, sc, 0);
>> + SNIC_HOST_ERR(shost, "Devrst: Parent Devs are not online.\n");
>> +
>> + goto dev_rst_end;
>> + }
>> +
>> + /* There is no tag when lun reset is issue through ioctl. */
>> + if (unlikely(tag <= SNIC_NO_TAG)) {
>> + SNIC_HOST_INFO(snic->shost,
>> + "Devrst: LUN Reset Recvd thru IOCTL.\n");
>> +
>> + rqi = snic_req_init(snic, 0);
>> + if (!rqi)
>> + goto dev_rst_end;
>> +
>> + memset(scsi_cmd_priv(sc), 0,
>> + sizeof(struct snic_internal_io_state));
>> + CMD_SP(sc) = (char *)rqi;
>> + CMD_FLAGS(sc) = SNIC_NO_FLAGS;
>> +
>> + /* Add special tag for dr coming from user spc */
>> + rqi->tm_tag = SNIC_TAG_IOCTL_DEV_RST;
>> + rqi->sc = sc;
>> + }
>> +
>> + ret = snic_send_dr_and_wait(snic, sc);
>> + if (ret) {
>> + SNIC_HOST_ERR(snic->shost,
>> + "Devrst: IO w/ Tag %x Failed w/ err = %d\n",
>> + tag, ret);
>> +
>> + snic_unlink_and_release_req(snic, sc, 0);
>> +
>> + goto dev_rst_end;
>> + }
>> +
>> + ret = snic_dr_finish(snic, sc);
>> +
>> +dev_rst_end:
>> + SNIC_TRC(snic->shost->host_no, tag, sc,
>> + jiffies_to_msecs(jiffies - start_time),
>> + 0, SNIC_TRC_CMD(sc), SNIC_TRC_CMD_STATE_FLAGS(sc));
>> +
>> + SNIC_SCSI_DBG(snic->shost,
>> + "Devrst: Returning from Device Reset : %s\n",
>> + (ret == SUCCESS) ? "SUCCESS" : "FAILED");
>> +
>> + return ret;
>> +} /* end of snic_device_reset */
>> +
>> +/*
>> + * SCSI Error handling calls driver's eh_host_reset if all prior
>> + * error handling levels return FAILED.
>> + *
>> + * Host Reset is the highest level of error recovery. If this fails,
>>then
>> + * host is offlined by SCSI.
>> + */
>> +/*
>> + * snic_issue_hba_reset : Queues FW Reset Request.
>> + */
>> +static int
>> +snic_issue_hba_reset(struct snic *snic, struct scsi_cmnd *sc)
>> +{
>> + struct snic_req_info *rqi = NULL;
>> + struct snic_host_req *req = NULL;
>> + spinlock_t *io_lock = NULL;
>> + DECLARE_COMPLETION_ONSTACK(wait);
>> + unsigned long flags;
>> + int ret = -ENOMEM;
>> +
>> + rqi = snic_req_init(snic, 0);
>> + if (!rqi) {
>> + ret = -ENOMEM;
>> +
>> + goto hba_rst_end;
>> + }
>> +
>> + if (snic_cmd_tag(sc) == SCSI_NO_TAG) {
>> + memset(scsi_cmd_priv(sc), 0,
>> + sizeof(struct snic_internal_io_state));
>> + SNIC_HOST_INFO(snic->shost, "issu_hr:Host reset thru ioctl.\n");
>> + rqi->sc = sc;
>> + }
>> +
>> + req = rqi_to_req(rqi);
>> +
>> + io_lock = snic_io_lock_hash(snic, sc);
>> + spin_lock_irqsave(io_lock, flags);
>> + SNIC_BUG_ON(CMD_SP(sc) != NULL);
>> + CMD_STATE(sc) = SNIC_IOREQ_PENDING;
>> + CMD_SP(sc) = (char *) rqi;
>> + CMD_FLAGS(sc) |= SNIC_IO_INITIALIZED;
>> + snic->remove_wait = &wait;
>> + spin_unlock_irqrestore(io_lock, flags);
>> +
>> + /* Initialize Request */
>> + snic_io_hdr_enc(&req->hdr, SNIC_REQ_HBA_RESET, 0, snic_cmd_tag(sc),
>> + snic->config.hid, 0, (u64)rqi);
>> +
>> + req->u.reset.flags = 0;
>> +
>> + ret = snic_queue_wq_desc(snic, req, sizeof(*req));
>> + if (ret) {
>> + SNIC_HOST_ERR(snic->shost,
>> + "issu_hr:Queuing HBA Reset Failed. w err %d\n",
>> + ret);
>> +
>> + goto hba_rst_err;
>> + }
>> +
>> + spin_lock_irqsave(io_lock, flags);
>> + CMD_FLAGS(sc) |= SNIC_HOST_RESET_ISSUED;
>> + spin_unlock_irqrestore(io_lock, flags);
>> + atomic64_inc(&snic->s_stats.reset.hba_resets);
>> + SNIC_HOST_INFO(snic->shost, "Queued HBA Reset Successfully.\n");
>> +
>> + wait_for_completion_timeout(snic->remove_wait,
>> + SNIC_HOST_RESET_TIMEOUT);
>> +
>> + if (snic_get_state(snic) == SNIC_FWRESET) {
>> + SNIC_HOST_ERR(snic->shost, "reset_cmpl: Reset Timedout.\n");
>> + ret = -ETIMEDOUT;
>> +
>> + goto hba_rst_err;
>> + }
>> +
>> + spin_lock_irqsave(io_lock, flags);
>> + snic->remove_wait = NULL;
>> + rqi = (struct snic_req_info *) CMD_SP(sc);
>> + CMD_SP(sc) = NULL;
>> + spin_unlock_irqrestore(io_lock, flags);
>> +
>> + if (rqi)
>> + snic_req_free(snic, rqi);
>> +
>> + ret = 0;
>> +
>> + return ret;
>> +
>> +hba_rst_err:
>> + spin_lock_irqsave(io_lock, flags);
>> + snic->remove_wait = NULL;
>> + rqi = (struct snic_req_info *) CMD_SP(sc);
>> + CMD_SP(sc) = NULL;
>> + spin_unlock_irqrestore(io_lock, flags);
>> +
>> + if (rqi)
>> + snic_req_free(snic, rqi);
>> +
>> +hba_rst_end:
>> + SNIC_HOST_ERR(snic->shost,
>> + "reset:HBA Reset Failed w/ err = %d.\n",
>> + ret);
>> +
>> + return ret;
>> +} /* end of snic_issue_hba_reset */
>> +
>> +int
>> +snic_reset(struct Scsi_Host *shost, struct scsi_cmnd *sc)
>> +{
>> + struct snic *snic = shost_priv(shost);
>> + enum snic_state sv_state;
>> + unsigned long flags;
>> + int ret = FAILED;
>> +
>> + /* Set snic state as SNIC_FWRESET*/
>> + sv_state = snic_get_state(snic);
>> +
>> + spin_lock_irqsave(&snic->snic_lock, flags);
>> + if (snic_get_state(snic) == SNIC_FWRESET) {
>> + spin_unlock_irqrestore(&snic->snic_lock, flags);
>> + SNIC_HOST_INFO(shost, "reset:prev reset is in progres\n");
>> +
>> + msleep(SNIC_HOST_RESET_TIMEOUT);
>> + ret = SUCCESS;
>> +
>> + goto reset_end;
>> + }
>> +
>> + snic_set_state(snic, SNIC_FWRESET);
>> + spin_unlock_irqrestore(&snic->snic_lock, flags);
>> +
>> +
>> + /* Wait for all the IOs that are entered in Qcmd */
>> + while (atomic_read(&snic->ios_inflight))
>> + schedule_timeout(msecs_to_jiffies(1));
>> +
>> + ret = snic_issue_hba_reset(snic, sc);
>> + if (ret) {
>> + SNIC_HOST_ERR(shost,
>> + "reset:Host Reset Failed w/ err %d.\n",
>> + ret);
>> + spin_lock_irqsave(&snic->snic_lock, flags);
>> + snic_set_state(snic, sv_state);
>> + spin_unlock_irqrestore(&snic->snic_lock, flags);
>> + atomic64_inc(&snic->s_stats.reset.hba_reset_fail);
>> + ret = FAILED;
>> +
>> + goto reset_end;
>> + }
>> +
>> + ret = SUCCESS;
>> +
>> +reset_end:
>> + return ret;
>> +} /* end of snic_reset */
>> +
>> +/*
>> + * SCSI Error handling calls driver's eh_host_reset if all prior
>> + * error handling levels return FAILED.
>> + *
>> + * Host Reset is the highest level of error recovery. If this fails,
>>then
>> + * host is offlined by SCSI.
