* [PATCH v4 00/10 iwl-next] idpf: refactor virtchnl messages
@ 2024-02-06 3:37 ` Alan Brady
0 siblings, 0 replies; 44+ messages in thread
From: Alan Brady @ 2024-02-06 3:37 UTC (permalink / raw)
To: intel-wired-lan
Cc: netdev, willemdebruijn.kernel, przemyslaw.kitszel, igor.bagnucki,
aleksander.lobakin, Alan Brady
The motivation for this series has two primary goals. We want to enable
support of multiple simultaneous messages and make the channel more
robust. The way it works right now, the driver can only send and receive
a single message at a time and if something goes really wrong, it can
lead to data corruption and strange bugs.
This works by conceptualizing a send and receive as a "virtchnl
transaction" (idpf_vc_xn) and introducing a "transaction manager"
(idpf_vc_xn_manager). The vcxn_mngr will init a ring of transactions
from which the driver will pop from a bitmap of free transactions to
track in-flight messages. Instead of needing to handle a complicated
send/recv for every a message, the driver now just needs to fill out a
xn_params struct and hand it over to idpf_vc_xn_exec which will take
care of all the messy bits. Once a message is sent and receives a reply,
we leverage the completion API to signal the received buffer is ready to
be used (assuming success, or an error code otherwise).
At a low-level, this implements the "sw cookie" field of the virtchnl
message descriptor to enable this. We have 16 bits we can put whatever
we want and the recipient is required to apply the same cookie to the
reply for that message. We use the first 8 bits as an index into the
array of transactions to enable fast lookups and we use the second 8
bits as a salt to make sure each cookie is unique for that message. As
transactions are received in arbitrary order, it's possible to reuse a
transaction index and the salt guards against index conflicts to make
certain the lookup is correct. As a primitive example, say index 1 is
used with salt 1. The message times out without receiving a reply so
index 1 is renewed to be ready for a new transaction, we report the
timeout, and send the message again. Since index 1 is free to be used
again now, index 1 is again sent but now salt is 2. This time we do get
a reply, however it could be that the reply is _actually_ for the
previous send index 1 with salt 1. Without the salt we would have no
way of knowing for sure if it's the correct reply, but with we will know
for certain.
Through this conversion we also get several other benefits. We can now
more appropriately handle asynchronously sent messages by providing
space for a callback to be defined. This notably allows us to handle MAC
filter failures better; previously we could potentially have stale,
failed filters in our list, which shouldn't really have a major impact
but is obviously not correct. I also managed to remove fairly
significant more lines than I added which is a win in my book.
Additionally, this converts some variables to use auto-variables where
appropriate. This makes the alloc paths much cleaner and less prone to
memory leaks. We also fix a few virtchnl related bugs while we're here.
Alan Brady (10):
idpf: implement virtchnl transaction manager
idpf: refactor vport virtchnl messages
idpf: refactor queue related virtchnl messages
idpf: refactor remaining virtchnl messages
idpf: add async_handler for MAC filter messages
idpf: refactor idpf_recv_mb_msg
idpf: cleanup virtchnl cruft
idpf: prevent deinit uninitialized virtchnl core
idpf: fix minor controlq issues
idpf: remove dealloc vector msg err in idpf_intr_rel
drivers/net/ethernet/intel/idpf/idpf.h | 194 +-
.../net/ethernet/intel/idpf/idpf_controlq.c | 7 +-
.../ethernet/intel/idpf/idpf_controlq_api.h | 5 +
drivers/net/ethernet/intel/idpf/idpf_lib.c | 38 +-
drivers/net/ethernet/intel/idpf/idpf_main.c | 3 +-
drivers/net/ethernet/intel/idpf/idpf_vf_dev.c | 2 +-
.../net/ethernet/intel/idpf/idpf_virtchnl.c | 2175 ++++++++---------
7 files changed, 1096 insertions(+), 1328 deletions(-)
--
v1 -> v2:
- don't take spin_lock in idpf_vc_xn_init, it's not needed
- fix set but unused error on payload_size var in idpf_recv_mb_msg
- prefer bitmap_fill and bitmap_zero if not setting an explicit
range per documention
- remove a couple unnecessary casts in idpf_send_get_stats_msg and
idpf_send_get_rx_ptype_msg
- split patch 4/6 such that the added functionality for MAC filters
is separate
v2 -> v3:
- fix 'mac' -> 'MAC' in async handler error messages
- fix size_t format specifier in async handler error message
- change some variables to use auto-variables instead
v3 -> v4:
- revert changes to idpf_send_mb_msg that were introduced in v3,
this will be addressed in future patch
- tweak idpf_recv_mb_msg refactoring to avoid bailing out of the
while loop when there are more messages to process and add comment
in idpf_vc_xn_forward_reply about ENXIO
- include some minor fixes to lower level ctrlq that seem like good
candidates to add here
- include fix to prevent deinit uninitialized vc core
- remove idpf_send_dealloc_vectors_msg error
2.40.1
^ permalink raw reply [flat|nested] 44+ messages in thread
* [Intel-wired-lan] [PATCH v4 00/10 iwl-next] idpf: refactor virtchnl messages
@ 2024-02-06 3:37 ` Alan Brady
0 siblings, 0 replies; 44+ messages in thread
From: Alan Brady @ 2024-02-06 3:37 UTC (permalink / raw)
To: intel-wired-lan
Cc: willemdebruijn.kernel, netdev, aleksander.lobakin, Alan Brady,
przemyslaw.kitszel, igor.bagnucki
The motivation for this series has two primary goals. We want to enable
support of multiple simultaneous messages and make the channel more
robust. The way it works right now, the driver can only send and receive
a single message at a time and if something goes really wrong, it can
lead to data corruption and strange bugs.
This works by conceptualizing a send and receive as a "virtchnl
transaction" (idpf_vc_xn) and introducing a "transaction manager"
(idpf_vc_xn_manager). The vcxn_mngr will init a ring of transactions
from which the driver will pop from a bitmap of free transactions to
track in-flight messages. Instead of needing to handle a complicated
send/recv for every a message, the driver now just needs to fill out a
xn_params struct and hand it over to idpf_vc_xn_exec which will take
care of all the messy bits. Once a message is sent and receives a reply,
we leverage the completion API to signal the received buffer is ready to
be used (assuming success, or an error code otherwise).
At a low-level, this implements the "sw cookie" field of the virtchnl
message descriptor to enable this. We have 16 bits we can put whatever
we want and the recipient is required to apply the same cookie to the
reply for that message. We use the first 8 bits as an index into the
array of transactions to enable fast lookups and we use the second 8
bits as a salt to make sure each cookie is unique for that message. As
transactions are received in arbitrary order, it's possible to reuse a
transaction index and the salt guards against index conflicts to make
certain the lookup is correct. As a primitive example, say index 1 is
used with salt 1. The message times out without receiving a reply so
index 1 is renewed to be ready for a new transaction, we report the
timeout, and send the message again. Since index 1 is free to be used
again now, index 1 is again sent but now salt is 2. This time we do get
a reply, however it could be that the reply is _actually_ for the
previous send index 1 with salt 1. Without the salt we would have no
way of knowing for sure if it's the correct reply, but with we will know
for certain.
Through this conversion we also get several other benefits. We can now
more appropriately handle asynchronously sent messages by providing
space for a callback to be defined. This notably allows us to handle MAC
filter failures better; previously we could potentially have stale,
failed filters in our list, which shouldn't really have a major impact
but is obviously not correct. I also managed to remove fairly
significant more lines than I added which is a win in my book.
Additionally, this converts some variables to use auto-variables where
appropriate. This makes the alloc paths much cleaner and less prone to
memory leaks. We also fix a few virtchnl related bugs while we're here.
Alan Brady (10):
idpf: implement virtchnl transaction manager
idpf: refactor vport virtchnl messages
idpf: refactor queue related virtchnl messages
idpf: refactor remaining virtchnl messages
idpf: add async_handler for MAC filter messages
idpf: refactor idpf_recv_mb_msg
idpf: cleanup virtchnl cruft
idpf: prevent deinit uninitialized virtchnl core
idpf: fix minor controlq issues
idpf: remove dealloc vector msg err in idpf_intr_rel
drivers/net/ethernet/intel/idpf/idpf.h | 194 +-
.../net/ethernet/intel/idpf/idpf_controlq.c | 7 +-
.../ethernet/intel/idpf/idpf_controlq_api.h | 5 +
drivers/net/ethernet/intel/idpf/idpf_lib.c | 38 +-
drivers/net/ethernet/intel/idpf/idpf_main.c | 3 +-
drivers/net/ethernet/intel/idpf/idpf_vf_dev.c | 2 +-
.../net/ethernet/intel/idpf/idpf_virtchnl.c | 2175 ++++++++---------
7 files changed, 1096 insertions(+), 1328 deletions(-)
--
v1 -> v2:
- don't take spin_lock in idpf_vc_xn_init, it's not needed
- fix set but unused error on payload_size var in idpf_recv_mb_msg
- prefer bitmap_fill and bitmap_zero if not setting an explicit
range per documention
- remove a couple unnecessary casts in idpf_send_get_stats_msg and
idpf_send_get_rx_ptype_msg
- split patch 4/6 such that the added functionality for MAC filters
is separate
v2 -> v3:
- fix 'mac' -> 'MAC' in async handler error messages
- fix size_t format specifier in async handler error message
- change some variables to use auto-variables instead
v3 -> v4:
- revert changes to idpf_send_mb_msg that were introduced in v3,
this will be addressed in future patch
- tweak idpf_recv_mb_msg refactoring to avoid bailing out of the
while loop when there are more messages to process and add comment
in idpf_vc_xn_forward_reply about ENXIO
- include some minor fixes to lower level ctrlq that seem like good
candidates to add here
- include fix to prevent deinit uninitialized vc core
- remove idpf_send_dealloc_vectors_msg error
2.40.1
^ permalink raw reply [flat|nested] 44+ messages in thread
* [PATCH v4 01/10 iwl-next] idpf: implement virtchnl transaction manager
2024-02-06 3:37 ` [Intel-wired-lan] " Alan Brady
@ 2024-02-06 3:37 ` Alan Brady
-1 siblings, 0 replies; 44+ messages in thread
From: Alan Brady @ 2024-02-06 3:37 UTC (permalink / raw)
To: intel-wired-lan
Cc: netdev, willemdebruijn.kernel, przemyslaw.kitszel, igor.bagnucki,
aleksander.lobakin, Alan Brady, Joshua Hay
This starts refactoring how virtchnl messages are handled by adding a
transaction manager (idpf_vc_xn_manager).
There are two primary motivations here which are to enable handling of
multiple messages at once and to make it more robust in general. As it
is right now, the driver may only have one pending message at a time and
there's no guarantee that the response we receive was actually intended
for the message we sent prior.
This works by utilizing a "cookie" field of the message descriptor. It
is arbitrary what data we put in the cookie and the response is required
to have the same cookie the original message was sent with. Then using a
"transaction" abstraction that uses the completion API to pair responses
to the message it belongs to.
The cookie works such that the first half is the index to the
transaction in our array, and the second half is a "salt" that gets
incremented every message. This enables quick lookups into the array and
also ensuring we have the correct message. The salt is necessary because
after, for example, a message times out and we deem the response was
lost for some reason, we could theoretically reuse the same index but
using a different salt ensures that when we do actually get a response
it's not the old message that timed out previously finally coming in.
Since the number of transactions allocated is U8_MAX and the salt is 8
bits, we can never have a conflict because we can't roll over the salt
without using more transactions than we have available.
This starts by only converting the VIRTCHNL2_OP_VERSION message to use
this new transaction API. Follow up patches will convert all virtchnl
messages to use the API.
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Reviewed-by: Igor Bagnucki <igor.bagnucki@intel.com>
Co-developed-by: Joshua Hay <joshua.a.hay@intel.com>
Signed-off-by: Joshua Hay <joshua.a.hay@intel.com>
Signed-off-by: Alan Brady <alan.brady@intel.com>
---
drivers/net/ethernet/intel/idpf/idpf.h | 101 +++-
.../ethernet/intel/idpf/idpf_controlq_api.h | 5 +
drivers/net/ethernet/intel/idpf/idpf_lib.c | 2 +
drivers/net/ethernet/intel/idpf/idpf_main.c | 1 +
drivers/net/ethernet/intel/idpf/idpf_vf_dev.c | 2 +-
.../net/ethernet/intel/idpf/idpf_virtchnl.c | 510 +++++++++++++++---
6 files changed, 548 insertions(+), 73 deletions(-)
diff --git a/drivers/net/ethernet/intel/idpf/idpf.h b/drivers/net/ethernet/intel/idpf/idpf.h
index 0acc125decb3..92103c3cf5a0 100644
--- a/drivers/net/ethernet/intel/idpf/idpf.h
+++ b/drivers/net/ethernet/intel/idpf/idpf.h
@@ -39,6 +39,11 @@ struct idpf_vport_max_q;
((IDPF_CTLQ_MAX_BUF_LEN - (struct_sz)) / (chunk_sz))
#define IDPF_WAIT_FOR_EVENT_TIMEO_MIN 2000
#define IDPF_WAIT_FOR_EVENT_TIMEO 60000
+#define IDPF_VC_XN_MIN_TIMEOUT_MSEC 2000
+#define IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC (60 * 1000)
+#define IDPF_VC_XN_IDX_M GENMASK(7, 0)
+#define IDPF_VC_XN_SALT_M GENMASK(15, 8)
+#define IDPF_VC_XN_RING_LEN U8_MAX
#define IDPF_MAX_WAIT 500
@@ -66,14 +71,12 @@ struct idpf_mac_filter {
/**
* enum idpf_state - State machine to handle bring up
- * @__IDPF_STARTUP: Start the state machine
* @__IDPF_VER_CHECK: Negotiate virtchnl version
* @__IDPF_GET_CAPS: Negotiate capabilities
* @__IDPF_INIT_SW: Init based on given capabilities
* @__IDPF_STATE_LAST: Must be last, used to determine size
*/
enum idpf_state {
- __IDPF_STARTUP,
__IDPF_VER_CHECK,
__IDPF_GET_CAPS,
__IDPF_INIT_SW,
@@ -314,6 +317,96 @@ struct idpf_port_stats {
struct virtchnl2_vport_stats vport_stats;
};
+/**
+ * enum idpf_vc_xn_state - Virtchnl transaction status
+ * @IDPF_VC_XN_IDLE: not expecting a reply, ready to be used
+ * @IDPF_VC_XN_WAITING: expecting a reply, not yet received
+ * @IDPF_VC_XN_COMPLETED_SUCCESS: a reply was expected and received,
+ * buffer updated
+ * @IDPF_VC_XN_COMPLETED_FAILED: a reply was expected and received, but there
+ * was an error, buffer not updated
+ * @IDPF_VC_XN_SHUTDOWN: transaction object cannot be used, VC torn down
+ * @IDPF_VC_XN_ASYNC: transaction sent asynchronously and doesn't have the
+ * return context; a callback may be provided to handle
+ * return
+ */
+enum idpf_vc_xn_state {
+ IDPF_VC_XN_IDLE = 1,
+ IDPF_VC_XN_WAITING,
+ IDPF_VC_XN_COMPLETED_SUCCESS,
+ IDPF_VC_XN_COMPLETED_FAILED,
+ IDPF_VC_XN_SHUTDOWN,
+ IDPF_VC_XN_ASYNC,
+};
+
+struct idpf_vc_xn;
+/* Callback for asynchronous messages */
+typedef int (*async_vc_cb) (struct idpf_adapter *, struct idpf_vc_xn *,
+ const struct idpf_ctlq_msg *);
+
+/**
+ * struct idpf_vc_xn - Data structure representing virtchnl transactions
+ * @completed: virtchnl event loop uses that to signal when a reply is
+ * available, uses kernel completion API
+ * @state: virtchnl event loop stores the data below, protected by the
+ * completion's lock.
+ * @reply_sz: Original size of reply, may be > reply_buf.iov_len; it will be
+ * truncated on its way to the receiver thread according to
+ * reply_buf.iov_len.
+ * @reply: Reference to the buffer(s) where the reply data should be written
+ * to. May be 0-length (then NULL address permitted) if the reply data
+ * should be ignored.
+ * @async_handler: if sent asynchronously, a callback can be provided to handle
+ * the reply when it's received
+ * @vc_op: corresponding opcode sent with this transaction
+ * @idx: index used as retrieval on reply receive, used for cookie
+ * @salt: changed every message to make unique, used for cookie
+ */
+struct idpf_vc_xn {
+ struct completion completed;
+ enum idpf_vc_xn_state state;
+ size_t reply_sz;
+ struct kvec reply;
+ async_vc_cb async_handler;
+ u32 vc_op;
+ u8 idx;
+ u8 salt;
+};
+
+/**
+ * struct idpf_vc_xn_params - Parameters for executing transaction
+ * @send_buf: kvec for send buffer
+ * @recv_buf: kvec for recv buffer, may be NULL, must then have zero length
+ * @timeout_ms: timeout to wait for reply
+ * @async: send message asynchronously, will not wait on completion
+ * @async_handler: If sent asynchronously, optional callback handler. The user
+ * must be careful when using async handlers as the memory for
+ * the recv_buf _cannot_ be on stack if this is async.
+ * @vc_op: virtchnl op to send
+ */
+struct idpf_vc_xn_params {
+ struct kvec send_buf;
+ struct kvec recv_buf;
+ int timeout_ms;
+ bool async;
+ async_vc_cb async_handler;
+ u32 vc_op;
+};
+
+/**
+ * struct idpf_vc_xn_manager - Manager for tracking transactions
+ * @ring: backing and lookup for transactions
+ * @free_xn_bm: bitmap for free transactions
+ * @xn_bm_lock: make bitmap access synchronous where necessary
+ * @salt: used to make cookie unique every message
+ */
+struct idpf_vc_xn_manager {
+ struct idpf_vc_xn ring[IDPF_VC_XN_RING_LEN];
+ DECLARE_BITMAP(free_xn_bm, IDPF_VC_XN_RING_LEN);
+ spinlock_t xn_bm_lock;
+ u8 salt;
+};
+
/**
* struct idpf_vport - Handle for netdevices and queue resources
* @num_txq: Number of allocated TX queues
@@ -604,6 +697,7 @@ struct idpf_vport_config {
* @vchnl_wq: Wait queue for virtchnl messages
* @vc_state: Virtchnl message state
* @vc_msg: Virtchnl message buffer
+ * @vcxn_mngr: Virtchnl transaction manager
* @dev_ops: See idpf_dev_ops
* @num_vfs: Number of allocated VFs through sysfs. PF does not directly talk
* to VFs but is used to initialize them
@@ -663,6 +757,7 @@ struct idpf_adapter {
wait_queue_head_t vchnl_wq;
DECLARE_BITMAP(vc_state, IDPF_VC_NBITS);
char vc_msg[IDPF_CTLQ_MAX_BUF_LEN];
+ struct idpf_vc_xn_manager vcxn_mngr;
struct idpf_dev_ops dev_ops;
int num_vfs;
bool crc_enable;
@@ -939,7 +1034,7 @@ int idpf_get_vec_ids(struct idpf_adapter *adapter,
int idpf_recv_mb_msg(struct idpf_adapter *adapter, u32 op,
void *msg, int msg_size);
int idpf_send_mb_msg(struct idpf_adapter *adapter, u32 op,
- u16 msg_size, u8 *msg);
+ u16 msg_size, u8 *msg, u16 cookie);
void idpf_set_ethtool_ops(struct net_device *netdev);
int idpf_vport_alloc_max_qs(struct idpf_adapter *adapter,
struct idpf_vport_max_q *max_q);
diff --git a/drivers/net/ethernet/intel/idpf/idpf_controlq_api.h b/drivers/net/ethernet/intel/idpf/idpf_controlq_api.h
index 8dee098bbfb0..e8e046ef2f0d 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_controlq_api.h
+++ b/drivers/net/ethernet/intel/idpf/idpf_controlq_api.h
@@ -69,6 +69,11 @@ struct idpf_ctlq_msg {
u8 context[IDPF_INDIRECT_CTX_SIZE];
struct idpf_dma_mem *payload;
} indirect;
+ struct {
+ u32 rsvd;
+ u16 data;
+ u16 flags;
+ } sw_cookie;
} ctx;
};
diff --git a/drivers/net/ethernet/intel/idpf/idpf_lib.c b/drivers/net/ethernet/intel/idpf/idpf_lib.c
index 58179bd733ff..6b4a7408246c 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_lib.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_lib.c
@@ -1823,6 +1823,8 @@ static int idpf_init_hard_reset(struct idpf_adapter *adapter)
goto unlock_mutex;
}
+ queue_delayed_work(adapter->mbx_wq, &adapter->mbx_task, 0);
+
/* Initialize the state machine, also allocate memory and request
* resources
*/
diff --git a/drivers/net/ethernet/intel/idpf/idpf_main.c b/drivers/net/ethernet/intel/idpf/idpf_main.c
index e1febc74cefd..c82233b112bd 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_main.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_main.c
@@ -30,6 +30,7 @@ static void idpf_remove(struct pci_dev *pdev)
idpf_sriov_configure(pdev, 0);
idpf_vc_core_deinit(adapter);
+
/* Be a good citizen and leave the device clean on exit */
adapter->dev_ops.reg_ops.trigger_reset(adapter, IDPF_HR_FUNC_RESET);
idpf_deinit_dflt_mbx(adapter);
diff --git a/drivers/net/ethernet/intel/idpf/idpf_vf_dev.c b/drivers/net/ethernet/intel/idpf/idpf_vf_dev.c
index 8ade4e3a9fe1..192d19a0c89a 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_vf_dev.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_vf_dev.c
@@ -137,7 +137,7 @@ static void idpf_vf_trigger_reset(struct idpf_adapter *adapter,
/* Do not send VIRTCHNL2_OP_RESET_VF message on driver unload */
if (trig_cause == IDPF_HR_FUNC_RESET &&
!test_bit(IDPF_REMOVE_IN_PROG, adapter->flags))
- idpf_send_mb_msg(adapter, VIRTCHNL2_OP_RESET_VF, 0, NULL);
+ idpf_send_mb_msg(adapter, VIRTCHNL2_OP_RESET_VF, 0, NULL, 0);
}
/**
diff --git a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
index d0cdd63b3d5b..92173bcb9fc3 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
@@ -93,13 +93,14 @@ static int idpf_mb_clean(struct idpf_adapter *adapter)
* @op: virtchnl opcode
* @msg_size: size of the payload
* @msg: pointer to buffer holding the payload
+ * @cookie: unique SW generated cookie per message
*
* Will prepare the control queue message and initiates the send api
*
* Returns 0 on success, negative on failure
*/
int idpf_send_mb_msg(struct idpf_adapter *adapter, u32 op,
- u16 msg_size, u8 *msg)
+ u16 msg_size, u8 *msg, u16 cookie)
{
struct idpf_ctlq_msg *ctlq_msg;
struct idpf_dma_mem *dma_mem;
@@ -139,8 +140,12 @@ int idpf_send_mb_msg(struct idpf_adapter *adapter, u32 op,
err = -ENOMEM;
goto dma_alloc_error;
}
- memcpy(dma_mem->va, msg, msg_size);
+
+ /* It's possible we're just sending an opcode but no buffer */
+ if (msg && msg_size)
+ memcpy(dma_mem->va, msg, msg_size);
ctlq_msg->ctx.indirect.payload = dma_mem;
+ ctlq_msg->ctx.sw_cookie.data = cookie;
err = idpf_ctlq_send(&adapter->hw, adapter->hw.asq, 1, ctlq_msg);
if (err)
@@ -361,6 +366,378 @@ static void idpf_recv_vchnl_op(struct idpf_adapter *adapter,
}
}
+/* API for virtchnl "transaction" support ("xn" for short).
+ *
+ * We are reusing the completion lock to serialize the accesses to the
+ * transaction state for simplicity, but it could be its own separate synchro
+ * as well. For now, this API is only used from within a workqueue context;
+ * raw_spin_lock() is enough.
+ */
+/**
+ * idpf_vc_xn_lock - Request exclusive access to vc transaction
+ * @xn: struct idpf_vc_xn* to access
+ */
+#define idpf_vc_xn_lock(xn) \
+ raw_spin_lock(&(xn)->completed.wait.lock)
+
+/**
+ * idpf_vc_xn_unlock - Release exclusive access to vc transaction
+ * @xn: struct idpf_vc_xn* to access
+ */
+#define idpf_vc_xn_unlock(xn) \
+ raw_spin_unlock(&(xn)->completed.wait.lock)
+
+/**
+ * idpf_vc_xn_release_bufs - Release reference to reply buffer(s) and
+ * reset the transaction state.
+ * @xn: struct idpf_vc_xn to update
+ */
+static void idpf_vc_xn_release_bufs(struct idpf_vc_xn *xn)
+{
+ xn->reply.iov_base = NULL;
+ xn->reply.iov_len = 0;
+
+ if (xn->state != IDPF_VC_XN_SHUTDOWN)
+ xn->state = IDPF_VC_XN_IDLE;
+}
+
+/**
+ * idpf_vc_xn_init - Initialize virtchnl transaction object
+ * @vcxn_mngr: pointer to vc transaction manager struct
+ */
+static void idpf_vc_xn_init(struct idpf_vc_xn_manager *vcxn_mngr)
+{
+ int i;
+
+ spin_lock_init(&vcxn_mngr->xn_bm_lock);
+
+ for (i = 0; i < ARRAY_SIZE(vcxn_mngr->ring); i++) {
+ struct idpf_vc_xn *xn = &vcxn_mngr->ring[i];
+
+ xn->state = IDPF_VC_XN_IDLE;
+ xn->idx = i;
+ idpf_vc_xn_release_bufs(xn);
+ init_completion(&xn->completed);
+ }
+
+ bitmap_fill(vcxn_mngr->free_xn_bm, IDPF_VC_XN_RING_LEN);
+}
+
+/**
+ * idpf_vc_xn_shutdown - Uninitialize virtchnl transaction object
+ * @vcxn_mngr: pointer to vc transaction manager struct
+ *
+ * All waiting threads will be woken-up and their transaction aborted. Further
+ * operations on that object will fail.
+ */
+static void idpf_vc_xn_shutdown(struct idpf_vc_xn_manager *vcxn_mngr)
+{
+ int i;
+
+ spin_lock_bh(&vcxn_mngr->xn_bm_lock);
+ bitmap_zero(vcxn_mngr->free_xn_bm, IDPF_VC_XN_RING_LEN);
+ spin_unlock_bh(&vcxn_mngr->xn_bm_lock);
+
+ for (i = 0; i < ARRAY_SIZE(vcxn_mngr->ring); i++) {
+ struct idpf_vc_xn *xn = &vcxn_mngr->ring[i];
+
+ idpf_vc_xn_lock(xn);
+ xn->state = IDPF_VC_XN_SHUTDOWN;
+ idpf_vc_xn_release_bufs(xn);
+ idpf_vc_xn_unlock(xn);
+ complete_all(&xn->completed);
+ }
+}
+
+/**
+ * idpf_vc_xn_pop_free - Pop a free transaction from free list
+ * @vcxn_mngr: transaction manager to pop from
+ *
+ * Returns NULL if no free transactions
+ */
+static
+struct idpf_vc_xn *idpf_vc_xn_pop_free(struct idpf_vc_xn_manager *vcxn_mngr)
+{
+ struct idpf_vc_xn *xn = NULL;
+ unsigned long free_idx;
+
+ spin_lock_bh(&vcxn_mngr->xn_bm_lock);
+ free_idx = find_first_bit(vcxn_mngr->free_xn_bm, IDPF_VC_XN_RING_LEN);
+ if (free_idx == IDPF_VC_XN_RING_LEN)
+ goto do_unlock;
+
+ clear_bit(free_idx, vcxn_mngr->free_xn_bm);
+ xn = &vcxn_mngr->ring[free_idx];
+ xn->salt = vcxn_mngr->salt++;
+
+do_unlock:
+ spin_unlock_bh(&vcxn_mngr->xn_bm_lock);
+
+ return xn;
+}
+
+/**
+ * idpf_vc_xn_push_free - Push a free transaction to free list
+ * @vcxn_mngr: transaction manager to push to
+ * @xn: transaction to push
+ */
+static void idpf_vc_xn_push_free(struct idpf_vc_xn_manager *vcxn_mngr,
+ struct idpf_vc_xn *xn)
+{
+ idpf_vc_xn_release_bufs(xn);
+ set_bit(xn->idx, vcxn_mngr->free_xn_bm);
+}
+
+/**
+ * idpf_vc_xn_exec - Perform a send/recv virtchnl transaction
+ * @adapter: driver specific private structure with vcxn_mngr
+ * @params: parameters for this particular transaction including
+ * -vc_op: virtchannel operation to send
+ * -send_buf: kvec iov for send buf and len
+ * -recv_buf: kvec iov for recv buf and len (ignored if NULL)
+ * -timeout_ms: timeout waiting for a reply (milliseconds)
+ * -async: don't wait for message reply, will lose caller context
+ * -async_handler: callback to handle async replies
+ *
+ * @returns >= 0 for success, the size of the initial reply (may or may not be
+ * >= @recv_buf.iov_len, but we never overflow @@recv_buf_iov_base). < 0 for
+ * error.
+ */
+static ssize_t idpf_vc_xn_exec(struct idpf_adapter *adapter,
+ struct idpf_vc_xn_params params)
+{
+ struct kvec *send_buf = ¶ms.send_buf;
+ struct idpf_vc_xn *xn;
+ ssize_t retval;
+ u16 cookie;
+
+ xn = idpf_vc_xn_pop_free(&adapter->vcxn_mngr);
+ /* no free transactions available */
+ if (!xn)
+ return -ENOSPC;
+
+ idpf_vc_xn_lock(xn);
+ if (xn->state == IDPF_VC_XN_SHUTDOWN) {
+ retval = -ENXIO;
+ goto only_unlock;
+ } else if (xn->state != IDPF_VC_XN_IDLE) {
+ /* We're just going to clobber this transaction even though
+ * it's not IDLE. If we don't reuse it we could theoretically
+ * eventually leak all the free transactions and not be able to
+ * send any messages. At least this way we make an attempt to
+ * remain functional even though something really bad is
+ * happening that's corrupting what was supposed to be free
+ * transactions.
+ */
+ WARN_ONCE(1, "There should only be idle transactions in free list (idx %d op %d)\n",
+ xn->idx, xn->vc_op);
+ }
+
+ xn->reply = params.recv_buf;
+ xn->reply_sz = 0;
+ xn->state = params.async ? IDPF_VC_XN_ASYNC : IDPF_VC_XN_WAITING;
+ xn->vc_op = params.vc_op;
+ xn->async_handler = params.async_handler;
+ idpf_vc_xn_unlock(xn);
+
+ if (!params.async)
+ reinit_completion(&xn->completed);
+ cookie = FIELD_PREP(IDPF_VC_XN_SALT_M, xn->salt) |
+ FIELD_PREP(IDPF_VC_XN_IDX_M, xn->idx);
+
+ retval = idpf_send_mb_msg(adapter, params.vc_op,
+ send_buf->iov_len, send_buf->iov_base,
+ cookie);
+ if (retval) {
+ idpf_vc_xn_lock(xn);
+ goto release_and_unlock;
+ }
+
+ if (params.async)
+ return 0;
+
+ wait_for_completion_timeout(&xn->completed,
+ msecs_to_jiffies(params.timeout_ms));
+
+ /* No need to check the return value; we check the final state of the
+ * transaction below. It's possible the transaction actually gets more
+ * timeout than specified if we get preempted here but after
+ * wait_for_completion_timeout returns. This should be non-issue
+ * however.
+ */
+ idpf_vc_xn_lock(xn);
+ switch (xn->state) {
+ case IDPF_VC_XN_SHUTDOWN:
+ retval = -ENXIO;
+ goto only_unlock;
+ case IDPF_VC_XN_WAITING:
+ dev_notice_ratelimited(&adapter->pdev->dev, "Transaction timed-out (op %d, %dms)\n",
+ params.vc_op, params.timeout_ms);
+ retval = -ETIME;
+ break;
+ case IDPF_VC_XN_COMPLETED_SUCCESS:
+ retval = xn->reply_sz;
+ break;
+ case IDPF_VC_XN_COMPLETED_FAILED:
+ dev_notice_ratelimited(&adapter->pdev->dev, "Transaction failed (op %d)\n",
+ params.vc_op);
+ retval = -EIO;
+ break;
+ default:
+ /* Invalid state. */
+ WARN_ON_ONCE(1);
+ retval = -EIO;
+ break;
+ }
+
+release_and_unlock:
+ idpf_vc_xn_push_free(&adapter->vcxn_mngr, xn);
+ /* If we receive a VC reply after here, it will be dropped. */
+only_unlock:
+ idpf_vc_xn_unlock(xn);
+
+ return retval;
+}
+
+/**
+ * idpf_vc_xn_forward_async - Handle async reply receives
+ * @adapter: private data struct
+ * @xn: transaction to handle
+ * @ctlq_msg: corresponding ctlq_msg
+ *
+ * For async sends we're going to lose the caller's context so, if an
+ * async_handler was provided, it can deal with the reply, otherwise we'll just
+ * check and report if there is an error.
+ */
+static int
+idpf_vc_xn_forward_async(struct idpf_adapter *adapter, struct idpf_vc_xn *xn,
+ const struct idpf_ctlq_msg *ctlq_msg)
+{
+ int err = 0;
+
+ if (ctlq_msg->cookie.mbx.chnl_opcode != xn->vc_op) {
+ dev_err_ratelimited(&adapter->pdev->dev, "Async message opcode does not match transaction opcode (msg: %d) (xn: %d)\n",
+ ctlq_msg->cookie.mbx.chnl_opcode, xn->vc_op);
+ xn->reply_sz = 0;
+ err = -EINVAL;
+ goto release_bufs;
+ }
+
+ if (xn->async_handler) {
+ err = xn->async_handler(adapter, xn, ctlq_msg);
+ goto release_bufs;
+ }
+
+ if (ctlq_msg->cookie.mbx.chnl_retval) {
+ xn->reply_sz = 0;
+ dev_err_ratelimited(&adapter->pdev->dev, "Async message failure (op %d)\n",
+ ctlq_msg->cookie.mbx.chnl_opcode);
+ err = -EINVAL;
+ }
+
+release_bufs:
+ idpf_vc_xn_push_free(&adapter->vcxn_mngr, xn);
+
+ return err;
+}
+
+/**
+ * idpf_vc_xn_forward_reply - copy a reply back to receiving thread
+ * @adapter: driver specific private structure with vcxn_mngr
+ * @ctlq_msg: controlq message to send back to receiving thread
+ */
+static int
+idpf_vc_xn_forward_reply(struct idpf_adapter *adapter,
+ const struct idpf_ctlq_msg *ctlq_msg)
+{
+ const void *payload = NULL;
+ size_t payload_size = 0;
+ struct idpf_vc_xn *xn;
+ u16 msg_info;
+ int err = 0;
+ u16 xn_idx;
+ u16 salt;
+
+ msg_info = ctlq_msg->ctx.sw_cookie.data;
+ xn_idx = FIELD_GET(IDPF_VC_XN_IDX_M, msg_info);
+ if (xn_idx >= ARRAY_SIZE(adapter->vcxn_mngr.ring)) {
+ dev_err_ratelimited(&adapter->pdev->dev, "Out of bounds cookie received: %02x\n",
+ xn_idx);
+ return -EINVAL;
+ }
+ xn = &adapter->vcxn_mngr.ring[xn_idx];
+ salt = FIELD_GET(IDPF_VC_XN_SALT_M, msg_info);
+ if (xn->salt != salt) {
+ dev_err_ratelimited(&adapter->pdev->dev, "Transaction salt does not match (%02x != %02x)\n",
+ xn->salt, salt);
+ return -EINVAL;
+ }
+
+ idpf_vc_xn_lock(xn);
+ switch (xn->state) {
+ case IDPF_VC_XN_WAITING:
+ /* success */
+ break;
+ case IDPF_VC_XN_IDLE:
+ dev_err_ratelimited(&adapter->pdev->dev, "Unexpected or belated VC reply (op %d)\n",
+ ctlq_msg->cookie.mbx.chnl_opcode);
+ err = -EINVAL;
+ goto out_unlock;
+ case IDPF_VC_XN_SHUTDOWN:
+ /* ENXIO is a bit special here as the recv msg loop uses that
+ * know if it should stop trying to clean the ring if we lost
+ * the virtchnl. We need to stop playing with registers and
+ * yield.