>> + */
>> +int
>> +snic_host_reset(struct scsi_cmnd *sc)
>> +{
>> + struct Scsi_Host *shost = sc->device->host;
>> + u32 start_time = jiffies;
>> + int ret = FAILED;
>> +
>> + SNIC_SCSI_DBG(shost,
>> + "host reset:sc %p sc_cmd 0x%x req %p tag %d flags 0x%llx\n",
>> + sc, sc->cmnd[0], sc->request,
>> + snic_cmd_tag(sc), CMD_FLAGS(sc));
>> +
>> + ret = snic_reset(shost, sc);
>> +
>> + SNIC_TRC(shost->host_no, snic_cmd_tag(sc), sc,
>> + jiffies_to_msecs(jiffies - start_time),
>> + 0, SNIC_TRC_CMD(sc), SNIC_TRC_CMD_STATE_FLAGS(sc));
>> +
>> + return ret;
>> +} /* end of snic_host_reset */
>> +
>As mentioned above:
>The send/receive queues are established during device initialization.
>So if this information get garbled for whatever reason or if the
>firmware becomes hosed you've got no way of resetting the HBA.
>I would rather see if there wouldn't be an out-of-be mechanism to reset
>the HBA (via PCI reset, say), and then send a new initialisation command.
>This way you can be sure to reset the HBA correctly.
We will use the out-of-band mechanism in a future-patch.
>
>Cheers,
>
>Hannes
>--
>Dr. Hannes Reinecke zSeries & Storage
>hare@suse.de +49 911 74053 688
>SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg
>GF: J. Hawn, J. Guild, F. Imendörffer, HRB 16746 (AG Nürnberg)
Thanks
Narsimhulu
>
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH v4 6/9] snic:Add low level queuing interfaces
2015-04-09 13:23 ` Hannes Reinecke
@ 2015-04-13 5:40 ` Narsimhulu Musini (nmusini)
0 siblings, 0 replies; 24+ messages in thread
From: Narsimhulu Musini (nmusini) @ 2015-04-13 5:40 UTC (permalink / raw)
To: Hannes Reinecke, JBottomley, linux-scsi, hch; +Cc: Sesidhar Baddela (sebaddel)
Hi Hannes,
Thank you for reviewing patches. Please find responses inline.
I will incorporate the comments and suggestions in next patch submittal.
On 09/04/15 6:53 pm, "Hannes Reinecke" <hare@suse.de> wrote:
>Hi Narsimhulu,
>
>please find my comment at the bottom.
>
>On 04/09/2015 01:49 PM, Narsimhulu Musini wrote:
>> These files contain low level queueing interfaces includes
>> hardware queues, and management of hardware features.
>>
>> Signed-off-by: Narsimhulu Musini <nmusini@cisco.com>
>> Signed-off-by: Sesidhar Baddela <sebaddel@cisco.com>
>> ---
>> * v3
>> - Cleaned up unused functions.
>>
>> * v2
>> - driver supports x86-64 arch, so removed cpu_to_XX API to maintain
>>consistency.
>>
>> drivers/scsi/snic/cq_desc.h | 76 ++++
>> drivers/scsi/snic/cq_enet_desc.h | 38 ++
>> drivers/scsi/snic/vnic_cq.c | 86 +++++
>> drivers/scsi/snic/vnic_cq.h | 120 ++++++
>> drivers/scsi/snic/vnic_cq_fw.h | 62 ++++
>> drivers/scsi/snic/vnic_dev.c | 749
>>++++++++++++++++++++++++++++++++++++++
>> drivers/scsi/snic/vnic_dev.h | 140 +++++++
>> drivers/scsi/snic/vnic_devcmd.h | 270 ++++++++++++++
>> drivers/scsi/snic/vnic_intr.c | 59 +++
>> drivers/scsi/snic/vnic_intr.h | 119 ++++++
>> drivers/scsi/snic/vnic_resource.h | 68 ++++
>> drivers/scsi/snic/vnic_snic.h | 54 +++
>> drivers/scsi/snic/vnic_stats.h | 68 ++++
>> drivers/scsi/snic/vnic_wq.c | 236 ++++++++++++
>> drivers/scsi/snic/vnic_wq.h | 187 ++++++++++
>> drivers/scsi/snic/wq_enet_desc.h | 91 +++++
>> 16 files changed, 2423 insertions(+)
>> create mode 100644 drivers/scsi/snic/cq_desc.h
>> create mode 100644 drivers/scsi/snic/cq_enet_desc.h
>> create mode 100644 drivers/scsi/snic/vnic_cq.c
>> create mode 100644 drivers/scsi/snic/vnic_cq.h
>> create mode 100644 drivers/scsi/snic/vnic_cq_fw.h
>> create mode 100644 drivers/scsi/snic/vnic_dev.c
>> create mode 100644 drivers/scsi/snic/vnic_dev.h
>> create mode 100644 drivers/scsi/snic/vnic_devcmd.h
>> create mode 100644 drivers/scsi/snic/vnic_intr.c
>> create mode 100644 drivers/scsi/snic/vnic_intr.h
>> create mode 100644 drivers/scsi/snic/vnic_resource.h
>> create mode 100644 drivers/scsi/snic/vnic_snic.h
>> create mode 100644 drivers/scsi/snic/vnic_stats.h
>> create mode 100644 drivers/scsi/snic/vnic_wq.c
>> create mode 100644 drivers/scsi/snic/vnic_wq.h
>> create mode 100644 drivers/scsi/snic/wq_enet_desc.h
>>
>> diff --git a/drivers/scsi/snic/cq_desc.h b/drivers/scsi/snic/cq_desc.h
>> new file mode 100644
>> index 0000000..630edfa
>> --- /dev/null
>> +++ b/drivers/scsi/snic/cq_desc.h
>> @@ -0,0 +1,76 @@
>> +/*
>> + * Copyright 2014 Cisco Systems, Inc. All rights reserved.
>> + *
>> + * This program is free software; you may redistribute it and/or modify
>> + * it under the terms of the GNU General Public License as published by
>> + * the Free Software Foundation; version 2 of the License.
>> + *
>> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
>> + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
>> + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
>> + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
>> + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
>> + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
>> + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
>> + * SOFTWARE.
>> + */
>> +
>> +#ifndef _CQ_DESC_H_
>> +#define _CQ_DESC_H_
>> +
>> +/*
>> + * Completion queue descriptor types
>> + */
>> +enum cq_desc_types {
>> + CQ_DESC_TYPE_WQ_ENET = 0,
>> + CQ_DESC_TYPE_DESC_COPY = 1,
>> + CQ_DESC_TYPE_WQ_EXCH = 2,
>> + CQ_DESC_TYPE_RQ_ENET = 3,
>> + CQ_DESC_TYPE_RQ_FCP = 4,
>> +};
>> +
>> +/* Completion queue descriptor: 16B
>> + *
>> + * All completion queues have this basic layout. The
>> + * type_specific area is unique for each completion
>> + * queue type.
>> + */
>> +struct cq_desc {
>> + u16 completed_index;
>> + u16 q_number;
>> + u8 type_specific[11];
>> + u8 type_color;
>> +};
>> +
>> +#define CQ_DESC_TYPE_BITS 4
>> +#define CQ_DESC_TYPE_MASK ((1 << CQ_DESC_TYPE_BITS) - 1)
>> +#define CQ_DESC_COLOR_MASK 1
>> +#define CQ_DESC_COLOR_SHIFT 7
>> +#define CQ_DESC_Q_NUM_BITS 10
>> +#define CQ_DESC_Q_NUM_MASK ((1 << CQ_DESC_Q_NUM_BITS) - 1)
>> +#define CQ_DESC_COMP_NDX_BITS 12
>> +#define CQ_DESC_COMP_NDX_MASK ((1 << CQ_DESC_COMP_NDX_BITS) - 1)
>> +
>> +static inline void cq_desc_dec(const struct cq_desc *desc_arg,
>> + u8 *type, u8 *color, u16 *q_number, u16 *completed_index)
>> +{
>> + const struct cq_desc *desc = desc_arg;
>> + const u8 type_color = desc->type_color;
>> +
>> + *color = (type_color >> CQ_DESC_COLOR_SHIFT) & CQ_DESC_COLOR_MASK;
>> +
>> + /*
>> + * Make sure color bit is read from desc *before* other fields
>> + * are read from desc. Hardware guarantees color bit is last
>> + * bit (byte) written. Adding the rmb() prevents the compiler
>> + * and/or CPU from reordering the reads which would potentially
>> + * result in reading stale values.
>> + */
>> + rmb();
>> +
>> + *type = type_color & CQ_DESC_TYPE_MASK;
>> + *q_number = desc->q_number & CQ_DESC_Q_NUM_MASK;
>> + *completed_index = desc->completed_index & CQ_DESC_COMP_NDX_MASK;
>> +}
>> +
>> +#endif /* _CQ_DESC_H_ */
>> diff --git a/drivers/scsi/snic/cq_enet_desc.h
>>b/drivers/scsi/snic/cq_enet_desc.h
>> new file mode 100644
>> index 0000000..99ecd20
>> --- /dev/null
>> +++ b/drivers/scsi/snic/cq_enet_desc.h
>> @@ -0,0 +1,38 @@
>> +/*
>> + * Copyright 2014 Cisco Systems, Inc. All rights reserved.