+ */
+ err = -ENXIO;
+ goto out_unlock;
+ case IDPF_VC_XN_ASYNC:
+ err = idpf_vc_xn_forward_async(adapter, xn, ctlq_msg);
+ idpf_vc_xn_unlock(xn);
+ return err;
+ default:
+ dev_err_ratelimited(&adapter->pdev->dev, "Overwriting VC reply (op %d)\n",
+ ctlq_msg->cookie.mbx.chnl_opcode);
+ err = -EBUSY;
+ goto out_unlock;
+ }
+
+ if (ctlq_msg->cookie.mbx.chnl_opcode != xn->vc_op) {
+ dev_err_ratelimited(&adapter->pdev->dev, "Message opcode does not match transaction opcode (msg: %d) (xn: %d)\n",
+ ctlq_msg->cookie.mbx.chnl_opcode, xn->vc_op);
+ xn->reply_sz = 0;
+ xn->state = IDPF_VC_XN_COMPLETED_FAILED;
+ err = -EINVAL;
+ goto out_unlock;
+ }
+
+ if (ctlq_msg->cookie.mbx.chnl_retval) {
+ xn->reply_sz = 0;
+ xn->state = IDPF_VC_XN_COMPLETED_FAILED;
+ err = -EINVAL;
+ goto out_unlock;
+ }
+
+ if (ctlq_msg->data_len) {
+ payload = ctlq_msg->ctx.indirect.payload->va;
+ payload_size = ctlq_msg->ctx.indirect.payload->size;
+ }
+
+ xn->reply_sz = payload_size;
+ xn->state = IDPF_VC_XN_COMPLETED_SUCCESS;
+
+ if (xn->reply.iov_base && xn->reply.iov_len && payload_size)
+ memcpy(xn->reply.iov_base, payload,
+ min_t(size_t, xn->reply.iov_len, payload_size));
+
+out_unlock:
+ idpf_vc_xn_unlock(xn);
+ /* we _cannot_ hold lock while calling complete */
+ complete(&xn->completed);
+
+ return err;
+}
+
/**
* idpf_recv_mb_msg - Receive message over mailbox
* @adapter: Driver specific private structure
@@ -429,6 +806,8 @@ int idpf_recv_mb_msg(struct idpf_adapter *adapter, u32 op,
*/
switch (ctlq_msg.cookie.mbx.chnl_opcode) {
case VIRTCHNL2_OP_VERSION:
+ err = idpf_vc_xn_forward_reply(adapter, &ctlq_msg);
+ break;
case VIRTCHNL2_OP_GET_CAPS:
if (ctlq_msg.cookie.mbx.chnl_retval) {
dev_err(&adapter->pdev->dev, "Failure initializing, vc op: %u retval: %u\n",
@@ -785,7 +1164,11 @@ static int idpf_wait_for_marker_event(struct idpf_vport *vport)
*/
static int idpf_send_ver_msg(struct idpf_adapter *adapter)
{
+ struct idpf_vc_xn_params xn_params = {};
struct virtchnl2_version_info vvi;
+ ssize_t reply_sz;
+ u32 major, minor;
+ int err = 0;
if (adapter->virt_ver_maj) {
vvi.major = cpu_to_le32(adapter->virt_ver_maj);
@@ -795,43 +1178,29 @@ static int idpf_send_ver_msg(struct idpf_adapter *adapter)
vvi.minor = cpu_to_le32(IDPF_VIRTCHNL_VERSION_MINOR);
}
- return idpf_send_mb_msg(adapter, VIRTCHNL2_OP_VERSION, sizeof(vvi),
- (u8 *)&vvi);
-}
-
-/**
- * idpf_recv_ver_msg - Receive virtchnl version message
- * @adapter: Driver specific private structure
- *
- * Receive virtchnl version message. Returns 0 on success, -EAGAIN if we need
- * to send version message again, otherwise negative on failure.
- */
-static int idpf_recv_ver_msg(struct idpf_adapter *adapter)
-{
- struct virtchnl2_version_info vvi;
- u32 major, minor;
- int err;
+ xn_params.vc_op = VIRTCHNL2_OP_VERSION;
+ xn_params.send_buf.iov_base = &vvi;
+ xn_params.send_buf.iov_len = sizeof(vvi);
+ xn_params.recv_buf = xn_params.send_buf;
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
- err = idpf_recv_mb_msg(adapter, VIRTCHNL2_OP_VERSION, &vvi,
- sizeof(vvi));
- if (err)
- return err;
+ reply_sz = idpf_vc_xn_exec(adapter, xn_params);
+ if (reply_sz < 0)
+ return reply_sz;
+ if (reply_sz < sizeof(vvi))
+ return -EIO;
major = le32_to_cpu(vvi.major);
minor = le32_to_cpu(vvi.minor);
if (major > IDPF_VIRTCHNL_VERSION_MAJOR) {
- dev_warn(&adapter->pdev->dev,
- "Virtchnl major version (%d) greater than supported\n",
- major);
-
+ dev_warn(&adapter->pdev->dev, "Virtchnl major version greater than supported\n");
return -EINVAL;
}
if (major == IDPF_VIRTCHNL_VERSION_MAJOR &&
minor > IDPF_VIRTCHNL_VERSION_MINOR)
- dev_warn(&adapter->pdev->dev,
- "Virtchnl minor version (%d) didn't match\n", minor);
+ dev_warn(&adapter->pdev->dev, "Virtchnl minor version didn't match\n");
/* If we have a mismatch, resend version to update receiver on what
* version we will use.
@@ -914,7 +1283,7 @@ static int idpf_send_get_caps_msg(struct idpf_adapter *adapter)
VIRTCHNL2_CAP_LOOPBACK);
return idpf_send_mb_msg(adapter, VIRTCHNL2_OP_GET_CAPS, sizeof(caps),
- (u8 *)&caps);
+ (u8 *)&caps, 0);
}
/**
@@ -1289,7 +1658,7 @@ int idpf_send_create_vport_msg(struct idpf_adapter *adapter,
mutex_lock(&adapter->vc_buf_lock);
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_CREATE_VPORT, buf_size,
- (u8 *)vport_msg);
+ (u8 *)vport_msg, 0);
if (err)
goto rel_lock;
@@ -1375,7 +1744,7 @@ int idpf_send_destroy_vport_msg(struct idpf_vport *vport)
mutex_lock(&vport->vc_buf_lock);
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_DESTROY_VPORT,
- sizeof(v_id), (u8 *)&v_id);
+ sizeof(v_id), (u8 *)&v_id, 0);
if (err)
goto rel_lock;
@@ -1406,7 +1775,7 @@ int idpf_send_enable_vport_msg(struct idpf_vport *vport)
mutex_lock(&vport->vc_buf_lock);
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_ENABLE_VPORT,
- sizeof(v_id), (u8 *)&v_id);
+ sizeof(v_id), (u8 *)&v_id, 0);
if (err)
goto rel_lock;
@@ -1437,7 +1806,7 @@ int idpf_send_disable_vport_msg(struct idpf_vport *vport)
mutex_lock(&vport->vc_buf_lock);
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_DISABLE_VPORT,
- sizeof(v_id), (u8 *)&v_id);
+ sizeof(v_id), (u8 *)&v_id, 0);
if (err)
goto rel_lock;
@@ -1556,7 +1925,7 @@ static int idpf_send_config_tx_queues_msg(struct idpf_vport *vport)
err = idpf_send_mb_msg(vport->adapter,
VIRTCHNL2_OP_CONFIG_TX_QUEUES,
- buf_sz, (u8 *)ctq);
+ buf_sz, (u8 *)ctq, 0);
if (err)
goto mbx_error;
@@ -1708,7 +2077,7 @@ static int idpf_send_config_rx_queues_msg(struct idpf_vport *vport)
err = idpf_send_mb_msg(vport->adapter,
VIRTCHNL2_OP_CONFIG_RX_QUEUES,
- buf_sz, (u8 *)crq);
+ buf_sz, (u8 *)crq, 0);
if (err)
goto mbx_error;
@@ -1875,7 +2244,7 @@ static int idpf_send_ena_dis_queues_msg(struct idpf_vport *vport, u32 vc_op)
qcs = &eq->chunks;
memcpy(qcs->chunks, &qc[k], chunk_sz * num_chunks);
- err = idpf_send_mb_msg(adapter, vc_op, buf_sz, (u8 *)eq);
+ err = idpf_send_mb_msg(adapter, vc_op, buf_sz, (u8 *)eq, 0);
if (err)
goto mbx_error;
@@ -2019,7 +2388,7 @@ int idpf_send_map_unmap_queue_vector_msg(struct idpf_vport *vport, bool map)
if (map) {
err = idpf_send_mb_msg(adapter,
VIRTCHNL2_OP_MAP_QUEUE_VECTOR,
- buf_sz, (u8 *)vqvm);
+ buf_sz, (u8 *)vqvm, 0);
if (!err)
err = idpf_wait_for_event(adapter, vport,
IDPF_VC_MAP_IRQ,
@@ -2027,7 +2396,7 @@ int idpf_send_map_unmap_queue_vector_msg(struct idpf_vport *vport, bool map)
} else {
err = idpf_send_mb_msg(adapter,
VIRTCHNL2_OP_UNMAP_QUEUE_VECTOR,
- buf_sz, (u8 *)vqvm);
+ buf_sz, (u8 *)vqvm, 0);
if (!err)
err =
idpf_min_wait_for_event(adapter, vport,
@@ -2157,7 +2526,7 @@ int idpf_send_delete_queues_msg(struct idpf_vport *vport)
mutex_lock(&vport->vc_buf_lock);
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_DEL_QUEUES,
- buf_size, (u8 *)eq);
+ buf_size, (u8 *)eq, 0);
if (err)
goto rel_lock;
@@ -2221,7 +2590,7 @@ int idpf_send_add_queues_msg(const struct idpf_vport *vport, u16 num_tx_q,
mutex_lock(&((struct idpf_vport *)vport)->vc_buf_lock);
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_ADD_QUEUES,
- sizeof(struct virtchnl2_add_queues), (u8 *)&aq);
+ sizeof(struct virtchnl2_add_queues), (u8 *)&aq, 0);
if (err)
goto rel_lock;
@@ -2280,7 +2649,7 @@ int idpf_send_alloc_vectors_msg(struct idpf_adapter *adapter, u16 num_vectors)
mutex_lock(&adapter->vc_buf_lock);
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_ALLOC_VECTORS,
- sizeof(ac), (u8 *)&ac);
+ sizeof(ac), (u8 *)&ac, 0);
if (err)
goto rel_lock;
@@ -2336,7 +2705,7 @@ int idpf_send_dealloc_vectors_msg(struct idpf_adapter *adapter)
mutex_lock(&adapter->vc_buf_lock);
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_DEALLOC_VECTORS, buf_size,
- (u8 *)vcs);
+ (u8 *)vcs, 0);
if (err)
goto rel_lock;
@@ -2382,7 +2751,7 @@ int idpf_send_set_sriov_vfs_msg(struct idpf_adapter *adapter, u16 num_vfs)
mutex_lock(&adapter->vc_buf_lock);
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_SET_SRIOV_VFS,
- sizeof(svi), (u8 *)&svi);
+ sizeof(svi), (u8 *)&svi, 0);
if (err)
goto rel_lock;
@@ -2420,7 +2789,7 @@ int idpf_send_get_stats_msg(struct idpf_vport *vport)
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_GET_STATS,
sizeof(struct virtchnl2_vport_stats),
- (u8 *)&stats_msg);
+ (u8 *)&stats_msg, 0);
if (err)
goto rel_lock;
@@ -2489,7 +2858,7 @@ int idpf_send_get_set_rss_lut_msg(struct idpf_vport *vport, bool get)
rl->lut[i] = cpu_to_le32(rss_data->rss_lut[i]);
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_SET_RSS_LUT,
- buf_size, (u8 *)rl);
+ buf_size, (u8 *)rl, 0);
if (err)
goto free_mem;
@@ -2500,7 +2869,7 @@ int idpf_send_get_set_rss_lut_msg(struct idpf_vport *vport, bool get)
}
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_GET_RSS_LUT,
- buf_size, (u8 *)rl);
+ buf_size, (u8 *)rl, 0);
if (err)
goto free_mem;
@@ -2559,7 +2928,7 @@ int idpf_send_get_set_rss_key_msg(struct idpf_vport *vport, bool get)
if (get) {
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_GET_RSS_KEY,
- buf_size, (u8 *)rk);
+ buf_size, (u8 *)rk, 0);
if (err)
goto error;
@@ -2591,7 +2960,7 @@ int idpf_send_get_set_rss_key_msg(struct idpf_vport *vport, bool get)
rk->key_flex[i] = rss_data->rss_key[i];
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_SET_RSS_KEY,
- buf_size, (u8 *)rk);
+ buf_size, (u8 *)rk, 0);
if (err)
goto error;
@@ -2688,7 +3057,7 @@ int idpf_send_get_rx_ptype_msg(struct idpf_vport *vport)
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_GET_PTYPE_INFO,
sizeof(struct virtchnl2_get_ptype_info),
- (u8 *)&get_ptype_info);
+ (u8 *)&get_ptype_info, 0);
if (err)
goto vc_buf_unlock;
@@ -2882,7 +3251,7 @@ int idpf_send_ena_dis_loopback_msg(struct idpf_vport *vport)
mutex_lock(&vport->vc_buf_lock);
err = idpf_send_mb_msg(vport->adapter, VIRTCHNL2_OP_LOOPBACK,
- sizeof(loopback), (u8 *)&loopback);
+ sizeof(loopback), (u8 *)&loopback, 0);
if (err)
goto rel_lock;
@@ -2958,7 +3327,7 @@ int idpf_init_dflt_mbx(struct idpf_adapter *adapter)
return -ENOENT;
}
- adapter->state = __IDPF_STARTUP;
+ adapter->state = __IDPF_VER_CHECK;
return 0;
}
@@ -3055,27 +3424,28 @@ int idpf_vc_core_init(struct idpf_adapter *adapter)
u16 num_max_vports;
int err = 0;
+ idpf_vc_xn_init(&adapter->vcxn_mngr);
+
while (adapter->state != __IDPF_INIT_SW) {
switch (adapter->state) {
- case __IDPF_STARTUP:
- if (idpf_send_ver_msg(adapter))
- goto init_failed;
- adapter->state = __IDPF_VER_CHECK;
- goto restart;
case __IDPF_VER_CHECK:
- err = idpf_recv_ver_msg(adapter);
- if (err == -EIO) {
- return err;
- } else if (err == -EAGAIN) {
- adapter->state = __IDPF_STARTUP;
+ err = idpf_send_ver_msg(adapter);
+ switch (err) {
+ case 0:
+ /* success, move state machine forward */
+ adapter->state = __IDPF_GET_CAPS;
+ err = idpf_send_get_caps_msg(adapter);
+ if (err)
+ goto init_failed;
+ fallthrough;
+ case -EAGAIN:
goto restart;
- } else if (err) {
+ default:
+ /* Something bad happened, try again but only a
+ * few times.
+ */
goto init_failed;
}
- if (idpf_send_get_caps_msg(adapter))
- goto init_failed;
- adapter->state = __IDPF_GET_CAPS;
- goto restart;
case __IDPF_GET_CAPS:
if (idpf_recv_get_caps_msg(adapter))
goto init_failed;
@@ -3168,7 +3538,8 @@ int idpf_vc_core_init(struct idpf_adapter *adapter)
* register writes might not have taken effect. Retry to initialize
* the mailbox again
*/
- adapter->state = __IDPF_STARTUP;
+ adapter->state = __IDPF_VER_CHECK;
+ idpf_vc_xn_shutdown(&adapter->vcxn_mngr);
idpf_deinit_dflt_mbx(adapter);
set_bit(IDPF_HR_DRV_LOAD, adapter->flags);
queue_delayed_work(adapter->vc_event_wq, &adapter->vc_event_task,
@@ -3186,6 +3557,7 @@ void idpf_vc_core_deinit(struct idpf_adapter *adapter)
{
int i;
+ idpf_vc_xn_shutdown(&adapter->vcxn_mngr);
idpf_deinit_task(adapter);
idpf_intr_rel(adapter);
/* Set all bits as we dont know on which vc_state the vhnl_wq is
@@ -3739,7 +4111,7 @@ int idpf_add_del_mac_filters(struct idpf_vport *vport,
if (async)
set_bit(mac_flag, vport_config->flags);
- err = idpf_send_mb_msg(adapter, vop, buf_size, (u8 *)ma_list);
+ err = idpf_send_mb_msg(adapter, vop, buf_size, (u8 *)ma_list, 0);
if (err)
goto mbx_error;
@@ -3794,7 +4166,7 @@ int idpf_set_promiscuous(struct idpf_adapter *adapter,
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_CONFIG_PROMISCUOUS_MODE,
sizeof(struct virtchnl2_promisc_info),
- (u8 *)&vpi);
+ (u8 *)&vpi, 0);
return err;
}
--
2.40.1
^ permalink raw reply related [flat|nested] 44+ messages in thread
* [Intel-wired-lan] [PATCH v4 01/10 iwl-next] idpf: implement virtchnl transaction manager
@ 2024-02-06 3:37 ` Alan Brady
0 siblings, 0 replies; 44+ messages in thread
From: Alan Brady @ 2024-02-06 3:37 UTC (permalink / raw)
To: intel-wired-lan
Cc: willemdebruijn.kernel, netdev, Joshua Hay, aleksander.lobakin,
Alan Brady, przemyslaw.kitszel, igor.bagnucki
This starts refactoring how virtchnl messages are handled by adding a
transaction manager (idpf_vc_xn_manager).
There are two primary motivations here which are to enable handling of
multiple messages at once and to make it more robust in general. As it
is right now, the driver may only have one pending message at a time and
there's no guarantee that the response we receive was actually intended
for the message we sent prior.
This works by utilizing a "cookie" field of the message descriptor. It
is arbitrary what data we put in the cookie and the response is required
to have the same cookie the original message was sent with. Then using a
"transaction" abstraction that uses the completion API to pair responses
to the message it belongs to.
The cookie works such that the first half is the index to the
transaction in our array, and the second half is a "salt" that gets
incremented every message. This enables quick lookups into the array and
also ensuring we have the correct message. The salt is necessary because
after, for example, a message times out and we deem the response was
lost for some reason, we could theoretically reuse the same index but
using a different salt ensures that when we do actually get a response
it's not the old message that timed out previously finally coming in.
Since the number of transactions allocated is U8_MAX and the salt is 8
bits, we can never have a conflict because we can't roll over the salt
without using more transactions than we have available.
This starts by only converting the VIRTCHNL2_OP_VERSION message to use
this new transaction API. Follow up patches will convert all virtchnl
messages to use the API.