>> + *
>> + * This program is free software; you may redistribute it and/or modify
>> + * it under the terms of the GNU General Public License as published by
>> + * the Free Software Foundation; version 2 of the License.
>> + *
>> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
>> + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
>> + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
>> + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
>> + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
>> + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
>> + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
>> + * SOFTWARE.
>> + */
>> +
>> +#ifndef _CQ_ENET_DESC_H_
>> +#define _CQ_ENET_DESC_H_
>> +
>> +#include "cq_desc.h"
>> +
>> +/* Ethernet completion queue descriptor: 16B */
>> +struct cq_enet_wq_desc {
>> + u16 completed_index;
>> + u16 q_number;
>> + u8 reserved[11];
>> + u8 type_color;
>> +};
>> +
>> +static inline void cq_enet_wq_desc_dec(struct cq_enet_wq_desc *desc,
>> + u8 *type, u8 *color, u16 *q_number, u16 *completed_index)
>> +{
>> + cq_desc_dec((struct cq_desc *)desc, type,
>> + color, q_number, completed_index);
>> +}
>> +
>> +#endif /* _CQ_ENET_DESC_H_ */
>> diff --git a/drivers/scsi/snic/vnic_cq.c b/drivers/scsi/snic/vnic_cq.c
>> new file mode 100644
>> index 0000000..88d4537
>> --- /dev/null
>> +++ b/drivers/scsi/snic/vnic_cq.c
>> @@ -0,0 +1,86 @@
>> +/*
>> + * Copyright 2014 Cisco Systems, Inc. All rights reserved.
>> + *
>> + * This program is free software; you may redistribute it and/or modify
>> + * it under the terms of the GNU General Public License as published by
>> + * the Free Software Foundation; version 2 of the License.
>> + *
>> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
>> + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
>> + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
>> + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
>> + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
>> + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
>> + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
>> + * SOFTWARE.
>> + */
>> +
>> +#include <linux/errno.h>
>> +#include <linux/types.h>
>> +#include <linux/pci.h>
>> +#include "vnic_dev.h"
>> +#include "vnic_cq.h"
>> +
>> +void vnic_cq_free(struct vnic_cq *cq)
>> +{
>> + vnic_dev_free_desc_ring(cq->vdev, &cq->ring);
>> +
>> + cq->ctrl = NULL;
>> +}
>> +
>> +int vnic_cq_alloc(struct vnic_dev *vdev, struct vnic_cq *cq, unsigned
>>int index,
>> + unsigned int desc_count, unsigned int desc_size)
>> +{
>> + int err;
>> +
>> + cq->index = index;
>> + cq->vdev = vdev;
>> +
>> + cq->ctrl = vnic_dev_get_res(vdev, RES_TYPE_CQ, index);
>> + if (!cq->ctrl) {
>> + pr_err("Failed to hook CQ[%d] resource\n", index);
>> +
>> + return -EINVAL;
>> + }
>> +
>> + err = vnic_dev_alloc_desc_ring(vdev, &cq->ring, desc_count,
>>desc_size);
>> + if (err)
>> + return err;
>> +
>> + return 0;
>> +}
>> +
>> +void vnic_cq_init(struct vnic_cq *cq, unsigned int flow_control_enable,
>> + unsigned int color_enable, unsigned int cq_head, unsigned int cq_tail,
>> + unsigned int cq_tail_color, unsigned int interrupt_enable,
>> + unsigned int cq_entry_enable, unsigned int cq_message_enable,
>> + unsigned int interrupt_offset, u64 cq_message_addr)
>> +{
>> + u64 paddr;
>> +
>> + paddr = (u64)cq->ring.base_addr | VNIC_PADDR_TARGET;
>> + writeq(paddr, &cq->ctrl->ring_base);
>> + iowrite32(cq->ring.desc_count, &cq->ctrl->ring_size);
>> + iowrite32(flow_control_enable, &cq->ctrl->flow_control_enable);
>> + iowrite32(color_enable, &cq->ctrl->color_enable);
>> + iowrite32(cq_head, &cq->ctrl->cq_head);
>> + iowrite32(cq_tail, &cq->ctrl->cq_tail);
>> + iowrite32(cq_tail_color, &cq->ctrl->cq_tail_color);
>> + iowrite32(interrupt_enable, &cq->ctrl->interrupt_enable);
>> + iowrite32(cq_entry_enable, &cq->ctrl->cq_entry_enable);
>> + iowrite32(cq_message_enable, &cq->ctrl->cq_message_enable);
>> + iowrite32(interrupt_offset, &cq->ctrl->interrupt_offset);
>> + writeq(cq_message_addr, &cq->ctrl->cq_message_addr);
>> +}
>> +
>> +void vnic_cq_clean(struct vnic_cq *cq)
>> +{
>> + cq->to_clean = 0;
>> + cq->last_color = 0;
>> +
>> + iowrite32(0, &cq->ctrl->cq_head);
>> + iowrite32(0, &cq->ctrl->cq_tail);
>> + iowrite32(1, &cq->ctrl->cq_tail_color);
>> +
>> + vnic_dev_clear_desc_ring(&cq->ring);
>> +}
>> diff --git a/drivers/scsi/snic/vnic_cq.h b/drivers/scsi/snic/vnic_cq.h
>> new file mode 100644
>> index 0000000..fb2dc61
>> --- /dev/null
>> +++ b/drivers/scsi/snic/vnic_cq.h
>> @@ -0,0 +1,120 @@
>> +/*
>> + * Copyright 2014 Cisco Systems, Inc. All rights reserved.
>> + *
>> + * This program is free software; you may redistribute it and/or modify
>> + * it under the terms of the GNU General Public License as published by
>> + * the Free Software Foundation; version 2 of the License.
>> + *
>> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
>> + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
>> + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
>> + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
>> + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
>> + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
>> + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
>> + * SOFTWARE.
>> + */
>> +
>> +#ifndef _VNIC_CQ_H_
>> +#define _VNIC_CQ_H_
>> +
>> +#include "cq_desc.h"
>> +#include "vnic_dev.h"
>> +
>> +/*
>> + * These defines avoid symbol clash between fnic and enic (Cisco 10G
>>Eth
>> + * Driver) when both are built with CONFIG options =y
>> + */
>> +#define vnic_cq_service snic_cq_service
>> +#define vnic_cq_free snic_cq_free
>> +#define vnic_cq_alloc snic_cq_alloc
>> +#define vnic_cq_init snic_cq_init
>> +#define vnic_cq_clean snic_cq_clean
>> +
>> +/* Completion queue control */
>> +struct vnic_cq_ctrl {
>> + u64 ring_base; /* 0x00 */
>> + u32 ring_size; /* 0x08 */
>> + u32 pad0;
>> + u32 flow_control_enable; /* 0x10 */
>> + u32 pad1;
>> + u32 color_enable; /* 0x18 */
>> + u32 pad2;
>> + u32 cq_head; /* 0x20 */
>> + u32 pad3;
>> + u32 cq_tail; /* 0x28 */
>> + u32 pad4;
>> + u32 cq_tail_color; /* 0x30 */
>> + u32 pad5;
>> + u32 interrupt_enable; /* 0x38 */
>> + u32 pad6;
>> + u32 cq_entry_enable; /* 0x40 */
>> + u32 pad7;
>> + u32 cq_message_enable; /* 0x48 */
>> + u32 pad8;
>> + u32 interrupt_offset; /* 0x50 */
>> + u32 pad9;
>> + u64 cq_message_addr; /* 0x58 */
>> + u32 pad10;
>> +};
>> +
>> +struct vnic_cq {
>> + unsigned int index;
>> + struct vnic_dev *vdev;
>> + struct vnic_cq_ctrl __iomem *ctrl; /* memory-mapped */
>> + struct vnic_dev_ring ring;
>> + unsigned int to_clean;
>> + unsigned int last_color;
>> +};
>> +
>> +static inline unsigned int vnic_cq_service(struct vnic_cq *cq,
>> + unsigned int work_to_do,
>> + int (*q_service)(struct vnic_dev *vdev, struct cq_desc *cq_desc,
>> + u8 type, u16 q_number, u16 completed_index, void *opaque),
>> + void *opaque)
>> +{
>> + struct cq_desc *cq_desc;
>> + unsigned int work_done = 0;
>> + u16 q_number, completed_index;
>> + u8 type, color;
>> +
>> + cq_desc = (struct cq_desc *)((u8 *)cq->ring.descs +
>> + cq->ring.desc_size * cq->to_clean);
>> + cq_desc_dec(cq_desc, &type, &color,
>> + &q_number, &completed_index);
>> +
>> + while (color != cq->last_color) {
>> +
>> + if ((*q_service)(cq->vdev, cq_desc, type,
>> + q_number, completed_index, opaque))
>> + break;
>> +
>> + cq->to_clean++;
>> + if (cq->to_clean == cq->ring.desc_count) {
>> + cq->to_clean = 0;
>> + cq->last_color = cq->last_color ? 0 : 1;
>> + }
>> +
>> + cq_desc = (struct cq_desc *)((u8 *)cq->ring.descs +
>> + cq->ring.desc_size * cq->to_clean);
>> + cq_desc_dec(cq_desc, &type, &color,
>> + &q_number, &completed_index);
>> +
>> + work_done++;
>> + if (work_done >= work_to_do)
>> + break;
>> + }
>> +
>> + return work_done;
>> +}
>> +
>> +void vnic_cq_free(struct vnic_cq *cq);
>> +int vnic_cq_alloc(struct vnic_dev *vdev, struct vnic_cq *cq, unsigned
>>int index,
>> + unsigned int desc_count, unsigned int desc_size);
>> +void vnic_cq_init(struct vnic_cq *cq, unsigned int flow_control_enable,
>> + unsigned int color_enable, unsigned int cq_head, unsigned int cq_tail,
>> + unsigned int cq_tail_color, unsigned int interrupt_enable,
>> + unsigned int cq_entry_enable, unsigned int message_enable,
>> + unsigned int interrupt_offset, u64 message_addr);
>> +void vnic_cq_clean(struct vnic_cq *cq);
>> +#endif /* _VNIC_CQ_H_ */
>> diff --git a/drivers/scsi/snic/vnic_cq_fw.h
>>b/drivers/scsi/snic/vnic_cq_fw.h
>> new file mode 100644
>> index 0000000..c2d1bbd
>> --- /dev/null
>> +++ b/drivers/scsi/snic/vnic_cq_fw.h
>> @@ -0,0 +1,62 @@
>> +/*
>> + * Copyright 2014 Cisco Systems, Inc. All rights reserved.
>> + *
>> + * This program is free software; you may redistribute it and/or modify
>> + * it under the terms of the GNU General Public License as published by
>> + * the Free Software Foundation; version 2 of the License.
>> + *
>> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
>> + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
>> + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
>> + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
>> + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
>> + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
>> + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
>> + * SOFTWARE.
>> + */
>> +
>> +#ifndef _VNIC_CQ_FW_H_
>> +#define _VNIC_CQ_FW_H_
>> +
>> +#include "snic_fwint.h"
>> +
>> +static inline unsigned int
>> +vnic_cq_fw_service(struct vnic_cq *cq,
>> + int (*q_service)(struct vnic_dev *vdev,
>> + unsigned int index,
>> + struct snic_fw_req *desc),
>> + unsigned int work_to_do)
>> +
>> +{
>> + struct snic_fw_req *desc;
>> + unsigned int work_done = 0;
>> + u8 color;
>> +
>> + desc = (struct snic_fw_req *)((u8 *)cq->ring.descs +
>> + cq->ring.desc_size * cq->to_clean);
>> + snic_color_dec(desc, &color);
>> +
>> + while (color != cq->last_color) {
>> +
>> + if ((*q_service)(cq->vdev, cq->index, desc))
>> + break;
>> +
>> + cq->to_clean++;
>> + if (cq->to_clean == cq->ring.desc_count) {
>> + cq->to_clean = 0;
>> + cq->last_color = cq->last_color ? 0 : 1;
>> + }
>> +
>> + desc = (struct snic_fw_req *)((u8 *)cq->ring.descs +
>> + cq->ring.desc_size * cq->to_clean);
>> + snic_color_dec(desc, &color);
>> +
>> + work_done++;
>> + if (work_done >= work_to_do)
>> + break;
>> + }
>> +
>> + return work_done;
>> +}
>> +
>> +#endif /* _VNIC_CQ_FW_H_ */
>> diff --git a/drivers/scsi/snic/vnic_dev.c b/drivers/scsi/snic/vnic_dev.c
>> new file mode 100644
>> index 0000000..68c63d5
>> --- /dev/null
>> +++ b/drivers/scsi/snic/vnic_dev.c
>> @@ -0,0 +1,749 @@
>> +/*
>> + * Copyright 2014 Cisco Systems, Inc. All rights reserved.
>> + *
>> + * This program is free software; you may redistribute it and/or modify
>> + * it under the terms of the GNU General Public License as published by
>> + * the Free Software Foundation; version 2 of the License.
>> + *
>> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
>> + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
>> + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
>> + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
>> + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
>> + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
>> + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
>> + * SOFTWARE.
>> + */
>> +
>> +#include <linux/kernel.h>
>> +#include <linux/errno.h>
>> +#include <linux/types.h>
>> +#include <linux/pci.h>
>> +#include <linux/delay.h>
>> +#include <linux/if_ether.h>
>> +#include <linux/slab.h>
>> +#include "vnic_resource.h"
>> +#include "vnic_devcmd.h"
>> +#include "vnic_dev.h"
>> +#include "vnic_stats.h"
>> +#include "vnic_wq.h"
>> +
>> +#define VNIC_DVCMD_TMO 10000 /* Devcmd Timeout value */
>> +#define VNIC_NOTIFY_INTR_MASK 0x0000ffff00000000ULL
>> +
>> +struct devcmd2_controller {
>> + struct vnic_wq_ctrl __iomem *wq_ctrl;
>> + struct vnic_dev_ring results_ring;
>> + struct vnic_wq wq;
>> + struct vnic_devcmd2 *cmd_ring;
>> + struct devcmd2_result *result;
>> + u16 next_result;
>> + u16 result_size;
>> + int color;
>> +};
>> +
>> +struct vnic_res {
>> + void __iomem *vaddr;
>> + unsigned int count;
>> +};
>> +
>> +struct vnic_dev {
>> + void *priv;
>> + struct pci_dev *pdev;
>> + struct vnic_res res[RES_TYPE_MAX];
>> + enum vnic_dev_intr_mode intr_mode;
>> + struct vnic_devcmd __iomem *devcmd;
>> + struct vnic_devcmd_notify *notify;
>> + struct vnic_devcmd_notify notify_copy;
>> + dma_addr_t notify_pa;
>> + u32 *linkstatus;
>> + dma_addr_t linkstatus_pa;
>> + struct vnic_stats *stats;
>> + dma_addr_t stats_pa;
>> + struct vnic_devcmd_fw_info *fw_info;
>> + dma_addr_t fw_info_pa;
>> + u64 args[VNIC_DEVCMD_NARGS];
>> + struct devcmd2_controller *devcmd2;
>> +
>> + int (*devcmd_rtn)(struct vnic_dev *vdev, enum vnic_devcmd_cmd cmd,
>> + int wait);
>> +};
>> +
>> +#define VNIC_MAX_RES_HDR_SIZE \
>> + (sizeof(struct vnic_resource_header) + \
>> + sizeof(struct vnic_resource) * RES_TYPE_MAX)
>> +#define VNIC_RES_STRIDE 128
>> +
>> +void *vnic_dev_priv(struct vnic_dev *vdev)
>> +{
>> + return vdev->priv;
>> +}
>> +
>> +static int vnic_dev_discover_res(struct vnic_dev *vdev,
>> + struct vnic_dev_bar *bar, unsigned int num_bars)
>> +{
>> + struct vnic_resource_header __iomem *rh;
>> + struct vnic_resource __iomem *r;
>> + u8 type;
>> +
>> + if (num_bars == 0)
>> + return -EINVAL;
>> +
>> + if (bar->len < VNIC_MAX_RES_HDR_SIZE) {
>> + pr_err("vNIC BAR0 res hdr length error\n");
>> +
>> + return -EINVAL;
>> + }
>> +
>> + rh = bar->vaddr;
>> + if (!rh) {
>> + pr_err("vNIC BAR0 res hdr not mem-mapped\n");
>> +
>> + return -EINVAL;
>> + }
>> +
>> + if (ioread32(&rh->magic) != VNIC_RES_MAGIC ||
>> + ioread32(&rh->version) != VNIC_RES_VERSION) {
>> + pr_err("vNIC BAR0 res magic/version error exp (%lx/%lx) curr
>>(%x/%x)\n",
>> + VNIC_RES_MAGIC, VNIC_RES_VERSION,
>> + ioread32(&rh->magic), ioread32(&rh->version));
>> +
>> + return -EINVAL;
>> + }
>> +
>> + r = (struct vnic_resource __iomem *)(rh + 1);
>> +
>> + while ((type = ioread8(&r->type)) != RES_TYPE_EOL) {
>> +
>> + u8 bar_num = ioread8(&r->bar);
>> + u32 bar_offset = ioread32(&r->bar_offset);
>> + u32 count = ioread32(&r->count);
>> + u32 len;
>> +
>> + r++;
>> +
>> + if (bar_num >= num_bars)
>> + continue;
>> +
>> + if (!bar[bar_num].len || !bar[bar_num].vaddr)
>> + continue;
>> +
>> + switch (type) {
>> + case RES_TYPE_WQ:
>> + case RES_TYPE_RQ:
>> + case RES_TYPE_CQ:
>> + case RES_TYPE_INTR_CTRL:
>> + /* each count is stride bytes long */
>> + len = count * VNIC_RES_STRIDE;
>> + if (len + bar_offset > bar->len) {
>> + pr_err("vNIC BAR0 resource %d out-of-bounds, offset 0x%x + size
>>0x%x > bar len 0x%lx\n",
>> + type, bar_offset,
>> + len,
>> + bar->len);
>> +
>> + return -EINVAL;
>> + }
>> + break;
>> +
>> + case RES_TYPE_INTR_PBA_LEGACY:
>> + case RES_TYPE_DEVCMD:
>> + case RES_TYPE_DEVCMD2:
>> + len = count;
>> + break;
>> +
>> + default:
>> + continue;
>> + }
>> +
>> + vdev->res[type].count = count;
>> + vdev->res[type].vaddr = (char __iomem *)bar->vaddr + bar_offset;
>> + }
>> +
>> + return 0;
>> +}
>> +
>> +unsigned int vnic_dev_get_res_count(struct vnic_dev *vdev,
>> + enum vnic_res_type type)
>> +{
>> + return vdev->res[type].count;
>> +}
>> +
>> +void __iomem *vnic_dev_get_res(struct vnic_dev *vdev, enum
>>vnic_res_type type,
>> + unsigned int index)
>> +{
>> + if (!vdev->res[type].vaddr)
>> + return NULL;
>> +
>> + switch (type) {
>> + case RES_TYPE_WQ:
>> + case RES_TYPE_RQ:
>> + case RES_TYPE_CQ:
>> + case RES_TYPE_INTR_CTRL:
>> + return (char __iomem *)vdev->res[type].vaddr +
>> + index * VNIC_RES_STRIDE;
>> +
>> + default:
>> + return (char __iomem *)vdev->res[type].vaddr;
>> + }
>> +}
>> +
>> +unsigned int vnic_dev_desc_ring_size(struct vnic_dev_ring *ring,
>> + unsigned int desc_count,
>> + unsigned int desc_size)
>> +{
>> + /* The base address of the desc rings must be 512 byte aligned.