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Reviewed-by: Igor Bagnucki <igor.bagnucki@intel.com>
Co-developed-by: Joshua Hay <joshua.a.hay@intel.com>
Signed-off-by: Joshua Hay <joshua.a.hay@intel.com>
Signed-off-by: Alan Brady <alan.brady@intel.com>
---
drivers/net/ethernet/intel/idpf/idpf.h | 101 +++-
.../ethernet/intel/idpf/idpf_controlq_api.h | 5 +
drivers/net/ethernet/intel/idpf/idpf_lib.c | 2 +
drivers/net/ethernet/intel/idpf/idpf_main.c | 1 +
drivers/net/ethernet/intel/idpf/idpf_vf_dev.c | 2 +-
.../net/ethernet/intel/idpf/idpf_virtchnl.c | 510 +++++++++++++++---
6 files changed, 548 insertions(+), 73 deletions(-)
diff --git a/drivers/net/ethernet/intel/idpf/idpf.h b/drivers/net/ethernet/intel/idpf/idpf.h
index 0acc125decb3..92103c3cf5a0 100644
--- a/drivers/net/ethernet/intel/idpf/idpf.h
+++ b/drivers/net/ethernet/intel/idpf/idpf.h
@@ -39,6 +39,11 @@ struct idpf_vport_max_q;
((IDPF_CTLQ_MAX_BUF_LEN - (struct_sz)) / (chunk_sz))
#define IDPF_WAIT_FOR_EVENT_TIMEO_MIN 2000
#define IDPF_WAIT_FOR_EVENT_TIMEO 60000
+#define IDPF_VC_XN_MIN_TIMEOUT_MSEC 2000
+#define IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC (60 * 1000)
+#define IDPF_VC_XN_IDX_M GENMASK(7, 0)
+#define IDPF_VC_XN_SALT_M GENMASK(15, 8)
+#define IDPF_VC_XN_RING_LEN U8_MAX
#define IDPF_MAX_WAIT 500
@@ -66,14 +71,12 @@ struct idpf_mac_filter {
/**
* enum idpf_state - State machine to handle bring up
- * @__IDPF_STARTUP: Start the state machine
* @__IDPF_VER_CHECK: Negotiate virtchnl version
* @__IDPF_GET_CAPS: Negotiate capabilities
* @__IDPF_INIT_SW: Init based on given capabilities
* @__IDPF_STATE_LAST: Must be last, used to determine size
*/
enum idpf_state {
- __IDPF_STARTUP,
__IDPF_VER_CHECK,
__IDPF_GET_CAPS,
__IDPF_INIT_SW,
@@ -314,6 +317,96 @@ struct idpf_port_stats {
struct virtchnl2_vport_stats vport_stats;
};
+/**
+ * enum idpf_vc_xn_state - Virtchnl transaction status
+ * @IDPF_VC_XN_IDLE: not expecting a reply, ready to be used
+ * @IDPF_VC_XN_WAITING: expecting a reply, not yet received
+ * @IDPF_VC_XN_COMPLETED_SUCCESS: a reply was expected and received,
+ * buffer updated
+ * @IDPF_VC_XN_COMPLETED_FAILED: a reply was expected and received, but there
+ * was an error, buffer not updated
+ * @IDPF_VC_XN_SHUTDOWN: transaction object cannot be used, VC torn down
+ * @IDPF_VC_XN_ASYNC: transaction sent asynchronously and doesn't have the
+ * return context; a callback may be provided to handle
+ * return
+ */
+enum idpf_vc_xn_state {
+ IDPF_VC_XN_IDLE = 1,
+ IDPF_VC_XN_WAITING,
+ IDPF_VC_XN_COMPLETED_SUCCESS,
+ IDPF_VC_XN_COMPLETED_FAILED,
+ IDPF_VC_XN_SHUTDOWN,
+ IDPF_VC_XN_ASYNC,
+};
+
+struct idpf_vc_xn;
+/* Callback for asynchronous messages */
+typedef int (*async_vc_cb) (struct idpf_adapter *, struct idpf_vc_xn *,
+ const struct idpf_ctlq_msg *);
+
+/**
+ * struct idpf_vc_xn - Data structure representing virtchnl transactions
+ * @completed: virtchnl event loop uses that to signal when a reply is
+ * available, uses kernel completion API
+ * @state: virtchnl event loop stores the data below, protected by the
+ * completion's lock.
+ * @reply_sz: Original size of reply, may be > reply_buf.iov_len; it will be
+ * truncated on its way to the receiver thread according to
+ * reply_buf.iov_len.
+ * @reply: Reference to the buffer(s) where the reply data should be written
+ * to. May be 0-length (then NULL address permitted) if the reply data
+ * should be ignored.
+ * @async_handler: if sent asynchronously, a callback can be provided to handle
+ * the reply when it's received
+ * @vc_op: corresponding opcode sent with this transaction
+ * @idx: index used as retrieval on reply receive, used for cookie
+ * @salt: changed every message to make unique, used for cookie
+ */
+struct idpf_vc_xn {
+ struct completion completed;
+ enum idpf_vc_xn_state state;
+ size_t reply_sz;
+ struct kvec reply;
+ async_vc_cb async_handler;
+ u32 vc_op;
+ u8 idx;
+ u8 salt;
+};
+
+/**
+ * struct idpf_vc_xn_params - Parameters for executing transaction
+ * @send_buf: kvec for send buffer
+ * @recv_buf: kvec for recv buffer, may be NULL, must then have zero length
+ * @timeout_ms: timeout to wait for reply
+ * @async: send message asynchronously, will not wait on completion
+ * @async_handler: If sent asynchronously, optional callback handler. The user
+ * must be careful when using async handlers as the memory for
+ * the recv_buf _cannot_ be on stack if this is async.
+ * @vc_op: virtchnl op to send
+ */
+struct idpf_vc_xn_params {
+ struct kvec send_buf;
+ struct kvec recv_buf;
+ int timeout_ms;
+ bool async;
+ async_vc_cb async_handler;
+ u32 vc_op;
+};
+
+/**
+ * struct idpf_vc_xn_manager - Manager for tracking transactions
+ * @ring: backing and lookup for transactions
+ * @free_xn_bm: bitmap for free transactions
+ * @xn_bm_lock: make bitmap access synchronous where necessary
+ * @salt: used to make cookie unique every message
+ */
+struct idpf_vc_xn_manager {
+ struct idpf_vc_xn ring[IDPF_VC_XN_RING_LEN];
+ DECLARE_BITMAP(free_xn_bm, IDPF_VC_XN_RING_LEN);
+ spinlock_t xn_bm_lock;
+ u8 salt;
+};
+
/**
* struct idpf_vport - Handle for netdevices and queue resources
* @num_txq: Number of allocated TX queues
@@ -604,6 +697,7 @@ struct idpf_vport_config {
* @vchnl_wq: Wait queue for virtchnl messages
* @vc_state: Virtchnl message state
* @vc_msg: Virtchnl message buffer
+ * @vcxn_mngr: Virtchnl transaction manager
* @dev_ops: See idpf_dev_ops
* @num_vfs: Number of allocated VFs through sysfs. PF does not directly talk
* to VFs but is used to initialize them
@@ -663,6 +757,7 @@ struct idpf_adapter {
wait_queue_head_t vchnl_wq;
DECLARE_BITMAP(vc_state, IDPF_VC_NBITS);
char vc_msg[IDPF_CTLQ_MAX_BUF_LEN];
+ struct idpf_vc_xn_manager vcxn_mngr;
struct idpf_dev_ops dev_ops;
int num_vfs;
bool crc_enable;
@@ -939,7 +1034,7 @@ int idpf_get_vec_ids(struct idpf_adapter *adapter,
int idpf_recv_mb_msg(struct idpf_adapter *adapter, u32 op,
void *msg, int msg_size);
int idpf_send_mb_msg(struct idpf_adapter *adapter, u32 op,
- u16 msg_size, u8 *msg);
+ u16 msg_size, u8 *msg, u16 cookie);
void idpf_set_ethtool_ops(struct net_device *netdev);
int idpf_vport_alloc_max_qs(struct idpf_adapter *adapter,
struct idpf_vport_max_q *max_q);
diff --git a/drivers/net/ethernet/intel/idpf/idpf_controlq_api.h b/drivers/net/ethernet/intel/idpf/idpf_controlq_api.h
index 8dee098bbfb0..e8e046ef2f0d 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_controlq_api.h
+++ b/drivers/net/ethernet/intel/idpf/idpf_controlq_api.h
@@ -69,6 +69,11 @@ struct idpf_ctlq_msg {
u8 context[IDPF_INDIRECT_CTX_SIZE];
struct idpf_dma_mem *payload;
} indirect;
+ struct {
+ u32 rsvd;
+ u16 data;
+ u16 flags;
+ } sw_cookie;
} ctx;
};
diff --git a/drivers/net/ethernet/intel/idpf/idpf_lib.c b/drivers/net/ethernet/intel/idpf/idpf_lib.c
index 58179bd733ff..6b4a7408246c 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_lib.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_lib.c
@@ -1823,6 +1823,8 @@ static int idpf_init_hard_reset(struct idpf_adapter *adapter)
goto unlock_mutex;
}
+ queue_delayed_work(adapter->mbx_wq, &adapter->mbx_task, 0);
+
/* Initialize the state machine, also allocate memory and request
* resources
*/
diff --git a/drivers/net/ethernet/intel/idpf/idpf_main.c b/drivers/net/ethernet/intel/idpf/idpf_main.c
index e1febc74cefd..c82233b112bd 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_main.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_main.c
@@ -30,6 +30,7 @@ static void idpf_remove(struct pci_dev *pdev)
idpf_sriov_configure(pdev, 0);
idpf_vc_core_deinit(adapter);
+
/* Be a good citizen and leave the device clean on exit */
adapter->dev_ops.reg_ops.trigger_reset(adapter, IDPF_HR_FUNC_RESET);
idpf_deinit_dflt_mbx(adapter);
diff --git a/drivers/net/ethernet/intel/idpf/idpf_vf_dev.c b/drivers/net/ethernet/intel/idpf/idpf_vf_dev.c
index 8ade4e3a9fe1..192d19a0c89a 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_vf_dev.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_vf_dev.c
@@ -137,7 +137,7 @@ static void idpf_vf_trigger_reset(struct idpf_adapter *adapter,
/* Do not send VIRTCHNL2_OP_RESET_VF message on driver unload */
if (trig_cause == IDPF_HR_FUNC_RESET &&
!test_bit(IDPF_REMOVE_IN_PROG, adapter->flags))
- idpf_send_mb_msg(adapter, VIRTCHNL2_OP_RESET_VF, 0, NULL);
+ idpf_send_mb_msg(adapter, VIRTCHNL2_OP_RESET_VF, 0, NULL, 0);
}
/**
diff --git a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
index d0cdd63b3d5b..92173bcb9fc3 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
@@ -93,13 +93,14 @@ static int idpf_mb_clean(struct idpf_adapter *adapter)
* @op: virtchnl opcode
* @msg_size: size of the payload
* @msg: pointer to buffer holding the payload
+ * @cookie: unique SW generated cookie per message
*
* Will prepare the control queue message and initiates the send api
*
* Returns 0 on success, negative on failure
*/
int idpf_send_mb_msg(struct idpf_adapter *adapter, u32 op,
- u16 msg_size, u8 *msg)
+ u16 msg_size, u8 *msg, u16 cookie)
{
struct idpf_ctlq_msg *ctlq_msg;
struct idpf_dma_mem *dma_mem;
@@ -139,8 +140,12 @@ int idpf_send_mb_msg(struct idpf_adapter *adapter, u32 op,
err = -ENOMEM;
goto dma_alloc_error;
}
- memcpy(dma_mem->va, msg, msg_size);
+
+ /* It's possible we're just sending an opcode but no buffer */
+ if (msg && msg_size)
+ memcpy(dma_mem->va, msg, msg_size);
ctlq_msg->ctx.indirect.payload = dma_mem;
+ ctlq_msg->ctx.sw_cookie.data = cookie;
err = idpf_ctlq_send(&adapter->hw, adapter->hw.asq, 1, ctlq_msg);
if (err)
@@ -361,6 +366,378 @@ static void idpf_recv_vchnl_op(struct idpf_adapter *adapter,
}
}
+/* API for virtchnl "transaction" support ("xn" for short).
+ *
+ * We are reusing the completion lock to serialize the accesses to the
+ * transaction state for simplicity, but it could be its own separate synchro
+ * as well. For now, this API is only used from within a workqueue context;
+ * raw_spin_lock() is enough.
+ */
+/**
+ * idpf_vc_xn_lock - Request exclusive access to vc transaction
+ * @xn: struct idpf_vc_xn* to access
+ */
+#define idpf_vc_xn_lock(xn) \
+ raw_spin_lock(&(xn)->completed.wait.lock)
+
+/**
+ * idpf_vc_xn_unlock - Release exclusive access to vc transaction
+ * @xn: struct idpf_vc_xn* to access
+ */
+#define idpf_vc_xn_unlock(xn) \
+ raw_spin_unlock(&(xn)->completed.wait.lock)
+
+/**
+ * idpf_vc_xn_release_bufs - Release reference to reply buffer(s) and
+ * reset the transaction state.
+ * @xn: struct idpf_vc_xn to update
+ */
+static void idpf_vc_xn_release_bufs(struct idpf_vc_xn *xn)
+{
+ xn->reply.iov_base = NULL;
+ xn->reply.iov_len = 0;
+
+ if (xn->state != IDPF_VC_XN_SHUTDOWN)
+ xn->state = IDPF_VC_XN_IDLE;
+}
+
+/**
+ * idpf_vc_xn_init - Initialize virtchnl transaction object
+ * @vcxn_mngr: pointer to vc transaction manager struct
+ */
+static void idpf_vc_xn_init(struct idpf_vc_xn_manager *vcxn_mngr)
+{
+ int i;
+
+ spin_lock_init(&vcxn_mngr->xn_bm_lock);
+
+ for (i = 0; i < ARRAY_SIZE(vcxn_mngr->ring); i++) {
+ struct idpf_vc_xn *xn = &vcxn_mngr->ring[i];
+
+ xn->state = IDPF_VC_XN_IDLE;
+ xn->idx = i;
+ idpf_vc_xn_release_bufs(xn);
+ init_completion(&xn->completed);
+ }
+
+ bitmap_fill(vcxn_mngr->free_xn_bm, IDPF_VC_XN_RING_LEN);
+}
+
+/**
+ * idpf_vc_xn_shutdown - Uninitialize virtchnl transaction object
+ * @vcxn_mngr: pointer to vc transaction manager struct
+ *
+ * All waiting threads will be woken-up and their transaction aborted. Further
+ * operations on that object will fail.
+ */
+static void idpf_vc_xn_shutdown(struct idpf_vc_xn_manager *vcxn_mngr)
+{
+ int i;
+
+ spin_lock_bh(&vcxn_mngr->xn_bm_lock);
+ bitmap_zero(vcxn_mngr->free_xn_bm, IDPF_VC_XN_RING_LEN);
+ spin_unlock_bh(&vcxn_mngr->xn_bm_lock);
+
+ for (i = 0; i < ARRAY_SIZE(vcxn_mngr->ring); i++) {
+ struct idpf_vc_xn *xn = &vcxn_mngr->ring[i];
+
+ idpf_vc_xn_lock(xn);
+ xn->state = IDPF_VC_XN_SHUTDOWN;
+ idpf_vc_xn_release_bufs(xn);
+ idpf_vc_xn_unlock(xn);
+ complete_all(&xn->completed);
+ }
+}
+
+/**
+ * idpf_vc_xn_pop_free - Pop a free transaction from free list
+ * @vcxn_mngr: transaction manager to pop from
+ *
+ * Returns NULL if no free transactions
+ */
+static
+struct idpf_vc_xn *idpf_vc_xn_pop_free(struct idpf_vc_xn_manager *vcxn_mngr)
+{
+ struct idpf_vc_xn *xn = NULL;
+ unsigned long free_idx;
+
+ spin_lock_bh(&vcxn_mngr->xn_bm_lock);
+ free_idx = find_first_bit(vcxn_mngr->free_xn_bm, IDPF_VC_XN_RING_LEN);
+ if (free_idx == IDPF_VC_XN_RING_LEN)
+ goto do_unlock;
+
+ clear_bit(free_idx, vcxn_mngr->free_xn_bm);
+ xn = &vcxn_mngr->ring[free_idx];
+ xn->salt = vcxn_mngr->salt++;
+
+do_unlock:
+ spin_unlock_bh(&vcxn_mngr->xn_bm_lock);
+
+ return xn;
+}
+
+/**
+ * idpf_vc_xn_push_free - Push a free transaction to free list
+ * @vcxn_mngr: transaction manager to push to
+ * @xn: transaction to push
+ */
+static void idpf_vc_xn_push_free(struct idpf_vc_xn_manager *vcxn_mngr,
+ struct idpf_vc_xn *xn)
+{
+ idpf_vc_xn_release_bufs(xn);
+ set_bit(xn->idx, vcxn_mngr->free_xn_bm);
+}
+
+/**
+ * idpf_vc_xn_exec - Perform a send/recv virtchnl transaction
+ * @adapter: driver specific private structure with vcxn_mngr
+ * @params: parameters for this particular transaction including
+ * -vc_op: virtchannel operation to send
+ * -send_buf: kvec iov for send buf and len
+ * -recv_buf: kvec iov for recv buf and len (ignored if NULL)
+ * -timeout_ms: timeout waiting for a reply (milliseconds)
+ * -async: don't wait for message reply, will lose caller context
+ * -async_handler: callback to handle async replies
+ *
+ * @returns >= 0 for success, the size of the initial reply (may or may not be
+ * >= @recv_buf.iov_len, but we never overflow @@recv_buf_iov_base). < 0 for
+ * error.
+ */
+static ssize_t idpf_vc_xn_exec(struct idpf_adapter *adapter,
+ struct idpf_vc_xn_params params)
+{
+ struct kvec *send_buf = ¶ms.send_buf;
+ struct idpf_vc_xn *xn;
+ ssize_t retval;
+ u16 cookie;
+
+ xn = idpf_vc_xn_pop_free(&adapter->vcxn_mngr);
+ /* no free transactions available */
+ if (!xn)
+ return -ENOSPC;
+
+ idpf_vc_xn_lock(xn);
+ if (xn->state == IDPF_VC_XN_SHUTDOWN) {
+ retval = -ENXIO;
+ goto only_unlock;
+ } else if (xn->state != IDPF_VC_XN_IDLE) {
+ /* We're just going to clobber this transaction even though
+ * it's not IDLE. If we don't reuse it we could theoretically
+ * eventually leak all the free transactions and not be able to
+ * send any messages. At least this way we make an attempt to
+ * remain functional even though something really bad is
+ * happening that's corrupting what was supposed to be free
+ * transactions.
+ */
+ WARN_ONCE(1, "There should only be idle transactions in free list (idx %d op %d)\n",
+ xn->idx, xn->vc_op);
+ }
+
+ xn->reply = params.recv_buf;
+ xn->reply_sz = 0;
+ xn->state = params.async ? IDPF_VC_XN_ASYNC : IDPF_VC_XN_WAITING;
+ xn->vc_op = params.vc_op;
+ xn->async_handler = params.async_handler;
+ idpf_vc_xn_unlock(xn);
+
+ if (!params.async)
+ reinit_completion(&xn->completed);
+ cookie = FIELD_PREP(IDPF_VC_XN_SALT_M, xn->salt) |
+ FIELD_PREP(IDPF_VC_XN_IDX_M, xn->idx);
+
+ retval = idpf_send_mb_msg(adapter, params.vc_op,
+ send_buf->iov_len, send_buf->iov_base,
+ cookie);
+ if (retval) {
+ idpf_vc_xn_lock(xn);
+ goto release_and_unlock;
+ }
+
+ if (params.async)
+ return 0;
+
+ wait_for_completion_timeout(&xn->completed,
+ msecs_to_jiffies(params.timeout_ms));
+
+ /* No need to check the return value; we check the final state of the
+ * transaction below. It's possible the transaction actually gets more
+ * timeout than specified if we get preempted here but after
+ * wait_for_completion_timeout returns. This should be non-issue
+ * however.