>> + * Descriptor count is aligned to groups of 32 descriptors. A
>> + * count of 0 means the maximum 4096 descriptors. Descriptor
>> + * size is aligned to 16 bytes.
>> + */
>> +
>> + unsigned int count_align = 32;
>> + unsigned int desc_align = 16;
>> +
>> + ring->base_align = 512;
>> +
>> + if (desc_count == 0)
>> + desc_count = 4096;
>> +
>> + ring->desc_count = ALIGN(desc_count, count_align);
>> +
>> + ring->desc_size = ALIGN(desc_size, desc_align);
>> +
>> + ring->size = ring->desc_count * ring->desc_size;
>> + ring->size_unaligned = ring->size + ring->base_align;
>> +
>> + return ring->size_unaligned;
>> +}
>> +
>> +void vnic_dev_clear_desc_ring(struct vnic_dev_ring *ring)
>> +{
>> + memset(ring->descs, 0, ring->size);
>> +}
>> +
>> +int vnic_dev_alloc_desc_ring(struct vnic_dev *vdev, struct
>>vnic_dev_ring *ring,
>> + unsigned int desc_count, unsigned int desc_size)
>> +{
>> + vnic_dev_desc_ring_size(ring, desc_count, desc_size);
>> +
>> + ring->descs_unaligned = pci_alloc_consistent(vdev->pdev,
>> + ring->size_unaligned,
>> + &ring->base_addr_unaligned);
>> +
>> + if (!ring->descs_unaligned) {
>> + pr_err("Failed to allocate ring (size=%d), aborting\n",
>> + (int)ring->size);
>> +
>> + return -ENOMEM;
>> + }
>> +
>> + ring->base_addr = ALIGN(ring->base_addr_unaligned,
>> + ring->base_align);
>> + ring->descs = (u8 *)ring->descs_unaligned +
>> + (ring->base_addr - ring->base_addr_unaligned);
>> +
>> + vnic_dev_clear_desc_ring(ring);
>> +
>> + ring->desc_avail = ring->desc_count - 1;
>> +
>> + return 0;
>> +}
>> +
>> +void vnic_dev_free_desc_ring(struct vnic_dev *vdev, struct
>>vnic_dev_ring *ring)
>> +{
>> + if (ring->descs) {
>> + pci_free_consistent(vdev->pdev,
>> + ring->size_unaligned,
>> + ring->descs_unaligned,
>> + ring->base_addr_unaligned);
>> + ring->descs = NULL;
>> + }
>> +}
>> +
>> +static int _vnic_dev_cmd2(struct vnic_dev *vdev, enum vnic_devcmd_cmd
>>cmd,
>> + int wait)
>> +{
>> + struct devcmd2_controller *dc2c = vdev->devcmd2;
>> + struct devcmd2_result *result = dc2c->result + dc2c->next_result;
>> + unsigned int i;
>> + int delay;
>> + int err;
>> + u32 posted;
>> + u32 new_posted;
>> +
>> + posted = ioread32(&dc2c->wq_ctrl->posted_index);
>> +
>> + if (posted == 0xFFFFFFFF) { /* check for hardware gone */
>> + /* Hardware surprise removal: return error */
>> + return -ENODEV;
>> + }
>> +
>> + new_posted = (posted + 1) % DEVCMD2_RING_SIZE;
>> + dc2c->cmd_ring[posted].cmd = cmd;
>> + dc2c->cmd_ring[posted].flags = 0;
>> +
>> + if ((_CMD_FLAGS(cmd) & _CMD_FLAGS_NOWAIT))
>> + dc2c->cmd_ring[posted].flags |= DEVCMD2_FNORESULT;
>> +
>> + if (_CMD_DIR(cmd) & _CMD_DIR_WRITE) {
>> + for (i = 0; i < VNIC_DEVCMD_NARGS; i++)
>> + dc2c->cmd_ring[posted].args[i] = vdev->args[i];
>> + }
>> + /* Adding write memory barrier prevents compiler and/or CPU
>> + * reordering, thus avoiding descriptor posting before
>> + * descriptor is initialized. Otherwise, hardware can read
>> + * stale descriptor fields.
>> + */
>> + wmb();
>> + iowrite32(new_posted, &dc2c->wq_ctrl->posted_index);
>> +
>> + if (dc2c->cmd_ring[posted].flags & DEVCMD2_FNORESULT)
>> + return 0;
>> +
>> + for (delay = 0; delay < wait; delay++) {
>> + udelay(100);
>> + if (result->color == dc2c->color) {
>> + dc2c->next_result++;
>> + if (dc2c->next_result == dc2c->result_size) {
>> + dc2c->next_result = 0;
>> + dc2c->color = dc2c->color ? 0 : 1;
>> + }
>> + if (result->error) {
>> + err = (int) result->error;
>> + if (err != ERR_ECMDUNKNOWN ||
>> + cmd != CMD_CAPABILITY)
>> + pr_err("Error %d devcmd %d\n",
>> + err, _CMD_N(cmd));
>> +
>> + return err;
>> + }
>> + if (_CMD_DIR(cmd) & _CMD_DIR_READ) {
>> + /*
>> + * Adding the rmb() prevents the compiler
>> + * and/or CPU from reordering the reads which
>> + * would potentially result in reading stale
>> + * values.
>> + */
>> + rmb();
>> + for (i = 0; i < VNIC_DEVCMD_NARGS; i++)
>> + vdev->args[i] = result->results[i];
>> + }
>> +
>> + return 0;
>> + }
>> + }
>> +
>> + pr_err("Timed out devcmd %d\n", _CMD_N(cmd));
>> +
>> + return -ETIMEDOUT;
>> +}
>> +
>> +static int vnic_dev_init_devcmd2(struct vnic_dev *vdev)
>> +{
>> + struct devcmd2_controller *dc2c = NULL;
>> + unsigned int fetch_idx;
>> + int ret;
>> + void __iomem *p;
>> +
>> + if (vdev->devcmd2)
>> + return 0;
>> +
>> + p = vnic_dev_get_res(vdev, RES_TYPE_DEVCMD2, 0);
>> + if (!p)
>> + return -ENODEV;
>> +
>> + dc2c = kzalloc(sizeof(*dc2c), GFP_ATOMIC);
>> + if (!dc2c)
>> + return -ENOMEM;
>> +
>> + vdev->devcmd2 = dc2c;
>> +
>> + dc2c->color = 1;
>> + dc2c->result_size = DEVCMD2_RING_SIZE;
>> +
>> + ret = vnic_wq_devcmd2_alloc(vdev,
>> + &dc2c->wq,
>> + DEVCMD2_RING_SIZE,
>> + DEVCMD2_DESC_SIZE);
>> + if (ret)
>> + goto err_free_devcmd2;
>> +
>> + fetch_idx = ioread32(&dc2c->wq.ctrl->fetch_index);
>> + if (fetch_idx == 0xFFFFFFFF) { /* check for hardware gone */
>> + /* Hardware surprise removal: reset fetch_index */
>> + fetch_idx = 0;
>> + }
>> +
>> + /*
>> + * Don't change fetch_index ever and
>> + * set posted_index same as fetch_index
>> + * when setting up the WQ for devcmd2.
>> + */
>> + vnic_wq_init_start(&dc2c->wq, 0, fetch_idx, fetch_idx, 0, 0);
>> + vnic_wq_enable(&dc2c->wq);
>> + ret = vnic_dev_alloc_desc_ring(vdev,
>> + &dc2c->results_ring,
>> + DEVCMD2_RING_SIZE,
>> + DEVCMD2_DESC_SIZE);
>> + if (ret)
>> + goto err_free_wq;
>> +
>> + dc2c->result = (struct devcmd2_result *) dc2c->results_ring.descs;
>> + dc2c->cmd_ring = (struct vnic_devcmd2 *) dc2c->wq.ring.descs;
>> + dc2c->wq_ctrl = dc2c->wq.ctrl;
>> + vdev->args[0] = (u64) dc2c->results_ring.base_addr |
>>VNIC_PADDR_TARGET;
>> + vdev->args[1] = DEVCMD2_RING_SIZE;
>> +
>> + ret = _vnic_dev_cmd2(vdev, CMD_INITIALIZE_DEVCMD2, VNIC_DVCMD_TMO);
>> + if (ret < 0)
>> + goto err_free_desc_ring;
>> +
>> + vdev->devcmd_rtn = &_vnic_dev_cmd2;
>> + pr_info("DEVCMD2 Initialized.\n");
>> +
>> + return ret;
>> +
>> +err_free_desc_ring:
>> + vnic_dev_free_desc_ring(vdev, &dc2c->results_ring);
>> +
>> +err_free_wq:
>> + vnic_wq_disable(&dc2c->wq);
>> + vnic_wq_free(&dc2c->wq);
>> +
>> +err_free_devcmd2:
>> + kfree(dc2c);
>> + vdev->devcmd2 = NULL;
>> +
>> + return ret;
>> +} /* end of vnic_dev_init_devcmd2 */
>> +
>> +static void vnic_dev_deinit_devcmd2(struct vnic_dev *vdev)
>> +{
>> + struct devcmd2_controller *dc2c = vdev->devcmd2;
>> +
>> + vdev->devcmd2 = NULL;
>> + vdev->devcmd_rtn = NULL;
>> +
>> + vnic_dev_free_desc_ring(vdev, &dc2c->results_ring);
>> + vnic_wq_disable(&dc2c->wq);
>> + vnic_wq_free(&dc2c->wq);
>> + kfree(dc2c);
>> +}
>> +
>> +int vnic_dev_cmd(struct vnic_dev *vdev, enum vnic_devcmd_cmd cmd,
>> + u64 *a0, u64 *a1, int wait)
>> +{
>> + int err;
>> +
>> + memset(vdev->args, 0, sizeof(vdev->args));
>> + vdev->args[0] = *a0;
>> + vdev->args[1] = *a1;
>> +
>> + err = (*vdev->devcmd_rtn)(vdev, cmd, wait);
>> +
>> + *a0 = vdev->args[0];
>> + *a1 = vdev->args[1];
>> +
>> + return err;
>> +}
>> +
>> +int vnic_dev_fw_info(struct vnic_dev *vdev,
>> + struct vnic_devcmd_fw_info **fw_info)
>> +{
>> + u64 a0, a1 = 0;
>> + int wait = VNIC_DVCMD_TMO;
>> + int err = 0;
>> +
>> + if (!vdev->fw_info) {
>> + vdev->fw_info = pci_alloc_consistent(vdev->pdev,
>> + sizeof(struct vnic_devcmd_fw_info),
>> + &vdev->fw_info_pa);
>> + if (!vdev->fw_info)
>> + return -ENOMEM;
>> +
>> + a0 = vdev->fw_info_pa;
>> +
>> + /* only get fw_info once and cache it */
>> + err = vnic_dev_cmd(vdev, CMD_MCPU_FW_INFO, &a0, &a1, wait);
>> + }
>> +
>> + *fw_info = vdev->fw_info;
>> +
>> + return err;
>> +}
>> +
>> +int vnic_dev_spec(struct vnic_dev *vdev, unsigned int offset, unsigned
>>int size,
>> + void *value)
>> +{
>> + u64 a0, a1;
>> + int wait = VNIC_DVCMD_TMO;
>> + int err;
>> +
>> + a0 = offset;
>> + a1 = size;
>> +
>> + err = vnic_dev_cmd(vdev, CMD_DEV_SPEC, &a0, &a1, wait);
>> +
>> + switch (size) {
>> + case 1:
>> + *(u8 *)value = (u8)a0;
>> + break;
>> + case 2:
>> + *(u16 *)value = (u16)a0;
>> + break;
>> + case 4:
>> + *(u32 *)value = (u32)a0;
>> + break;
>> + case 8:
>> + *(u64 *)value = a0;
>> + break;
>> + default:
>> + BUG();
>> + break;
>> + }
>> +
>> + return err;
>> +}
>> +
>> +int vnic_dev_stats_clear(struct vnic_dev *vdev)
>> +{
>> + u64 a0 = 0, a1 = 0;
>> + int wait = VNIC_DVCMD_TMO;
>> +
>> + return vnic_dev_cmd(vdev, CMD_STATS_CLEAR, &a0, &a1, wait);
>> +}
>> +
>> +int vnic_dev_stats_dump(struct vnic_dev *vdev, struct vnic_stats
>>**stats)
>> +{
>> + u64 a0, a1;
>> + int wait = VNIC_DVCMD_TMO;
>> +
>> + if (!vdev->stats) {
>> + vdev->stats = pci_alloc_consistent(vdev->pdev,
>> + sizeof(struct vnic_stats), &vdev->stats_pa);
>> + if (!vdev->stats)
>> + return -ENOMEM;
>> + }
>> +
>> + *stats = vdev->stats;
>> + a0 = vdev->stats_pa;
>> + a1 = sizeof(struct vnic_stats);
>> +
>> + return vnic_dev_cmd(vdev, CMD_STATS_DUMP, &a0, &a1, wait);
>> +}
>> +
>> +int vnic_dev_close(struct vnic_dev *vdev)
>> +{
>> + u64 a0 = 0, a1 = 0;
>> + int wait = VNIC_DVCMD_TMO;
>> +
>> + return vnic_dev_cmd(vdev, CMD_CLOSE, &a0, &a1, wait);
>> +}
>> +
>> +int vnic_dev_enable_wait(struct vnic_dev *vdev)
>> +{
>> + u64 a0 = 0, a1 = 0;
>> + int wait = VNIC_DVCMD_TMO;
>> + int err = 0;
>> +
>> + err = vnic_dev_cmd(vdev, CMD_ENABLE_WAIT, &a0, &a1, wait);
>> + if (err == ERR_ECMDUNKNOWN)
>> + return vnic_dev_cmd(vdev, CMD_ENABLE, &a0, &a1, wait);
>> +
>> + return err;
>> +}
>> +
>> +int vnic_dev_disable(struct vnic_dev *vdev)
>> +{
>> + u64 a0 = 0, a1 = 0;
>> + int wait = VNIC_DVCMD_TMO;
>> +
>> + return vnic_dev_cmd(vdev, CMD_DISABLE, &a0, &a1, wait);
>> +}
>> +
>> +int vnic_dev_open(struct vnic_dev *vdev, int arg)
>> +{
>> + u64 a0 = (u32)arg, a1 = 0;
>> + int wait = VNIC_DVCMD_TMO;
>> +
>> + return vnic_dev_cmd(vdev, CMD_OPEN, &a0, &a1, wait);
>> +}
>> +
>> +int vnic_dev_open_done(struct vnic_dev *vdev, int *done)
>> +{
>> + u64 a0 = 0, a1 = 0;
>> + int wait = VNIC_DVCMD_TMO;
>> + int err;
>> +
>> + *done = 0;
>> +
>> + err = vnic_dev_cmd(vdev, CMD_OPEN_STATUS, &a0, &a1, wait);
>> + if (err)
>> + return err;
>> +
>> + *done = (a0 == 0);
>> +
>> + return 0;
>> +}
>> +
>> +int vnic_dev_notify_set(struct vnic_dev *vdev, u16 intr)
>> +{
>> + u64 a0, a1;
>> + int wait = VNIC_DVCMD_TMO;
>> +
>> + if (!vdev->notify) {
>> + vdev->notify = pci_alloc_consistent(vdev->pdev,
>> + sizeof(struct vnic_devcmd_notify),
>> + &vdev->notify_pa);
>> + if (!vdev->notify)
>> + return -ENOMEM;
>> + }
>> +
>> + a0 = vdev->notify_pa;
>> + a1 = ((u64)intr << 32) & VNIC_NOTIFY_INTR_MASK;
>> + a1 += sizeof(struct vnic_devcmd_notify);
>> +
>> + return vnic_dev_cmd(vdev, CMD_NOTIFY, &a0, &a1, wait);
>> +}
>> +
>> +void vnic_dev_notify_unset(struct vnic_dev *vdev)
>> +{
>> + u64 a0, a1;
>> + int wait = VNIC_DVCMD_TMO;
>> +
>> + a0 = 0; /* paddr = 0 to unset notify buffer */
>> + a1 = VNIC_NOTIFY_INTR_MASK; /* intr num = -1 to unreg for intr */