+ */
+ idpf_vc_xn_lock(xn);
+ switch (xn->state) {
+ case IDPF_VC_XN_SHUTDOWN:
+ retval = -ENXIO;
+ goto only_unlock;
+ case IDPF_VC_XN_WAITING:
+ dev_notice_ratelimited(&adapter->pdev->dev, "Transaction timed-out (op %d, %dms)\n",
+ params.vc_op, params.timeout_ms);
+ retval = -ETIME;
+ break;
+ case IDPF_VC_XN_COMPLETED_SUCCESS:
+ retval = xn->reply_sz;
+ break;
+ case IDPF_VC_XN_COMPLETED_FAILED:
+ dev_notice_ratelimited(&adapter->pdev->dev, "Transaction failed (op %d)\n",
+ params.vc_op);
+ retval = -EIO;
+ break;
+ default:
+ /* Invalid state. */
+ WARN_ON_ONCE(1);
+ retval = -EIO;
+ break;
+ }
+
+release_and_unlock:
+ idpf_vc_xn_push_free(&adapter->vcxn_mngr, xn);
+ /* If we receive a VC reply after here, it will be dropped. */
+only_unlock:
+ idpf_vc_xn_unlock(xn);
+
+ return retval;
+}
+
+/**
+ * idpf_vc_xn_forward_async - Handle async reply receives
+ * @adapter: private data struct
+ * @xn: transaction to handle
+ * @ctlq_msg: corresponding ctlq_msg
+ *
+ * For async sends we're going to lose the caller's context so, if an
+ * async_handler was provided, it can deal with the reply, otherwise we'll just
+ * check and report if there is an error.
+ */
+static int
+idpf_vc_xn_forward_async(struct idpf_adapter *adapter, struct idpf_vc_xn *xn,
+ const struct idpf_ctlq_msg *ctlq_msg)
+{
+ int err = 0;
+
+ if (ctlq_msg->cookie.mbx.chnl_opcode != xn->vc_op) {
+ dev_err_ratelimited(&adapter->pdev->dev, "Async message opcode does not match transaction opcode (msg: %d) (xn: %d)\n",
+ ctlq_msg->cookie.mbx.chnl_opcode, xn->vc_op);
+ xn->reply_sz = 0;
+ err = -EINVAL;
+ goto release_bufs;
+ }
+
+ if (xn->async_handler) {
+ err = xn->async_handler(adapter, xn, ctlq_msg);
+ goto release_bufs;
+ }
+
+ if (ctlq_msg->cookie.mbx.chnl_retval) {
+ xn->reply_sz = 0;
+ dev_err_ratelimited(&adapter->pdev->dev, "Async message failure (op %d)\n",
+ ctlq_msg->cookie.mbx.chnl_opcode);
+ err = -EINVAL;
+ }
+
+release_bufs:
+ idpf_vc_xn_push_free(&adapter->vcxn_mngr, xn);
+
+ return err;
+}
+
+/**
+ * idpf_vc_xn_forward_reply - copy a reply back to receiving thread
+ * @adapter: driver specific private structure with vcxn_mngr
+ * @ctlq_msg: controlq message to send back to receiving thread
+ */
+static int
+idpf_vc_xn_forward_reply(struct idpf_adapter *adapter,
+ const struct idpf_ctlq_msg *ctlq_msg)
+{
+ const void *payload = NULL;
+ size_t payload_size = 0;
+ struct idpf_vc_xn *xn;
+ u16 msg_info;
+ int err = 0;
+ u16 xn_idx;
+ u16 salt;
+
+ msg_info = ctlq_msg->ctx.sw_cookie.data;
+ xn_idx = FIELD_GET(IDPF_VC_XN_IDX_M, msg_info);
+ if (xn_idx >= ARRAY_SIZE(adapter->vcxn_mngr.ring)) {
+ dev_err_ratelimited(&adapter->pdev->dev, "Out of bounds cookie received: %02x\n",
+ xn_idx);
+ return -EINVAL;
+ }
+ xn = &adapter->vcxn_mngr.ring[xn_idx];
+ salt = FIELD_GET(IDPF_VC_XN_SALT_M, msg_info);
+ if (xn->salt != salt) {
+ dev_err_ratelimited(&adapter->pdev->dev, "Transaction salt does not match (%02x != %02x)\n",
+ xn->salt, salt);
+ return -EINVAL;
+ }
+
+ idpf_vc_xn_lock(xn);
+ switch (xn->state) {
+ case IDPF_VC_XN_WAITING:
+ /* success */
+ break;
+ case IDPF_VC_XN_IDLE:
+ dev_err_ratelimited(&adapter->pdev->dev, "Unexpected or belated VC reply (op %d)\n",
+ ctlq_msg->cookie.mbx.chnl_opcode);
+ err = -EINVAL;
+ goto out_unlock;
+ case IDPF_VC_XN_SHUTDOWN:
+ /* ENXIO is a bit special here as the recv msg loop uses that
+ * know if it should stop trying to clean the ring if we lost
+ * the virtchnl. We need to stop playing with registers and
+ * yield.
+ */
+ err = -ENXIO;
+ goto out_unlock;
+ case IDPF_VC_XN_ASYNC:
+ err = idpf_vc_xn_forward_async(adapter, xn, ctlq_msg);
+ idpf_vc_xn_unlock(xn);
+ return err;
+ default:
+ dev_err_ratelimited(&adapter->pdev->dev, "Overwriting VC reply (op %d)\n",
+ ctlq_msg->cookie.mbx.chnl_opcode);
+ err = -EBUSY;
+ goto out_unlock;
+ }
+
+ if (ctlq_msg->cookie.mbx.chnl_opcode != xn->vc_op) {
+ dev_err_ratelimited(&adapter->pdev->dev, "Message opcode does not match transaction opcode (msg: %d) (xn: %d)\n",
+ ctlq_msg->cookie.mbx.chnl_opcode, xn->vc_op);
+ xn->reply_sz = 0;
+ xn->state = IDPF_VC_XN_COMPLETED_FAILED;
+ err = -EINVAL;
+ goto out_unlock;
+ }
+
+ if (ctlq_msg->cookie.mbx.chnl_retval) {
+ xn->reply_sz = 0;
+ xn->state = IDPF_VC_XN_COMPLETED_FAILED;
+ err = -EINVAL;
+ goto out_unlock;
+ }
+
+ if (ctlq_msg->data_len) {
+ payload = ctlq_msg->ctx.indirect.payload->va;
+ payload_size = ctlq_msg->ctx.indirect.payload->size;
+ }
+
+ xn->reply_sz = payload_size;
+ xn->state = IDPF_VC_XN_COMPLETED_SUCCESS;
+
+ if (xn->reply.iov_base && xn->reply.iov_len && payload_size)
+ memcpy(xn->reply.iov_base, payload,
+ min_t(size_t, xn->reply.iov_len, payload_size));
+
+out_unlock:
+ idpf_vc_xn_unlock(xn);
+ /* we _cannot_ hold lock while calling complete */
+ complete(&xn->completed);
+
+ return err;
+}
+
/**
* idpf_recv_mb_msg - Receive message over mailbox
* @adapter: Driver specific private structure
@@ -429,6 +806,8 @@ int idpf_recv_mb_msg(struct idpf_adapter *adapter, u32 op,
*/
switch (ctlq_msg.cookie.mbx.chnl_opcode) {
case VIRTCHNL2_OP_VERSION:
+ err = idpf_vc_xn_forward_reply(adapter, &ctlq_msg);
+ break;
case VIRTCHNL2_OP_GET_CAPS:
if (ctlq_msg.cookie.mbx.chnl_retval) {
dev_err(&adapter->pdev->dev, "Failure initializing, vc op: %u retval: %u\n",
@@ -785,7 +1164,11 @@ static int idpf_wait_for_marker_event(struct idpf_vport *vport)
*/
static int idpf_send_ver_msg(struct idpf_adapter *adapter)
{
+ struct idpf_vc_xn_params xn_params = {};
struct virtchnl2_version_info vvi;
+ ssize_t reply_sz;
+ u32 major, minor;
+ int err = 0;
if (adapter->virt_ver_maj) {
vvi.major = cpu_to_le32(adapter->virt_ver_maj);
@@ -795,43 +1178,29 @@ static int idpf_send_ver_msg(struct idpf_adapter *adapter)
vvi.minor = cpu_to_le32(IDPF_VIRTCHNL_VERSION_MINOR);
}
- return idpf_send_mb_msg(adapter, VIRTCHNL2_OP_VERSION, sizeof(vvi),
- (u8 *)&vvi);
-}
-
-/**
- * idpf_recv_ver_msg - Receive virtchnl version message
- * @adapter: Driver specific private structure
- *
- * Receive virtchnl version message. Returns 0 on success, -EAGAIN if we need
- * to send version message again, otherwise negative on failure.
- */
-static int idpf_recv_ver_msg(struct idpf_adapter *adapter)
-{
- struct virtchnl2_version_info vvi;
- u32 major, minor;
- int err;
+ xn_params.vc_op = VIRTCHNL2_OP_VERSION;
+ xn_params.send_buf.iov_base = &vvi;
+ xn_params.send_buf.iov_len = sizeof(vvi);
+ xn_params.recv_buf = xn_params.send_buf;
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
- err = idpf_recv_mb_msg(adapter, VIRTCHNL2_OP_VERSION, &vvi,
- sizeof(vvi));
- if (err)
- return err;
+ reply_sz = idpf_vc_xn_exec(adapter, xn_params);
+ if (reply_sz < 0)
+ return reply_sz;
+ if (reply_sz < sizeof(vvi))
+ return -EIO;
major = le32_to_cpu(vvi.major);
minor = le32_to_cpu(vvi.minor);
if (major > IDPF_VIRTCHNL_VERSION_MAJOR) {
- dev_warn(&adapter->pdev->dev,
- "Virtchnl major version (%d) greater than supported\n",
- major);
-
+ dev_warn(&adapter->pdev->dev, "Virtchnl major version greater than supported\n");
return -EINVAL;
}
if (major == IDPF_VIRTCHNL_VERSION_MAJOR &&
minor > IDPF_VIRTCHNL_VERSION_MINOR)
- dev_warn(&adapter->pdev->dev,
- "Virtchnl minor version (%d) didn't match\n", minor);
+ dev_warn(&adapter->pdev->dev, "Virtchnl minor version didn't match\n");
/* If we have a mismatch, resend version to update receiver on what
* version we will use.
@@ -914,7 +1283,7 @@ static int idpf_send_get_caps_msg(struct idpf_adapter *adapter)
VIRTCHNL2_CAP_LOOPBACK);
return idpf_send_mb_msg(adapter, VIRTCHNL2_OP_GET_CAPS, sizeof(caps),
- (u8 *)&caps);
+ (u8 *)&caps, 0);
}
/**
@@ -1289,7 +1658,7 @@ int idpf_send_create_vport_msg(struct idpf_adapter *adapter,
mutex_lock(&adapter->vc_buf_lock);
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_CREATE_VPORT, buf_size,
- (u8 *)vport_msg);
+ (u8 *)vport_msg, 0);
if (err)
goto rel_lock;
@@ -1375,7 +1744,7 @@ int idpf_send_destroy_vport_msg(struct idpf_vport *vport)
mutex_lock(&vport->vc_buf_lock);
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_DESTROY_VPORT,
- sizeof(v_id), (u8 *)&v_id);
+ sizeof(v_id), (u8 *)&v_id, 0);
if (err)
goto rel_lock;
@@ -1406,7 +1775,7 @@ int idpf_send_enable_vport_msg(struct idpf_vport *vport)
mutex_lock(&vport->vc_buf_lock);
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_ENABLE_VPORT,
- sizeof(v_id), (u8 *)&v_id);
+ sizeof(v_id), (u8 *)&v_id, 0);
if (err)
goto rel_lock;
@@ -1437,7 +1806,7 @@ int idpf_send_disable_vport_msg(struct idpf_vport *vport)
mutex_lock(&vport->vc_buf_lock);
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_DISABLE_VPORT,
- sizeof(v_id), (u8 *)&v_id);
+ sizeof(v_id), (u8 *)&v_id, 0);
if (err)
goto rel_lock;
@@ -1556,7 +1925,7 @@ static int idpf_send_config_tx_queues_msg(struct idpf_vport *vport)
err = idpf_send_mb_msg(vport->adapter,
VIRTCHNL2_OP_CONFIG_TX_QUEUES,
- buf_sz, (u8 *)ctq);
+ buf_sz, (u8 *)ctq, 0);
if (err)
goto mbx_error;
@@ -1708,7 +2077,7 @@ static int idpf_send_config_rx_queues_msg(struct idpf_vport *vport)
err = idpf_send_mb_msg(vport->adapter,
VIRTCHNL2_OP_CONFIG_RX_QUEUES,
- buf_sz, (u8 *)crq);
+ buf_sz, (u8 *)crq, 0);
if (err)
goto mbx_error;
@@ -1875,7 +2244,7 @@ static int idpf_send_ena_dis_queues_msg(struct idpf_vport *vport, u32 vc_op)
qcs = &eq->chunks;
memcpy(qcs->chunks, &qc[k], chunk_sz * num_chunks);
- err = idpf_send_mb_msg(adapter, vc_op, buf_sz, (u8 *)eq);
+ err = idpf_send_mb_msg(adapter, vc_op, buf_sz, (u8 *)eq, 0);
if (err)
goto mbx_error;
@@ -2019,7 +2388,7 @@ int idpf_send_map_unmap_queue_vector_msg(struct idpf_vport *vport, bool map)
if (map) {
err = idpf_send_mb_msg(adapter,
VIRTCHNL2_OP_MAP_QUEUE_VECTOR,
- buf_sz, (u8 *)vqvm);
+ buf_sz, (u8 *)vqvm, 0);
if (!err)
err = idpf_wait_for_event(adapter, vport,
IDPF_VC_MAP_IRQ,
@@ -2027,7 +2396,7 @@ int idpf_send_map_unmap_queue_vector_msg(struct idpf_vport *vport, bool map)
} else {
err = idpf_send_mb_msg(adapter,
VIRTCHNL2_OP_UNMAP_QUEUE_VECTOR,
- buf_sz, (u8 *)vqvm);
+ buf_sz, (u8 *)vqvm, 0);
if (!err)
err =
idpf_min_wait_for_event(adapter, vport,
@@ -2157,7 +2526,7 @@ int idpf_send_delete_queues_msg(struct idpf_vport *vport)
mutex_lock(&vport->vc_buf_lock);
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_DEL_QUEUES,
- buf_size, (u8 *)eq);
+ buf_size, (u8 *)eq, 0);
if (err)
goto rel_lock;
@@ -2221,7 +2590,7 @@ int idpf_send_add_queues_msg(const struct idpf_vport *vport, u16 num_tx_q,
mutex_lock(&((struct idpf_vport *)vport)->vc_buf_lock);
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_ADD_QUEUES,
- sizeof(struct virtchnl2_add_queues), (u8 *)&aq);
+ sizeof(struct virtchnl2_add_queues), (u8 *)&aq, 0);
if (err)
goto rel_lock;
@@ -2280,7 +2649,7 @@ int idpf_send_alloc_vectors_msg(struct idpf_adapter *adapter, u16 num_vectors)
mutex_lock(&adapter->vc_buf_lock);
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_ALLOC_VECTORS,
- sizeof(ac), (u8 *)&ac);
+ sizeof(ac), (u8 *)&ac, 0);
if (err)
goto rel_lock;
@@ -2336,7 +2705,7 @@ int idpf_send_dealloc_vectors_msg(struct idpf_adapter *adapter)
mutex_lock(&adapter->vc_buf_lock);
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_DEALLOC_VECTORS, buf_size,
- (u8 *)vcs);
+ (u8 *)vcs, 0);
if (err)
goto rel_lock;
@@ -2382,7 +2751,7 @@ int idpf_send_set_sriov_vfs_msg(struct idpf_adapter *adapter, u16 num_vfs)
mutex_lock(&adapter->vc_buf_lock);
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_SET_SRIOV_VFS,
- sizeof(svi), (u8 *)&svi);
+ sizeof(svi), (u8 *)&svi, 0);
if (err)
goto rel_lock;
@@ -2420,7 +2789,7 @@ int idpf_send_get_stats_msg(struct idpf_vport *vport)
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_GET_STATS,
sizeof(struct virtchnl2_vport_stats),
- (u8 *)&stats_msg);
+ (u8 *)&stats_msg, 0);
if (err)
goto rel_lock;
@@ -2489,7 +2858,7 @@ int idpf_send_get_set_rss_lut_msg(struct idpf_vport *vport, bool get)
rl->lut[i] = cpu_to_le32(rss_data->rss_lut[i]);
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_SET_RSS_LUT,
- buf_size, (u8 *)rl);
+ buf_size, (u8 *)rl, 0);
if (err)
goto free_mem;
@@ -2500,7 +2869,7 @@ int idpf_send_get_set_rss_lut_msg(struct idpf_vport *vport, bool get)
}
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_GET_RSS_LUT,
- buf_size, (u8 *)rl);
+ buf_size, (u8 *)rl, 0);
if (err)
goto free_mem;
@@ -2559,7 +2928,7 @@ int idpf_send_get_set_rss_key_msg(struct idpf_vport *vport, bool get)
if (get) {
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_GET_RSS_KEY,
- buf_size, (u8 *)rk);
+ buf_size, (u8 *)rk, 0);
if (err)
goto error;
@@ -2591,7 +2960,7 @@ int idpf_send_get_set_rss_key_msg(struct idpf_vport *vport, bool get)
rk->key_flex[i] = rss_data->rss_key[i];
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_SET_RSS_KEY,
- buf_size, (u8 *)rk);
+ buf_size, (u8 *)rk, 0);
if (err)
goto error;
@@ -2688,7 +3057,7 @@ int idpf_send_get_rx_ptype_msg(struct idpf_vport *vport)
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_GET_PTYPE_INFO,
sizeof(struct virtchnl2_get_ptype_info),
- (u8 *)&get_ptype_info);
+ (u8 *)&get_ptype_info, 0);
if (err)
goto vc_buf_unlock;
@@ -2882,7 +3251,7 @@ int idpf_send_ena_dis_loopback_msg(struct idpf_vport *vport)
mutex_lock(&vport->vc_buf_lock);
err = idpf_send_mb_msg(vport->adapter, VIRTCHNL2_OP_LOOPBACK,
- sizeof(loopback), (u8 *)&loopback);
+ sizeof(loopback), (u8 *)&loopback, 0);
if (err)
goto rel_lock;
@@ -2958,7 +3327,7 @@ int idpf_init_dflt_mbx(struct idpf_adapter *adapter)
return -ENOENT;
}
- adapter->state = __IDPF_STARTUP;
+ adapter->state = __IDPF_VER_CHECK;
return 0;
}
@@ -3055,27 +3424,28 @@ int idpf_vc_core_init(struct idpf_adapter *adapter)
u16 num_max_vports;
int err = 0;
+ idpf_vc_xn_init(&adapter->vcxn_mngr);
+
while (adapter->state != __IDPF_INIT_SW) {
switch (adapter->state) {
- case __IDPF_STARTUP:
- if (idpf_send_ver_msg(adapter))
- goto init_failed;
- adapter->state = __IDPF_VER_CHECK;
- goto restart;
case __IDPF_VER_CHECK:
- err = idpf_recv_ver_msg(adapter);
- if (err == -EIO) {
- return err;
- } else if (err == -EAGAIN) {
- adapter->state = __IDPF_STARTUP;
+ err = idpf_send_ver_msg(adapter);
+ switch (err) {
+ case 0:
+ /* success, move state machine forward */
+ adapter->state = __IDPF_GET_CAPS;
+ err = idpf_send_get_caps_msg(adapter);
+ if (err)
+ goto init_failed;
+ fallthrough;
+ case -EAGAIN:
goto restart;
- } else if (err) {
+ default:
+ /* Something bad happened, try again but only a
+ * few times.