>> + a1 += sizeof(struct vnic_devcmd_notify);
>> +
>> + vnic_dev_cmd(vdev, CMD_NOTIFY, &a0, &a1, wait);
>> +}
>> +
>> +static int vnic_dev_notify_ready(struct vnic_dev *vdev)
>> +{
>> + u32 *words;
>> + unsigned int nwords = sizeof(struct vnic_devcmd_notify) / 4;
>> + unsigned int i;
>> + u32 csum;
>> +
>> + if (!vdev->notify)
>> + return 0;
>> +
>> + do {
>> + csum = 0;
>> + memcpy(&vdev->notify_copy, vdev->notify,
>> + sizeof(struct vnic_devcmd_notify));
>> + words = (u32 *)&vdev->notify_copy;
>> + for (i = 1; i < nwords; i++)
>> + csum += words[i];
>> + } while (csum != words[0]);
>> +
>> + return 1;
>> +}
>> +
>> +int vnic_dev_init(struct vnic_dev *vdev, int arg)
>> +{
>> + u64 a0 = (u32)arg, a1 = 0;
>> + int wait = VNIC_DVCMD_TMO;
>> +
>> + return vnic_dev_cmd(vdev, CMD_INIT, &a0, &a1, wait);
>> +}
>> +
>> +int vnic_dev_link_status(struct vnic_dev *vdev)
>> +{
>> + if (vdev->linkstatus)
>> + return *vdev->linkstatus;
>> +
>> + if (!vnic_dev_notify_ready(vdev))
>> + return 0;
>> +
>> + return vdev->notify_copy.link_state;
>> +}
>> +
>> +u32 vnic_dev_link_down_cnt(struct vnic_dev *vdev)
>> +{
>> + if (!vnic_dev_notify_ready(vdev))
>> + return 0;
>> +
>> + return vdev->notify_copy.link_down_cnt;
>> +}
>> +
>> +void vnic_dev_set_intr_mode(struct vnic_dev *vdev,
>> + enum vnic_dev_intr_mode intr_mode)
>> +{
>> + vdev->intr_mode = intr_mode;
>> +}
>> +
>> +enum vnic_dev_intr_mode vnic_dev_get_intr_mode(
>> + struct vnic_dev *vdev)
>> +{
>> + return vdev->intr_mode;
>> +}
>> +
>> +void vnic_dev_unregister(struct vnic_dev *vdev)
>> +{
>> + if (vdev) {
>> + if (vdev->notify)
>> + pci_free_consistent(vdev->pdev,
>> + sizeof(struct vnic_devcmd_notify),
>> + vdev->notify,
>> + vdev->notify_pa);
>> + if (vdev->linkstatus)
>> + pci_free_consistent(vdev->pdev,
>> + sizeof(u32),
>> + vdev->linkstatus,
>> + vdev->linkstatus_pa);
>> + if (vdev->stats)
>> + pci_free_consistent(vdev->pdev,
>> + sizeof(struct vnic_stats),
>> + vdev->stats, vdev->stats_pa);
>> + if (vdev->fw_info)
>> + pci_free_consistent(vdev->pdev,
>> + sizeof(struct vnic_devcmd_fw_info),
>> + vdev->fw_info, vdev->fw_info_pa);
>> + if (vdev->devcmd2)
>> + vnic_dev_deinit_devcmd2(vdev);
>> + kfree(vdev);
>> + }
>> +}
>> +
>> +struct vnic_dev *vnic_dev_alloc_discover(struct vnic_dev *vdev,
>> + void *priv,
>> + struct pci_dev *pdev,
>> + struct vnic_dev_bar *bar,
>> + unsigned int num_bars)
>> +{
>> + if (!vdev) {
>> + vdev = kzalloc(sizeof(struct vnic_dev), GFP_ATOMIC);
>> + if (!vdev)
>> + return NULL;
>> + }
>> +
>> + vdev->priv = priv;
>> + vdev->pdev = pdev;
>> +
>> + if (vnic_dev_discover_res(vdev, bar, num_bars))
>> + goto err_out;
>> +
>> + return vdev;
>> +
>> +err_out:
>> + vnic_dev_unregister(vdev);
>> +
>> + return NULL;
>> +} /* end of vnic_dev_alloc_discover */
>> +
>> +/*
>> + * fallback option is left to keep the interface common for other
>>vnics.
>> + */
>> +int vnic_dev_cmd_init(struct vnic_dev *vdev, int fallback)
>> +{
>> + int err = -ENODEV;
>> + void __iomem *p;
>> +
>> + p = vnic_dev_get_res(vdev, RES_TYPE_DEVCMD2, 0);
>> + if (p)
>> + err = vnic_dev_init_devcmd2(vdev);
>> + else
>> + pr_err("DEVCMD2 resource not found.\n");
>> +
>> + return err;
>> +} /* end of vnic_dev_cmd_init */
>> diff --git a/drivers/scsi/snic/vnic_dev.h b/drivers/scsi/snic/vnic_dev.h
>> new file mode 100644
>> index 0000000..19f3e76
>> --- /dev/null
>> +++ b/drivers/scsi/snic/vnic_dev.h
>> @@ -0,0 +1,140 @@
>> +/*
>> + * Copyright 2014 Cisco Systems, Inc. All rights reserved.
>> + *
>> + * This program is free software; you may redistribute it and/or modify
>> + * it under the terms of the GNU General Public License as published by
>> + * the Free Software Foundation; version 2 of the License.
>> + *
>> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
>> + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
>> + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
>> + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
>> + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
>> + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
>> + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
>> + * SOFTWARE.
>> + */
>> +
>> +#ifndef _VNIC_DEV_H_
>> +#define _VNIC_DEV_H_
>> +
>> +#include "vnic_resource.h"
>> +#include "vnic_devcmd.h"
>> +
>> +/*
>> + * These defines avoid symbol clash between fnic and enic (Cisco 10G
>>Eth
>> + * Driver) when both are built with CONFIG options =y
>> + */
>So why did you choose the same name then?
>Seeing that you rename them anyway, please use the correct names and do
>away with this 'define'.
Sure, I will rename them.
>
>Cheers,
>
>Hannes
>--
>Dr. Hannes Reinecke zSeries & Storage
>hare@suse.de +49 911 74053 688
>SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg
>GF: J. Hawn, J. Guild, F. Imendörffer, HRB 16746 (AG Nürnberg)
Thanks
Narsimhulu
>
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH v4 4/9] snic:Add snic target discovery
2015-04-13 5:25 ` Narsimhulu Musini (nmusini)
@ 2015-04-13 14:29 ` Hannes Reinecke
2015-04-14 10:15 ` Narsimhulu Musini (nmusini)
0 siblings, 1 reply; 24+ messages in thread
From: Hannes Reinecke @ 2015-04-13 14:29 UTC (permalink / raw)
To: Narsimhulu Musini (nmusini), JBottomley, linux-scsi, hch
Cc: Sesidhar Baddela (sebaddel)
On 04/13/2015 07:25 AM, Narsimhulu Musini (nmusini) wrote:
> Hi Hannes,
>
> Thank you for reviewing patches. Please find responses inline.
>
>
>
> On 09/04/15 6:29 pm, "Hannes Reinecke" <hare@suse.de> wrote:
>
>>
>> On 04/09/2015 01:49 PM, Narsimhulu Musini wrote:
[ .. ]
>>> +/* snic_tgt_create: checks for existence of snic_tgt, if it doesn't
>>> + * it creates one.