+ */
goto init_failed;
}
- if (idpf_send_get_caps_msg(adapter))
- goto init_failed;
- adapter->state = __IDPF_GET_CAPS;
- goto restart;
case __IDPF_GET_CAPS:
if (idpf_recv_get_caps_msg(adapter))
goto init_failed;
@@ -3168,7 +3538,8 @@ int idpf_vc_core_init(struct idpf_adapter *adapter)
* register writes might not have taken effect. Retry to initialize
* the mailbox again
*/
- adapter->state = __IDPF_STARTUP;
+ adapter->state = __IDPF_VER_CHECK;
+ idpf_vc_xn_shutdown(&adapter->vcxn_mngr);
idpf_deinit_dflt_mbx(adapter);
set_bit(IDPF_HR_DRV_LOAD, adapter->flags);
queue_delayed_work(adapter->vc_event_wq, &adapter->vc_event_task,
@@ -3186,6 +3557,7 @@ void idpf_vc_core_deinit(struct idpf_adapter *adapter)
{
int i;
+ idpf_vc_xn_shutdown(&adapter->vcxn_mngr);
idpf_deinit_task(adapter);
idpf_intr_rel(adapter);
/* Set all bits as we dont know on which vc_state the vhnl_wq is
@@ -3739,7 +4111,7 @@ int idpf_add_del_mac_filters(struct idpf_vport *vport,
if (async)
set_bit(mac_flag, vport_config->flags);
- err = idpf_send_mb_msg(adapter, vop, buf_size, (u8 *)ma_list);
+ err = idpf_send_mb_msg(adapter, vop, buf_size, (u8 *)ma_list, 0);
if (err)
goto mbx_error;
@@ -3794,7 +4166,7 @@ int idpf_set_promiscuous(struct idpf_adapter *adapter,
err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_CONFIG_PROMISCUOUS_MODE,
sizeof(struct virtchnl2_promisc_info),
- (u8 *)&vpi);
+ (u8 *)&vpi, 0);
return err;
}
--
2.40.1
^ permalink raw reply related [flat|nested] 44+ messages in thread
* [PATCH v4 02/10 iwl-next] idpf: refactor vport virtchnl messages
2024-02-06 3:37 ` [Intel-wired-lan] " Alan Brady
@ 2024-02-06 3:37 ` Alan Brady
-1 siblings, 0 replies; 44+ messages in thread
From: Alan Brady @ 2024-02-06 3:37 UTC (permalink / raw)
To: intel-wired-lan
Cc: netdev, willemdebruijn.kernel, przemyslaw.kitszel, igor.bagnucki,
aleksander.lobakin, Alan Brady, Joshua Hay
This reworks the way vport related virtchnl messages work to take
advantage of the added transaction API. It is fairly mechanical as, to
use the transaction API, the function just needs to fill out an
appropriate idpf_vc_xn_params struct to pass to idpf_vc_xn_exec which
will take care of the actual send and recv.
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Reviewed-by: Igor Bagnucki <igor.bagnucki@intel.com>
Co-developed-by: Joshua Hay <joshua.a.hay@intel.com>
Signed-off-by: Joshua Hay <joshua.a.hay@intel.com>
Signed-off-by: Alan Brady <alan.brady@intel.com>
---
.../net/ethernet/intel/idpf/idpf_virtchnl.c | 185 +++++++-----------
1 file changed, 69 insertions(+), 116 deletions(-)
diff --git a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
index 92173bcb9fc3..2f8d44f3ce69 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
@@ -761,7 +761,6 @@ int idpf_recv_mb_msg(struct idpf_adapter *adapter, u32 op,
while (1) {
struct idpf_vport_config *vport_config;
- int payload_size = 0;
/* Try to get one message */
num_q_msg = 1;
@@ -798,47 +797,17 @@ int idpf_recv_mb_msg(struct idpf_adapter *adapter, u32 op,
if (err)
goto post_buffs;
- if (ctlq_msg.data_len)
- payload_size = ctlq_msg.ctx.indirect.payload->size;
-
/* All conditions are met. Either a message requested is
* received or we received a message to be processed
*/
switch (ctlq_msg.cookie.mbx.chnl_opcode) {
case VIRTCHNL2_OP_VERSION:
- err = idpf_vc_xn_forward_reply(adapter, &ctlq_msg);
- break;
case VIRTCHNL2_OP_GET_CAPS:
- if (ctlq_msg.cookie.mbx.chnl_retval) {
- dev_err(&adapter->pdev->dev, "Failure initializing, vc op: %u retval: %u\n",
- ctlq_msg.cookie.mbx.chnl_opcode,
- ctlq_msg.cookie.mbx.chnl_retval);
- err = -EBADMSG;
- } else if (msg) {
- memcpy(msg, ctlq_msg.ctx.indirect.payload->va,
- min_t(int, payload_size, msg_size));
- }
- work_done = true;
- break;
case VIRTCHNL2_OP_CREATE_VPORT:
- idpf_recv_vchnl_op(adapter, NULL, &ctlq_msg,
- IDPF_VC_CREATE_VPORT,
- IDPF_VC_CREATE_VPORT_ERR);
- break;
case VIRTCHNL2_OP_ENABLE_VPORT:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_ENA_VPORT,
- IDPF_VC_ENA_VPORT_ERR);
- break;
case VIRTCHNL2_OP_DISABLE_VPORT:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_DIS_VPORT,
- IDPF_VC_DIS_VPORT_ERR);
- break;
case VIRTCHNL2_OP_DESTROY_VPORT:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_DESTROY_VPORT,
- IDPF_VC_DESTROY_VPORT_ERR);
+ err = idpf_vc_xn_forward_reply(adapter, &ctlq_msg);
break;
case VIRTCHNL2_OP_CONFIG_TX_QUEUES:
idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
@@ -1225,7 +1194,9 @@ static int idpf_send_ver_msg(struct idpf_adapter *adapter)
*/
static int idpf_send_get_caps_msg(struct idpf_adapter *adapter)
{
- struct virtchnl2_get_capabilities caps = { };
+ struct virtchnl2_get_capabilities caps = {};
+ struct idpf_vc_xn_params xn_params = {};
+ ssize_t reply_sz;
caps.csum_caps =
cpu_to_le32(VIRTCHNL2_CAP_TX_CSUM_L3_IPV4 |
@@ -1282,21 +1253,20 @@ static int idpf_send_get_caps_msg(struct idpf_adapter *adapter)
VIRTCHNL2_CAP_PROMISC |
VIRTCHNL2_CAP_LOOPBACK);
- return idpf_send_mb_msg(adapter, VIRTCHNL2_OP_GET_CAPS, sizeof(caps),
- (u8 *)&caps, 0);
-}
+ xn_params.vc_op = VIRTCHNL2_OP_GET_CAPS;
+ xn_params.send_buf.iov_base = ∩︀
+ xn_params.send_buf.iov_len = sizeof(caps);
+ xn_params.recv_buf.iov_base = &adapter->caps;
+ xn_params.recv_buf.iov_len = sizeof(adapter->caps);
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
-/**
- * idpf_recv_get_caps_msg - Receive virtchnl get capabilities message
- * @adapter: Driver specific private structure
- *
- * Receive virtchnl get capabilities message. Returns 0 on success, negative on
- * failure.
- */
-static int idpf_recv_get_caps_msg(struct idpf_adapter *adapter)
-{
- return idpf_recv_mb_msg(adapter, VIRTCHNL2_OP_GET_CAPS, &adapter->caps,
- sizeof(struct virtchnl2_get_capabilities));
+ reply_sz = idpf_vc_xn_exec(adapter, xn_params);
+ if (reply_sz < 0)
+ return reply_sz;
+ if (reply_sz < sizeof(adapter->caps))
+ return -EIO;
+
+ return 0;
}
/**
@@ -1623,8 +1593,10 @@ int idpf_send_create_vport_msg(struct idpf_adapter *adapter,
struct idpf_vport_max_q *max_q)
{
struct virtchnl2_create_vport *vport_msg;
+ struct idpf_vc_xn_params xn_params = {};
u16 idx = adapter->next_vport;
int err, buf_size;
+ ssize_t reply_sz;
buf_size = sizeof(struct virtchnl2_create_vport);
if (!adapter->vport_params_reqd[idx]) {
@@ -1655,35 +1627,38 @@ int idpf_send_create_vport_msg(struct idpf_adapter *adapter,
return err;
}
- mutex_lock(&adapter->vc_buf_lock);
-
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_CREATE_VPORT, buf_size,
- (u8 *)vport_msg, 0);
- if (err)
- goto rel_lock;
-
- err = idpf_wait_for_event(adapter, NULL, IDPF_VC_CREATE_VPORT,
- IDPF_VC_CREATE_VPORT_ERR);
- if (err) {
- dev_err(&adapter->pdev->dev, "Failed to receive create vport message");
-
- goto rel_lock;
- }
-
if (!adapter->vport_params_recvd[idx]) {
adapter->vport_params_recvd[idx] = kzalloc(IDPF_CTLQ_MAX_BUF_LEN,
GFP_KERNEL);
if (!adapter->vport_params_recvd[idx]) {
err = -ENOMEM;
- goto rel_lock;
+ goto free_vport_params;
}
}
- vport_msg = adapter->vport_params_recvd[idx];
- memcpy(vport_msg, adapter->vc_msg, IDPF_CTLQ_MAX_BUF_LEN);
+ xn_params.vc_op = VIRTCHNL2_OP_CREATE_VPORT;
+ xn_params.send_buf.iov_base = vport_msg;
+ xn_params.send_buf.iov_len = buf_size;
+ xn_params.recv_buf.iov_base = adapter->vport_params_recvd[idx];
+ xn_params.recv_buf.iov_len = IDPF_CTLQ_MAX_BUF_LEN;
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
+ reply_sz = idpf_vc_xn_exec(adapter, xn_params);
+ if (reply_sz < 0) {
+ err = reply_sz;
+ goto free_vport_params;
+ }
+ if (reply_sz < IDPF_CTLQ_MAX_BUF_LEN) {
+ err = -EIO;
+ goto free_vport_params;
+ }
-rel_lock:
- mutex_unlock(&adapter->vc_buf_lock);
+ return 0;
+
+free_vport_params:
+ kfree(adapter->vport_params_recvd[idx]);
+ adapter->vport_params_recvd[idx] = NULL;
+ kfree(adapter->vport_params_reqd[idx]);
+ adapter->vport_params_reqd[idx] = NULL;
return err;
}
@@ -1735,26 +1710,19 @@ int idpf_check_supported_desc_ids(struct idpf_vport *vport)
*/
int idpf_send_destroy_vport_msg(struct idpf_vport *vport)
{
- struct idpf_adapter *adapter = vport->adapter;
+ struct idpf_vc_xn_params xn_params = {};
struct virtchnl2_vport v_id;
- int err;
+ ssize_t reply_sz;
v_id.vport_id = cpu_to_le32(vport->vport_id);
- mutex_lock(&vport->vc_buf_lock);
-
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_DESTROY_VPORT,
- sizeof(v_id), (u8 *)&v_id, 0);
- if (err)
- goto rel_lock;
-
- err = idpf_min_wait_for_event(adapter, vport, IDPF_VC_DESTROY_VPORT,
- IDPF_VC_DESTROY_VPORT_ERR);
-
-rel_lock:
- mutex_unlock(&vport->vc_buf_lock);
+ xn_params.vc_op = VIRTCHNL2_OP_DESTROY_VPORT;
+ xn_params.send_buf.iov_base = &v_id;
+ xn_params.send_buf.iov_len = sizeof(v_id);
+ xn_params.timeout_ms = IDPF_VC_XN_MIN_TIMEOUT_MSEC;
+ reply_sz = idpf_vc_xn_exec(vport->adapter, xn_params);
- return err;
+ return reply_sz < 0 ? reply_sz : 0;
}
/**
@@ -1766,26 +1734,19 @@ int idpf_send_destroy_vport_msg(struct idpf_vport *vport)
*/
int idpf_send_enable_vport_msg(struct idpf_vport *vport)
{
- struct idpf_adapter *adapter = vport->adapter;
+ struct idpf_vc_xn_params xn_params = {};
struct virtchnl2_vport v_id;
- int err;
+ ssize_t reply_sz;
v_id.vport_id = cpu_to_le32(vport->vport_id);
- mutex_lock(&vport->vc_buf_lock);
-
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_ENABLE_VPORT,
- sizeof(v_id), (u8 *)&v_id, 0);
- if (err)
- goto rel_lock;
-
- err = idpf_wait_for_event(adapter, vport, IDPF_VC_ENA_VPORT,
- IDPF_VC_ENA_VPORT_ERR);
-
-rel_lock:
- mutex_unlock(&vport->vc_buf_lock);
+ xn_params.vc_op = VIRTCHNL2_OP_ENABLE_VPORT;
+ xn_params.send_buf.iov_base = &v_id;
+ xn_params.send_buf.iov_len = sizeof(v_id);
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
+ reply_sz = idpf_vc_xn_exec(vport->adapter, xn_params);
- return err;
+ return reply_sz < 0 ? reply_sz : 0;
}
/**
@@ -1797,26 +1758,19 @@ int idpf_send_enable_vport_msg(struct idpf_vport *vport)
*/
int idpf_send_disable_vport_msg(struct idpf_vport *vport)
{
- struct idpf_adapter *adapter = vport->adapter;
+ struct idpf_vc_xn_params xn_params = {};
struct virtchnl2_vport v_id;
- int err;
+ ssize_t reply_sz;
v_id.vport_id = cpu_to_le32(vport->vport_id);
- mutex_lock(&vport->vc_buf_lock);
-
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_DISABLE_VPORT,
- sizeof(v_id), (u8 *)&v_id, 0);
- if (err)
- goto rel_lock;
-
- err = idpf_min_wait_for_event(adapter, vport, IDPF_VC_DIS_VPORT,
- IDPF_VC_DIS_VPORT_ERR);
+ xn_params.vc_op = VIRTCHNL2_OP_DISABLE_VPORT;
+ xn_params.send_buf.iov_base = &v_id;
+ xn_params.send_buf.iov_len = sizeof(v_id);
+ xn_params.timeout_ms = IDPF_VC_XN_MIN_TIMEOUT_MSEC;
+ reply_sz = idpf_vc_xn_exec(vport->adapter, xn_params);
-rel_lock:
- mutex_unlock(&vport->vc_buf_lock);
-
- return err;
+ return reply_sz < 0 ? reply_sz : 0;
}
/**
@@ -3434,9 +3388,6 @@ int idpf_vc_core_init(struct idpf_adapter *adapter)
case 0:
/* success, move state machine forward */
adapter->state = __IDPF_GET_CAPS;
- err = idpf_send_get_caps_msg(adapter);
- if (err)
- goto init_failed;
fallthrough;
case -EAGAIN:
goto restart;
@@ -3447,13 +3398,15 @@ int idpf_vc_core_init(struct idpf_adapter *adapter)
goto init_failed;
}
case __IDPF_GET_CAPS:
- if (idpf_recv_get_caps_msg(adapter))
+ err = idpf_send_get_caps_msg(adapter);
+ if (err)
goto init_failed;
adapter->state = __IDPF_INIT_SW;
break;
default:
dev_err(&adapter->pdev->dev, "Device is in bad state: %d\n",
adapter->state);
+ err = -EINVAL;
goto init_failed;
}
break;
--
2.40.1
^ permalink raw reply related [flat|nested] 44+ messages in thread
* [Intel-wired-lan] [PATCH v4 02/10 iwl-next] idpf: refactor vport virtchnl messages
@ 2024-02-06 3:37 ` Alan Brady
0 siblings, 0 replies; 44+ messages in thread
From: Alan Brady @ 2024-02-06 3:37 UTC (permalink / raw)
To: intel-wired-lan
Cc: willemdebruijn.kernel, netdev, Joshua Hay, aleksander.lobakin,
Alan Brady, przemyslaw.kitszel, igor.bagnucki
This reworks the way vport related virtchnl messages work to take
advantage of the added transaction API. It is fairly mechanical as, to
use the transaction API, the function just needs to fill out an
appropriate idpf_vc_xn_params struct to pass to idpf_vc_xn_exec which
will take care of the actual send and recv.
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Reviewed-by: Igor Bagnucki <igor.bagnucki@intel.com>
Co-developed-by: Joshua Hay <joshua.a.hay@intel.com>
Signed-off-by: Joshua Hay <joshua.a.hay@intel.com>
Signed-off-by: Alan Brady <alan.brady@intel.com>
---
.../net/ethernet/intel/idpf/idpf_virtchnl.c | 185 +++++++-----------
1 file changed, 69 insertions(+), 116 deletions(-)
diff --git a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
index 92173bcb9fc3..2f8d44f3ce69 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
@@ -761,7 +761,6 @@ int idpf_recv_mb_msg(struct idpf_adapter *adapter, u32 op,
while (1) {
struct idpf_vport_config *vport_config;
- int payload_size = 0;
/* Try to get one message */
num_q_msg = 1;
@@ -798,47 +797,17 @@ int idpf_recv_mb_msg(struct idpf_adapter *adapter, u32 op,
if (err)
goto post_buffs;
- if (ctlq_msg.data_len)
- payload_size = ctlq_msg.ctx.indirect.payload->size;
-
/* All conditions are met. Either a message requested is
* received or we received a message to be processed
*/
switch (ctlq_msg.cookie.mbx.chnl_opcode) {
case VIRTCHNL2_OP_VERSION:
- err = idpf_vc_xn_forward_reply(adapter, &ctlq_msg);
- break;
case VIRTCHNL2_OP_GET_CAPS:
- if (ctlq_msg.cookie.mbx.chnl_retval) {
- dev_err(&adapter->pdev->dev, "Failure initializing, vc op: %u retval: %u\n",
- ctlq_msg.cookie.mbx.chnl_opcode,
- ctlq_msg.cookie.mbx.chnl_retval);
- err = -EBADMSG;
- } else if (msg) {
- memcpy(msg, ctlq_msg.ctx.indirect.payload->va,
- min_t(int, payload_size, msg_size));
- }
- work_done = true;
- break;
case VIRTCHNL2_OP_CREATE_VPORT:
- idpf_recv_vchnl_op(adapter, NULL, &ctlq_msg,
- IDPF_VC_CREATE_VPORT,
- IDPF_VC_CREATE_VPORT_ERR);
- break;
case VIRTCHNL2_OP_ENABLE_VPORT:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_ENA_VPORT,
- IDPF_VC_ENA_VPORT_ERR);
- break;
case VIRTCHNL2_OP_DISABLE_VPORT:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_DIS_VPORT,
- IDPF_VC_DIS_VPORT_ERR);
- break;
case VIRTCHNL2_OP_DESTROY_VPORT:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_DESTROY_VPORT,
- IDPF_VC_DESTROY_VPORT_ERR);
+ err = idpf_vc_xn_forward_reply(adapter, &ctlq_msg);
break;
case VIRTCHNL2_OP_CONFIG_TX_QUEUES:
idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
@@ -1225,7 +1194,9 @@ static int idpf_send_ver_msg(struct idpf_adapter *adapter)
*/
static int idpf_send_get_caps_msg(struct idpf_adapter *adapter)
{
- struct virtchnl2_get_capabilities caps = { };
+ struct virtchnl2_get_capabilities caps = {};
+ struct idpf_vc_xn_params xn_params = {};
+ ssize_t reply_sz;
caps.csum_caps =
cpu_to_le32(VIRTCHNL2_CAP_TX_CSUM_L3_IPV4 |
@@ -1282,21 +1253,20 @@ static int idpf_send_get_caps_msg(struct idpf_adapter *adapter)
VIRTCHNL2_CAP_PROMISC |
VIRTCHNL2_CAP_LOOPBACK);
- return idpf_send_mb_msg(adapter, VIRTCHNL2_OP_GET_CAPS, sizeof(caps),
- (u8 *)&caps, 0);
-}
+ xn_params.vc_op = VIRTCHNL2_OP_GET_CAPS;
+ xn_params.send_buf.iov_base = ∩︀
+ xn_params.send_buf.iov_len = sizeof(caps);
+ xn_params.recv_buf.iov_base = &adapter->caps;
+ xn_params.recv_buf.iov_len = sizeof(adapter->caps);
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
-/**
- * idpf_recv_get_caps_msg - Receive virtchnl get capabilities message
- * @adapter: Driver specific private structure
- *
- * Receive virtchnl get capabilities message. Returns 0 on success, negative on
- * failure.
- */
-static int idpf_recv_get_caps_msg(struct idpf_adapter *adapter)
-{
- return idpf_recv_mb_msg(adapter, VIRTCHNL2_OP_GET_CAPS, &adapter->caps,
- sizeof(struct virtchnl2_get_capabilities));
+ reply_sz = idpf_vc_xn_exec(adapter, xn_params);
+ if (reply_sz < 0)
+ return reply_sz;
+ if (reply_sz < sizeof(adapter->caps))
+ return -EIO;
+
+ return 0;
}
/**
@@ -1623,8 +1593,10 @@ int idpf_send_create_vport_msg(struct idpf_adapter *adapter,
struct idpf_vport_max_q *max_q)
{
struct virtchnl2_create_vport *vport_msg;
+ struct idpf_vc_xn_params xn_params = {};
u16 idx = adapter->next_vport;
int err, buf_size;
+ ssize_t reply_sz;
buf_size = sizeof(struct virtchnl2_create_vport);
if (!adapter->vport_params_reqd[idx]) {
@@ -1655,35 +1627,38 @@ int idpf_send_create_vport_msg(struct idpf_adapter *adapter,
return err;
}
- mutex_lock(&adapter->vc_buf_lock);
-
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_CREATE_VPORT, buf_size,
- (u8 *)vport_msg, 0);
- if (err)
- goto rel_lock;
-
- err = idpf_wait_for_event(adapter, NULL, IDPF_VC_CREATE_VPORT,
- IDPF_VC_CREATE_VPORT_ERR);
- if (err) {
- dev_err(&adapter->pdev->dev, "Failed to receive create vport message");
-
- goto rel_lock;
- }
-
if (!adapter->vport_params_recvd[idx]) {
adapter->vport_params_recvd[idx] = kzalloc(IDPF_CTLQ_MAX_BUF_LEN,
GFP_KERNEL);
if (!adapter->vport_params_recvd[idx]) {
err = -ENOMEM;
- goto rel_lock;
+ goto free_vport_params;
}
}
- vport_msg = adapter->vport_params_recvd[idx];
- memcpy(vport_msg, adapter->vc_msg, IDPF_CTLQ_MAX_BUF_LEN);
+ xn_params.vc_op = VIRTCHNL2_OP_CREATE_VPORT;
+ xn_params.send_buf.iov_base = vport_msg;
+ xn_params.send_buf.iov_len = buf_size;
+ xn_params.recv_buf.iov_base = adapter->vport_params_recvd[idx];
+ xn_params.recv_buf.iov_len = IDPF_CTLQ_MAX_BUF_LEN;
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
+ reply_sz = idpf_vc_xn_exec(adapter, xn_params);
+ if (reply_sz < 0) {
+ err = reply_sz;
+ goto free_vport_params;
+ }
+ if (reply_sz < IDPF_CTLQ_MAX_BUF_LEN) {
+ err = -EIO;
+ goto free_vport_params;
+ }
-rel_lock:
- mutex_unlock(&adapter->vc_buf_lock);
+ return 0;
+
+free_vport_params:
+ kfree(adapter->vport_params_recvd[idx]);
+ adapter->vport_params_recvd[idx] = NULL;
+ kfree(adapter->vport_params_reqd[idx]);
+ adapter->vport_params_reqd[idx] = NULL;
return err;
}
@@ -1735,26 +1710,19 @@ int idpf_check_supported_desc_ids(struct idpf_vport *vport)
*/
int idpf_send_destroy_vport_msg(struct idpf_vport *vport)
{
- struct idpf_adapter *adapter = vport->adapter;
+ struct idpf_vc_xn_params xn_params = {};
struct virtchnl2_vport v_id;
- int err;
+ ssize_t reply_sz;
v_id.vport_id = cpu_to_le32(vport->vport_id);
- mutex_lock(&vport->vc_buf_lock);
-
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_DESTROY_VPORT,
- sizeof(v_id), (u8 *)&v_id, 0);
- if (err)
- goto rel_lock;
-
- err = idpf_min_wait_for_event(adapter, vport, IDPF_VC_DESTROY_VPORT,
- IDPF_VC_DESTROY_VPORT_ERR);
-
-rel_lock:
- mutex_unlock(&vport->vc_buf_lock);
+ xn_params.vc_op = VIRTCHNL2_OP_DESTROY_VPORT;
+ xn_params.send_buf.iov_base = &v_id;
+ xn_params.send_buf.iov_len = sizeof(v_id);
+ xn_params.timeout_ms = IDPF_VC_XN_MIN_TIMEOUT_MSEC;
+ reply_sz = idpf_vc_xn_exec(vport->adapter, xn_params);
- return err;
+ return reply_sz < 0 ? reply_sz : 0;
}
/**
@@ -1766,26 +1734,19 @@ int idpf_send_destroy_vport_msg(struct idpf_vport *vport)
*/
int idpf_send_enable_vport_msg(struct idpf_vport *vport)
{
- struct idpf_adapter *adapter = vport->adapter;
+ struct idpf_vc_xn_params xn_params = {};
struct virtchnl2_vport v_id;
- int err;
+ ssize_t reply_sz;
v_id.vport_id = cpu_to_le32(vport->vport_id);
- mutex_lock(&vport->vc_buf_lock);
-
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_ENABLE_VPORT,
- sizeof(v_id), (u8 *)&v_id, 0);
- if (err)
- goto rel_lock;
-
- err = idpf_wait_for_event(adapter, vport, IDPF_VC_ENA_VPORT,
- IDPF_VC_ENA_VPORT_ERR);
-
-rel_lock:
- mutex_unlock(&vport->vc_buf_lock);
+ xn_params.vc_op = VIRTCHNL2_OP_ENABLE_VPORT;
+ xn_params.send_buf.iov_base = &v_id;
+ xn_params.send_buf.iov_len = sizeof(v_id);
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
+ reply_sz = idpf_vc_xn_exec(vport->adapter, xn_params);
- return err;
+ return reply_sz < 0 ? reply_sz : 0;
}
/**
@@ -1797,26 +1758,19 @@ int idpf_send_enable_vport_msg(struct idpf_vport *vport)
*/
int idpf_send_disable_vport_msg(struct idpf_vport *vport)
{
- struct idpf_adapter *adapter = vport->adapter;
+ struct idpf_vc_xn_params xn_params = {};
struct virtchnl2_vport v_id;
- int err;
+ ssize_t reply_sz;
v_id.vport_id = cpu_to_le32(vport->vport_id);
- mutex_lock(&vport->vc_buf_lock);
-
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_DISABLE_VPORT,
- sizeof(v_id), (u8 *)&v_id, 0);
- if (err)
- goto rel_lock;
-
- err = idpf_min_wait_for_event(adapter, vport, IDPF_VC_DIS_VPORT,
- IDPF_VC_DIS_VPORT_ERR);
+ xn_params.vc_op = VIRTCHNL2_OP_DISABLE_VPORT;
+ xn_params.send_buf.iov_base = &v_id;
+ xn_params.send_buf.iov_len = sizeof(v_id);
+ xn_params.timeout_ms = IDPF_VC_XN_MIN_TIMEOUT_MSEC;
+ reply_sz = idpf_vc_xn_exec(vport->adapter, xn_params);
-rel_lock:
- mutex_unlock(&vport->vc_buf_lock);
-
- return err;
+ return reply_sz < 0 ? reply_sz : 0;
}
/**
@@ -3434,9 +3388,6 @@ int idpf_vc_core_init(struct idpf_adapter *adapter)
case 0:
/* success, move state machine forward */
adapter->state = __IDPF_GET_CAPS;
- err = idpf_send_get_caps_msg(adapter);
- if (err)
- goto init_failed;
fallthrough;
case -EAGAIN:
goto restart;
@@ -3447,13 +3398,15 @@ int idpf_vc_core_init(struct idpf_adapter *adapter)
goto init_failed;
}
case __IDPF_GET_CAPS:
- if (idpf_recv_get_caps_msg(adapter))
+ err = idpf_send_get_caps_msg(adapter);
+ if (err)
goto init_failed;
adapter->state = __IDPF_INIT_SW;
break;
default:
dev_err(&adapter->pdev->dev, "Device is in bad state: %d\n",
adapter->state);
+ err = -EINVAL;
goto init_failed;
}
break;
--
2.40.1
^ permalink raw reply related [flat|nested] 44+ messages in thread
* [PATCH v4 03/10 iwl-next] idpf: refactor queue related virtchnl messages
2024-02-06 3:37 ` [Intel-wired-lan] " Alan Brady
@ 2024-02-06 3:37 ` Alan Brady
-1 siblings, 0 replies; 44+ messages in thread
From: Alan Brady @ 2024-02-06 3:37 UTC (permalink / raw)
To: intel-wired-lan
Cc: netdev, willemdebruijn.kernel, przemyslaw.kitszel, igor.bagnucki,
aleksander.lobakin, Alan Brady
This reworks queue specific virtchnl messages to use the added
transaction API. It is fairly mechanical and generally makes the
functions using it more simple. Functions using transaction API no
longer need to take the vc_buf_lock since it's not using it anymore.
After filling out an idpf_vc_xn_params struct, idpf_vc_xn_exec takes
care of the send and recv handling.
This also converts those functions where appropriate to use
auto-variables instead of manually calling kfree. This greatly
simplifies the memory alloc paths and makes them less prone memory
leaks.
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Reviewed-by: Igor Bagnucki <igor.bagnucki@intel.com>
Signed-off-by: Alan Brady <alan.brady@intel.com>
---
drivers/net/ethernet/intel/idpf/idpf.h | 2 +-
.../net/ethernet/intel/idpf/idpf_virtchnl.c | 393 ++++++------------
2 files changed, 136 insertions(+), 259 deletions(-)
diff --git a/drivers/net/ethernet/intel/idpf/idpf.h b/drivers/net/ethernet/intel/idpf/idpf.h
index 92103c3cf5a0..393f6e46012f 100644
--- a/drivers/net/ethernet/intel/idpf/idpf.h
+++ b/drivers/net/ethernet/intel/idpf/idpf.h
@@ -648,7 +648,7 @@ struct idpf_vector_lifo {
struct idpf_vport_config {
struct idpf_vport_user_config_data user_config;
struct idpf_vport_max_q max_q;
- void *req_qs_chunks;
+ struct virtchnl2_add_queues *req_qs_chunks;
spinlock_t mac_filter_list_lock;
DECLARE_BITMAP(flags, IDPF_VPORT_CONFIG_FLAGS_NBITS);
};
diff --git a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
index 2f8d44f3ce69..193c284719ae 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
@@ -807,47 +807,15 @@ int idpf_recv_mb_msg(struct idpf_adapter *adapter, u32 op,
case VIRTCHNL2_OP_ENABLE_VPORT:
case VIRTCHNL2_OP_DISABLE_VPORT:
case VIRTCHNL2_OP_DESTROY_VPORT:
- err = idpf_vc_xn_forward_reply(adapter, &ctlq_msg);
- break;
case VIRTCHNL2_OP_CONFIG_TX_QUEUES:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_CONFIG_TXQ,
- IDPF_VC_CONFIG_TXQ_ERR);
- break;
case VIRTCHNL2_OP_CONFIG_RX_QUEUES:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_CONFIG_RXQ,
- IDPF_VC_CONFIG_RXQ_ERR);
- break;
case VIRTCHNL2_OP_ENABLE_QUEUES:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_ENA_QUEUES,
- IDPF_VC_ENA_QUEUES_ERR);
- break;
case VIRTCHNL2_OP_DISABLE_QUEUES:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_DIS_QUEUES,
- IDPF_VC_DIS_QUEUES_ERR);
- break;
case VIRTCHNL2_OP_ADD_QUEUES:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_ADD_QUEUES,
- IDPF_VC_ADD_QUEUES_ERR);
- break;
case VIRTCHNL2_OP_DEL_QUEUES:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_DEL_QUEUES,
- IDPF_VC_DEL_QUEUES_ERR);
- break;
case VIRTCHNL2_OP_MAP_QUEUE_VECTOR:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_MAP_IRQ,
- IDPF_VC_MAP_IRQ_ERR);
- break;
case VIRTCHNL2_OP_UNMAP_QUEUE_VECTOR:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_UNMAP_IRQ,
- IDPF_VC_UNMAP_IRQ_ERR);
+ err = idpf_vc_xn_forward_reply(adapter, &ctlq_msg);
break;
case VIRTCHNL2_OP_GET_STATS:
idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
@@ -1782,11 +1750,13 @@ int idpf_send_disable_vport_msg(struct idpf_vport *vport)
*/
static int idpf_send_config_tx_queues_msg(struct idpf_vport *vport)
{
- struct virtchnl2_config_tx_queues *ctq;
+ struct virtchnl2_config_tx_queues *ctq __free(kfree) = NULL;
+ struct virtchnl2_txq_info *qi __free(kfree) = NULL;
+ struct idpf_vc_xn_params xn_params = {};
u32 config_sz, chunk_sz, buf_sz;
int totqs, num_msgs, num_chunks;
- struct virtchnl2_txq_info *qi;
- int err = 0, i, k = 0;
+ ssize_t reply_sz;
+ int i, k = 0;
totqs = vport->num_txq + vport->num_complq;
qi = kcalloc(totqs, sizeof(struct virtchnl2_txq_info), GFP_KERNEL);
@@ -1847,10 +1817,8 @@ static int idpf_send_config_tx_queues_msg(struct idpf_vport *vport)
}
/* Make sure accounting agrees */
- if (k != totqs) {
- err = -EINVAL;
- goto error;
- }
+ if (k != totqs)
+ return -EINVAL;
/* Chunk up the queue contexts into multiple messages to avoid
* sending a control queue message buffer that is too large
@@ -1864,12 +1832,11 @@ static int idpf_send_config_tx_queues_msg(struct idpf_vport *vport)
buf_sz = struct_size(ctq, qinfo, num_chunks);
ctq = kzalloc(buf_sz, GFP_KERNEL);
- if (!ctq) {
- err = -ENOMEM;
- goto error;
- }
+ if (!ctq)
+ return -ENOMEM;
- mutex_lock(&vport->vc_buf_lock);
+ xn_params.vc_op = VIRTCHNL2_OP_CONFIG_TX_QUEUES;
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
for (i = 0, k = 0; i < num_msgs; i++) {
memset(ctq, 0, buf_sz);
@@ -1877,17 +1844,11 @@ static int idpf_send_config_tx_queues_msg(struct idpf_vport *vport)
ctq->num_qinfo = cpu_to_le16(num_chunks);
memcpy(ctq->qinfo, &qi[k], chunk_sz * num_chunks);
- err = idpf_send_mb_msg(vport->adapter,
- VIRTCHNL2_OP_CONFIG_TX_QUEUES,
- buf_sz, (u8 *)ctq, 0);
- if (err)
- goto mbx_error;
-
- err = idpf_wait_for_event(vport->adapter, vport,
- IDPF_VC_CONFIG_TXQ,
- IDPF_VC_CONFIG_TXQ_ERR);
- if (err)
- goto mbx_error;
+ xn_params.send_buf.iov_base = ctq;
+ xn_params.send_buf.iov_len = buf_sz;
+ reply_sz = idpf_vc_xn_exec(vport->adapter, xn_params);
+ if (reply_sz < 0)
+ return reply_sz;
k += num_chunks;
totqs -= num_chunks;
@@ -1896,13 +1857,7 @@ static int idpf_send_config_tx_queues_msg(struct idpf_vport *vport)
buf_sz = struct_size(ctq, qinfo, num_chunks);
}
-mbx_error:
- mutex_unlock(&vport->vc_buf_lock);
- kfree(ctq);
-error:
- kfree(qi);
-
- return err;
+ return 0;
}
/**
@@ -1914,11 +1869,13 @@ static int idpf_send_config_tx_queues_msg(struct idpf_vport *vport)
*/
static int idpf_send_config_rx_queues_msg(struct idpf_vport *vport)
{
- struct virtchnl2_config_rx_queues *crq;
+ struct virtchnl2_config_rx_queues *crq __free(kfree) = NULL;
+ struct virtchnl2_rxq_info *qi __free(kfree) = NULL;
+ struct idpf_vc_xn_params xn_params = {};
u32 config_sz, chunk_sz, buf_sz;
int totqs, num_msgs, num_chunks;
- struct virtchnl2_rxq_info *qi;
- int err = 0, i, k = 0;
+ ssize_t reply_sz;
+ int i, k = 0;
totqs = vport->num_rxq + vport->num_bufq;
qi = kcalloc(totqs, sizeof(struct virtchnl2_rxq_info), GFP_KERNEL);
@@ -1999,10 +1956,8 @@ static int idpf_send_config_rx_queues_msg(struct idpf_vport *vport)
}
/* Make sure accounting agrees */
- if (k != totqs) {
- err = -EINVAL;
- goto error;
- }
+ if (k != totqs)
+ return -EINVAL;
/* Chunk up the queue contexts into multiple messages to avoid
* sending a control queue message buffer that is too large
@@ -2016,12 +1971,11 @@ static int idpf_send_config_rx_queues_msg(struct idpf_vport *vport)
buf_sz = struct_size(crq, qinfo, num_chunks);
crq = kzalloc(buf_sz, GFP_KERNEL);
- if (!crq) {
- err = -ENOMEM;
- goto error;
- }
+ if (!crq)
+ return -ENOMEM;
- mutex_lock(&vport->vc_buf_lock);
+ xn_params.vc_op = VIRTCHNL2_OP_CONFIG_RX_QUEUES;
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
for (i = 0, k = 0; i < num_msgs; i++) {
memset(crq, 0, buf_sz);
@@ -2029,17 +1983,11 @@ static int idpf_send_config_rx_queues_msg(struct idpf_vport *vport)
crq->num_qinfo = cpu_to_le16(num_chunks);
memcpy(crq->qinfo, &qi[k], chunk_sz * num_chunks);
- err = idpf_send_mb_msg(vport->adapter,
- VIRTCHNL2_OP_CONFIG_RX_QUEUES,
- buf_sz, (u8 *)crq, 0);
- if (err)
- goto mbx_error;
-
- err = idpf_wait_for_event(vport->adapter, vport,
- IDPF_VC_CONFIG_RXQ,
- IDPF_VC_CONFIG_RXQ_ERR);
- if (err)
- goto mbx_error;
+ xn_params.send_buf.iov_base = crq;
+ xn_params.send_buf.iov_len = buf_sz;
+ reply_sz = idpf_vc_xn_exec(vport->adapter, xn_params);
+ if (reply_sz < 0)
+ return reply_sz;
k += num_chunks;
totqs -= num_chunks;
@@ -2048,42 +1996,28 @@ static int idpf_send_config_rx_queues_msg(struct idpf_vport *vport)
buf_sz = struct_size(crq, qinfo, num_chunks);
}
-mbx_error:
- mutex_unlock(&vport->vc_buf_lock);
- kfree(crq);
-error:
- kfree(qi);
-
- return err;
+ return 0;
}
/**
* idpf_send_ena_dis_queues_msg - Send virtchnl enable or disable
* queues message
* @vport: virtual port data structure
- * @vc_op: virtchnl op code to send
+ * @ena: if true enable, false disable
*
* Send enable or disable queues virtchnl message. Returns 0 on success,
* negative on failure.