>>> + */
>>> +static struct snic_tgt *
>>> +snic_tgt_create(struct snic *snic, struct snic_tgt_id *tgtid)
>>> +{
>>> + struct snic_tgt *tgt = NULL;
>>> + unsigned long flags;
>>> + int ret;
>>> +
>>> + tgt = snic_tgt_lookup(snic, tgtid);
>>> + if (tgt) {
>>> + /* update the information if required */
>>> + return tgt;
>>> + }
>>> +
>>> + tgt = kzalloc(sizeof(*tgt), GFP_KERNEL);
>>> + if (!tgt) {
>>> + SNIC_HOST_ERR(snic->shost, "Failure to allocate snic_tgt.\n");
>>> + ret = -ENOMEM;
>>> +
>>> + return tgt;
>>> + }
>>> +
>>> + INIT_LIST_HEAD(&tgt->list);
>>> + tgt->id = tgtid->tgt_id;
>>> + tgt->channel = 0;
>>> +
>>> + SNIC_BUG_ON(tgtid->tgt_type > SNIC_TGT_SAN);
>>> + tgt->tdata.typ = tgtid->tgt_type;
>>> +
>>> + /*
>>> + * Plugging into SML Device Tree
>>> + */
>>> + tgt->tdata.disc_id = 0;
>>> + tgt->state = SNIC_TGT_STAT_INIT;
>>> + device_initialize(&tgt->dev);
>>> + tgt->dev.parent = get_device(&snic->shost->shost_gendev);
>>> + tgt->dev.release = snic_tgt_dev_release;
>> Why do you use your own scsi target instantiation here?
>> If it's equivalent to the scsi target than you should rather use the
>> 'scsi_target' structure here and attach driver-specific information
>> to either hostdata or starget_data.
> I got your idea, we followed an approach similar to fc (fc_rport_create()
> in scsi_transport_fc.c).
> Both are valid approaches. We opted for current approach, and gone through
> good amount of testing.
> I would like to keep the way it is. Please share your thoughts.
scsi_transport_fc() creates its own sysfs hierarchy for the remote
port, which is inserted between scsi_host and scsi_target:
/sys/devices/pci0000:00/0000:00:03.0/0000:07:00.3/host2/rport-2:0-13/target2:0:11/2:0:11:7
here, the fc_rport structure corresponds to the FC sysfs hierarchy,
containing all FC-related sysfs attributes. And as such makes sense
have a distinct object.
In your case no such hiearchy exists, and there are no additional
sysfs attributes which would warrant the creation of a distinct object.
Cheers,
Hannes
--
Dr. Hannes Reinecke zSeries & Storage
hare@suse.de +49 911 74053 688
SUSE LINUX GmbH, Maxfeldstr. 5, 90409 Nürnberg
GF: F. Imendörffer, J. Smithard, J. Guild, D. Upmanyu, G. Norton
HRB 21284 (AG Nürnberg)
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH v4 4/9] snic:Add snic target discovery
2015-04-13 14:29 ` Hannes Reinecke
@ 2015-04-14 10:15 ` Narsimhulu Musini (nmusini)
0 siblings, 0 replies; 24+ messages in thread
From: Narsimhulu Musini (nmusini) @ 2015-04-14 10:15 UTC (permalink / raw)
To: Hannes Reinecke, JBottomley, linux-scsi, hch; +Cc: Sesidhar Baddela (sebaddel)
Hi Hannes,
Thank you for reviewing patches. Please find responses inline.
On 13/04/15 7:59 pm, "Hannes Reinecke" <hare@suse.de> wrote:
>On 04/13/2015 07:25 AM, Narsimhulu Musini (nmusini) wrote:
>> Hi Hannes,
>>
>> Thank you for reviewing patches. Please find responses inline.
>>
>>
>>
>> On 09/04/15 6:29 pm, "Hannes Reinecke" <hare@suse.de> wrote:
>>
>>>
>>> On 04/09/2015 01:49 PM, Narsimhulu Musini wrote:
>[ .. ]
>>>> +/* snic_tgt_create: checks for existence of snic_tgt, if it doesn't
>>>> + * it creates one.
>>>> + */
>>>> +static struct snic_tgt *
>>>> +snic_tgt_create(struct snic *snic, struct snic_tgt_id *tgtid)
>>>> +{
>>>> + struct snic_tgt *tgt = NULL;
>>>> + unsigned long flags;
>>>> + int ret;
>>>> +
>>>> + tgt = snic_tgt_lookup(snic, tgtid);
>>>> + if (tgt) {
>>>> + /* update the information if required */
>>>> + return tgt;
>>>> + }
>>>> +
>>>> + tgt = kzalloc(sizeof(*tgt), GFP_KERNEL);
>>>> + if (!tgt) {
>>>> + SNIC_HOST_ERR(snic->shost, "Failure to allocate snic_tgt.\n");
>>>> + ret = -ENOMEM;
>>>> +
>>>> + return tgt;
>>>> + }
>>>> +
>>>> + INIT_LIST_HEAD(&tgt->list);
>>>> + tgt->id = tgtid->tgt_id;
>>>> + tgt->channel = 0;
>>>> +
>>>> + SNIC_BUG_ON(tgtid->tgt_type > SNIC_TGT_SAN);
>>>> + tgt->tdata.typ = tgtid->tgt_type;
>>>> +
>>>> + /*
>>>> + * Plugging into SML Device Tree
>>>> + */
>>>> + tgt->tdata.disc_id = 0;
>>>> + tgt->state = SNIC_TGT_STAT_INIT;
>>>> + device_initialize(&tgt->dev);
>>>> + tgt->dev.parent = get_device(&snic->shost->shost_gendev);
>>>> + tgt->dev.release = snic_tgt_dev_release;
>>> Why do you use your own scsi target instantiation here?
>>> If it's equivalent to the scsi target than you should rather use the
>>> 'scsi_target' structure here and attach driver-specific information
>>> to either hostdata or starget_data.
>> I got your idea, we followed an approach similar to fc
>>(fc_rport_create()
>> in scsi_transport_fc.c).
>> Both are valid approaches. We opted for current approach, and gone
>>through
>> good amount of testing.
>> I would like to keep the way it is. Please share your thoughts.
>
>scsi_transport_fc() creates its own sysfs hierarchy for the remote
>port, which is inserted between scsi_host and scsi_target:
>
>/sys/devices/pci0000:00/0000:00:03.0/0000:07:00.3/host2/rport-2:0-13/targe
>t2:0:11/2:0:11:7
>
>here, the fc_rport structure corresponds to the FC sysfs hierarchy,
>containing all FC-related sysfs attributes. And as such makes sense
>have a distinct object.
>
>In your case no such hiearchy exists, and there are no additional
>sysfs attributes which would warrant the creation of a distinct object.
We will add sysfs attributes for targets in future patches. And about
hierarchy, The current approach provides the option, which we would like
to avail in future.
>
>Cheers,
>
>Hannes
>--
>Dr. Hannes Reinecke zSeries & Storage
>hare@suse.de +49 911 74053 688
>SUSE LINUX GmbH, Maxfeldstr. 5, 90409 Nürnberg
>GF: F. Imendörffer, J. Smithard, J. Guild, D. Upmanyu, G. Norton
>HRB 21284 (AG Nürnberg)
Thanks
simha
>
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply [flat|nested] 24+ messages in thread
end of thread, other threads:[~2015-04-14 10:15 UTC | newest]
Thread overview: 24+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2015-04-09 11:49 [PATCH v4 0/9] snic:initial submission of snic driver for Cisco SCSI HBA Narsimhulu Musini
2015-04-09 11:49 ` [PATCH v4 1/9] snic: snic module infrastructure Narsimhulu Musini
2015-04-09 12:40 ` Hannes Reinecke
2015-04-09 11:49 ` [PATCH v4 2/9] snic:Add interrupt, resource firmware interfaces Narsimhulu Musini
2015-04-09 12:46 ` Hannes Reinecke
2015-04-09 11:49 ` [PATCH v4 3/9] snic:Add meta request, handling of meta requests Narsimhulu Musini
2015-04-09 12:50 ` Hannes Reinecke
2015-04-09 11:49 ` [PATCH v4 4/9] snic:Add snic target discovery Narsimhulu Musini
2015-04-09 12:59 ` Hannes Reinecke
2015-04-13 5:25 ` Narsimhulu Musini (nmusini)
2015-04-13 14:29 ` Hannes Reinecke
2015-04-14 10:15 ` Narsimhulu Musini (nmusini)
2015-04-09 11:49 ` [PATCH v4 5/9] snic:add SCSI handling, AEN, and fwreset handling Narsimhulu Musini
2015-04-09 13:16 ` Hannes Reinecke
2015-04-13 5:37 ` Narsimhulu Musini (nmusini)
2015-04-09 11:49 ` [PATCH v4 6/9] snic:Add low level queuing interfaces Narsimhulu Musini
2015-04-09 13:23 ` Hannes Reinecke
2015-04-13 5:40 ` Narsimhulu Musini (nmusini)
2015-04-09 11:49 ` [PATCH v4 7/9] snic:Add sysfs entries to list stats and trace data Narsimhulu Musini
2015-04-09 13:23 ` Hannes Reinecke
2015-04-09 11:49 ` [PATCH v4 8/9] snic:Add event tracing to capture IO events Narsimhulu Musini
2015-04-09 13:24 ` Hannes Reinecke
2015-04-09 11:49 ` [PATCH v4 9/9] snic:Add Makefile, patch Kconfig, MAINTAINERS Narsimhulu Musini
2015-04-09 13:25 ` Hannes Reinecke
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.