*/
-static int idpf_send_ena_dis_queues_msg(struct idpf_vport *vport, u32 vc_op)
+static int idpf_send_ena_dis_queues_msg(struct idpf_vport *vport, bool ena)
{
+ struct virtchnl2_del_ena_dis_queues *eq __free(kfree) = NULL;
+ struct virtchnl2_queue_chunk *qc __free(kfree) = NULL;
u32 num_msgs, num_chunks, num_txq, num_rxq, num_q;
- struct idpf_adapter *adapter = vport->adapter;
- struct virtchnl2_del_ena_dis_queues *eq;
+ struct idpf_vc_xn_params xn_params = {};
struct virtchnl2_queue_chunks *qcs;
- struct virtchnl2_queue_chunk *qc;
u32 config_sz, chunk_sz, buf_sz;
- int i, j, k = 0, err = 0;
-
- /* validate virtchnl op */
- switch (vc_op) {
- case VIRTCHNL2_OP_ENABLE_QUEUES:
- case VIRTCHNL2_OP_DISABLE_QUEUES:
- break;
- default:
- return -EINVAL;
- }
+ ssize_t reply_sz;
+ int i, j, k = 0;
num_txq = vport->num_txq + vport->num_complq;
num_rxq = vport->num_rxq + vport->num_bufq;
@@ -2102,10 +2036,8 @@ static int idpf_send_ena_dis_queues_msg(struct idpf_vport *vport, u32 vc_op)
qc[k].num_queues = cpu_to_le32(IDPF_NUMQ_PER_CHUNK);
}
}
- if (vport->num_txq != k) {
- err = -EINVAL;
- goto error;
- }
+ if (vport->num_txq != k)
+ return -EINVAL;
if (!idpf_is_queue_model_split(vport->txq_model))
goto setup_rx;
@@ -2117,10 +2049,8 @@ static int idpf_send_ena_dis_queues_msg(struct idpf_vport *vport, u32 vc_op)
qc[k].start_queue_id = cpu_to_le32(tx_qgrp->complq->q_id);
qc[k].num_queues = cpu_to_le32(IDPF_NUMQ_PER_CHUNK);
}
- if (vport->num_complq != (k - vport->num_txq)) {
- err = -EINVAL;
- goto error;
- }
+ if (vport->num_complq != (k - vport->num_txq))
+ return -EINVAL;
setup_rx:
for (i = 0; i < vport->num_rxq_grp; i++) {
@@ -2146,10 +2076,8 @@ static int idpf_send_ena_dis_queues_msg(struct idpf_vport *vport, u32 vc_op)
qc[k].num_queues = cpu_to_le32(IDPF_NUMQ_PER_CHUNK);
}
}
- if (vport->num_rxq != k - (vport->num_txq + vport->num_complq)) {
- err = -EINVAL;
- goto error;
- }
+ if (vport->num_rxq != k - (vport->num_txq + vport->num_complq))
+ return -EINVAL;
if (!idpf_is_queue_model_split(vport->rxq_model))
goto send_msg;
@@ -2168,10 +2096,8 @@ static int idpf_send_ena_dis_queues_msg(struct idpf_vport *vport, u32 vc_op)
}
if (vport->num_bufq != k - (vport->num_txq +
vport->num_complq +
- vport->num_rxq)) {
- err = -EINVAL;
- goto error;
- }
+ vport->num_rxq))
+ return -EINVAL;
send_msg:
/* Chunk up the queue info into multiple messages */
@@ -2184,12 +2110,16 @@ static int idpf_send_ena_dis_queues_msg(struct idpf_vport *vport, u32 vc_op)
buf_sz = struct_size(eq, chunks.chunks, num_chunks);
eq = kzalloc(buf_sz, GFP_KERNEL);
- if (!eq) {
- err = -ENOMEM;
- goto error;
- }
+ if (!eq)
+ return -ENOMEM;
- mutex_lock(&vport->vc_buf_lock);
+ if (ena) {
+ xn_params.vc_op = VIRTCHNL2_OP_ENABLE_QUEUES;
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
+ } else {
+ xn_params.vc_op = VIRTCHNL2_OP_DISABLE_QUEUES;
+ xn_params.timeout_ms = IDPF_VC_XN_MIN_TIMEOUT_MSEC;
+ }
for (i = 0, k = 0; i < num_msgs; i++) {
memset(eq, 0, buf_sz);
@@ -2198,20 +2128,11 @@ static int idpf_send_ena_dis_queues_msg(struct idpf_vport *vport, u32 vc_op)
qcs = &eq->chunks;
memcpy(qcs->chunks, &qc[k], chunk_sz * num_chunks);
- err = idpf_send_mb_msg(adapter, vc_op, buf_sz, (u8 *)eq, 0);
- if (err)
- goto mbx_error;
-
- if (vc_op == VIRTCHNL2_OP_ENABLE_QUEUES)
- err = idpf_wait_for_event(adapter, vport,
- IDPF_VC_ENA_QUEUES,
- IDPF_VC_ENA_QUEUES_ERR);
- else
- err = idpf_min_wait_for_event(adapter, vport,
- IDPF_VC_DIS_QUEUES,
- IDPF_VC_DIS_QUEUES_ERR);
- if (err)
- goto mbx_error;
+ xn_params.send_buf.iov_base = eq;
+ xn_params.send_buf.iov_len = buf_sz;
+ reply_sz = idpf_vc_xn_exec(vport->adapter, xn_params);
+ if (reply_sz < 0)
+ return reply_sz;
k += num_chunks;
num_q -= num_chunks;
@@ -2220,13 +2141,7 @@ static int idpf_send_ena_dis_queues_msg(struct idpf_vport *vport, u32 vc_op)
buf_sz = struct_size(eq, chunks.chunks, num_chunks);
}
-mbx_error:
- mutex_unlock(&vport->vc_buf_lock);
- kfree(eq);
-error:
- kfree(qc);
-
- return err;
+ return 0;
}
/**
@@ -2240,12 +2155,13 @@ static int idpf_send_ena_dis_queues_msg(struct idpf_vport *vport, u32 vc_op)
*/
int idpf_send_map_unmap_queue_vector_msg(struct idpf_vport *vport, bool map)
{
- struct idpf_adapter *adapter = vport->adapter;
- struct virtchnl2_queue_vector_maps *vqvm;
- struct virtchnl2_queue_vector *vqv;
+ struct virtchnl2_queue_vector_maps *vqvm __free(kfree) = NULL;
+ struct virtchnl2_queue_vector *vqv __free(kfree) = NULL;
+ struct idpf_vc_xn_params xn_params = {};
u32 config_sz, chunk_sz, buf_sz;
u32 num_msgs, num_chunks, num_q;
- int i, j, k = 0, err = 0;
+ ssize_t reply_sz;
+ int i, j, k = 0;
num_q = vport->num_txq + vport->num_rxq;
@@ -2275,10 +2191,8 @@ int idpf_send_map_unmap_queue_vector_msg(struct idpf_vport *vport, bool map)
}
}
- if (vport->num_txq != k) {
- err = -EINVAL;
- goto error;
- }
+ if (vport->num_txq != k)
+ return -EINVAL;
for (i = 0; i < vport->num_rxq_grp; i++) {
struct idpf_rxq_group *rx_qgrp = &vport->rxq_grps[i];
@@ -2305,15 +2219,11 @@ int idpf_send_map_unmap_queue_vector_msg(struct idpf_vport *vport, bool map)
}
if (idpf_is_queue_model_split(vport->txq_model)) {
- if (vport->num_rxq != k - vport->num_complq) {
- err = -EINVAL;
- goto error;
- }
+ if (vport->num_rxq != k - vport->num_complq)
+ return -EINVAL;
} else {
- if (vport->num_rxq != k - vport->num_txq) {
- err = -EINVAL;
- goto error;
- }
+ if (vport->num_rxq != k - vport->num_txq)
+ return -EINVAL;
}
/* Chunk up the vector info into multiple messages */
@@ -2326,39 +2236,28 @@ int idpf_send_map_unmap_queue_vector_msg(struct idpf_vport *vport, bool map)
buf_sz = struct_size(vqvm, qv_maps, num_chunks);
vqvm = kzalloc(buf_sz, GFP_KERNEL);
- if (!vqvm) {
- err = -ENOMEM;
- goto error;
- }
+ if (!vqvm)
+ return -ENOMEM;
- mutex_lock(&vport->vc_buf_lock);
+ if (map) {
+ xn_params.vc_op = VIRTCHNL2_OP_MAP_QUEUE_VECTOR;
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
+ } else {
+ xn_params.vc_op = VIRTCHNL2_OP_UNMAP_QUEUE_VECTOR;
+ xn_params.timeout_ms = IDPF_VC_XN_MIN_TIMEOUT_MSEC;
+ }
for (i = 0, k = 0; i < num_msgs; i++) {
memset(vqvm, 0, buf_sz);
+ xn_params.send_buf.iov_base = vqvm;
+ xn_params.send_buf.iov_len = buf_sz;
vqvm->vport_id = cpu_to_le32(vport->vport_id);
vqvm->num_qv_maps = cpu_to_le16(num_chunks);
memcpy(vqvm->qv_maps, &vqv[k], chunk_sz * num_chunks);
- if (map) {
- err = idpf_send_mb_msg(adapter,
- VIRTCHNL2_OP_MAP_QUEUE_VECTOR,
- buf_sz, (u8 *)vqvm, 0);
- if (!err)
- err = idpf_wait_for_event(adapter, vport,
- IDPF_VC_MAP_IRQ,
- IDPF_VC_MAP_IRQ_ERR);
- } else {
- err = idpf_send_mb_msg(adapter,
- VIRTCHNL2_OP_UNMAP_QUEUE_VECTOR,
- buf_sz, (u8 *)vqvm, 0);
- if (!err)
- err =
- idpf_min_wait_for_event(adapter, vport,
- IDPF_VC_UNMAP_IRQ,
- IDPF_VC_UNMAP_IRQ_ERR);
- }
- if (err)
- goto mbx_error;
+ reply_sz = idpf_vc_xn_exec(vport->adapter, xn_params);
+ if (reply_sz < 0)
+ return reply_sz;
k += num_chunks;
num_q -= num_chunks;
@@ -2367,13 +2266,7 @@ int idpf_send_map_unmap_queue_vector_msg(struct idpf_vport *vport, bool map)
buf_sz = struct_size(vqvm, qv_maps, num_chunks);
}
-mbx_error:
- mutex_unlock(&vport->vc_buf_lock);
- kfree(vqvm);
-error:
- kfree(vqv);
-
- return err;
+ return 0;
}
/**
@@ -2385,7 +2278,7 @@ int idpf_send_map_unmap_queue_vector_msg(struct idpf_vport *vport, bool map)
*/
int idpf_send_enable_queues_msg(struct idpf_vport *vport)
{
- return idpf_send_ena_dis_queues_msg(vport, VIRTCHNL2_OP_ENABLE_QUEUES);
+ return idpf_send_ena_dis_queues_msg(vport, true);
}
/**
@@ -2399,7 +2292,7 @@ int idpf_send_disable_queues_msg(struct idpf_vport *vport)
{
int err, i;
- err = idpf_send_ena_dis_queues_msg(vport, VIRTCHNL2_OP_DISABLE_QUEUES);
+ err = idpf_send_ena_dis_queues_msg(vport, false);
if (err)
return err;
@@ -2445,22 +2338,21 @@ static void idpf_convert_reg_to_queue_chunks(struct virtchnl2_queue_chunk *dchun
*/
int idpf_send_delete_queues_msg(struct idpf_vport *vport)
{
- struct idpf_adapter *adapter = vport->adapter;
+ struct virtchnl2_del_ena_dis_queues *eq __free(kfree) = NULL;
struct virtchnl2_create_vport *vport_params;
struct virtchnl2_queue_reg_chunks *chunks;
- struct virtchnl2_del_ena_dis_queues *eq;
+ struct idpf_vc_xn_params xn_params = {};
struct idpf_vport_config *vport_config;
u16 vport_idx = vport->idx;
- int buf_size, err;
+ ssize_t reply_sz;
u16 num_chunks;
+ int buf_size;
- vport_config = adapter->vport_config[vport_idx];
+ vport_config = vport->adapter->vport_config[vport_idx];
if (vport_config->req_qs_chunks) {
- struct virtchnl2_add_queues *vc_aq =
- (struct virtchnl2_add_queues *)vport_config->req_qs_chunks;
- chunks = &vc_aq->chunks;
+ chunks = &vport_config->req_qs_chunks->chunks;
} else {
- vport_params = adapter->vport_params_recvd[vport_idx];
+ vport_params = vport->adapter->vport_params_recvd[vport_idx];
chunks = &vport_params->chunks;
}
@@ -2477,21 +2369,13 @@ int idpf_send_delete_queues_msg(struct idpf_vport *vport)
idpf_convert_reg_to_queue_chunks(eq->chunks.chunks, chunks->chunks,
num_chunks);
- mutex_lock(&vport->vc_buf_lock);
-
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_DEL_QUEUES,
- buf_size, (u8 *)eq, 0);
- if (err)
- goto rel_lock;
-
- err = idpf_min_wait_for_event(adapter, vport, IDPF_VC_DEL_QUEUES,
- IDPF_VC_DEL_QUEUES_ERR);
-
-rel_lock:
- mutex_unlock(&vport->vc_buf_lock);
- kfree(eq);
+ xn_params.vc_op = VIRTCHNL2_OP_DEL_QUEUES;
+ xn_params.timeout_ms = IDPF_VC_XN_MIN_TIMEOUT_MSEC;
+ xn_params.send_buf.iov_base = eq;
+ xn_params.send_buf.iov_len = buf_size;
+ reply_sz = idpf_vc_xn_exec(vport->adapter, xn_params);
- return err;
+ return reply_sz < 0 ? reply_sz : 0;
}
/**
@@ -2526,14 +2410,21 @@ int idpf_send_config_queues_msg(struct idpf_vport *vport)
int idpf_send_add_queues_msg(const struct idpf_vport *vport, u16 num_tx_q,
u16 num_complq, u16 num_rx_q, u16 num_rx_bufq)
{
- struct idpf_adapter *adapter = vport->adapter;
+ struct virtchnl2_add_queues *vc_msg __free(kfree) = NULL;
+ struct idpf_vc_xn_params xn_params = {};
struct idpf_vport_config *vport_config;
- struct virtchnl2_add_queues aq = { };
- struct virtchnl2_add_queues *vc_msg;
+ struct virtchnl2_add_queues aq = {};
u16 vport_idx = vport->idx;
- int size, err;
+ ssize_t reply_sz;
+ int size;
- vport_config = adapter->vport_config[vport_idx];
+ vc_msg = kzalloc(IDPF_CTLQ_MAX_BUF_LEN, GFP_KERNEL);
+ if (!vc_msg)
+ return -ENOMEM;
+
+ vport_config = vport->adapter->vport_config[vport_idx];
+ kfree(vport_config->req_qs_chunks);
+ vport_config->req_qs_chunks = NULL;
aq.vport_id = cpu_to_le32(vport->vport_id);
aq.num_tx_q = cpu_to_le16(num_tx_q);
@@ -2541,47 +2432,33 @@ int idpf_send_add_queues_msg(const struct idpf_vport *vport, u16 num_tx_q,
aq.num_rx_q = cpu_to_le16(num_rx_q);
aq.num_rx_bufq = cpu_to_le16(num_rx_bufq);
- mutex_lock(&((struct idpf_vport *)vport)->vc_buf_lock);
-
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_ADD_QUEUES,
- sizeof(struct virtchnl2_add_queues), (u8 *)&aq, 0);
- if (err)
- goto rel_lock;
-
- /* We want vport to be const to prevent incidental code changes making
- * changes to the vport config. We're making a special exception here
- * to discard const to use the virtchnl.
- */
- err = idpf_wait_for_event(adapter, (struct idpf_vport *)vport,
- IDPF_VC_ADD_QUEUES, IDPF_VC_ADD_QUEUES_ERR);
- if (err)
- goto rel_lock;
-
- kfree(vport_config->req_qs_chunks);
- vport_config->req_qs_chunks = NULL;
+ xn_params.vc_op = VIRTCHNL2_OP_ADD_QUEUES;
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
+ xn_params.send_buf.iov_base = &aq;
+ xn_params.send_buf.iov_len = sizeof(aq);
+ xn_params.recv_buf.iov_base = vc_msg;
+ xn_params.recv_buf.iov_len = IDPF_CTLQ_MAX_BUF_LEN;
+ reply_sz = idpf_vc_xn_exec(vport->adapter, xn_params);
+ if (reply_sz < 0)
+ return reply_sz;
- vc_msg = (struct virtchnl2_add_queues *)vport->vc_msg;
/* compare vc_msg num queues with vport num queues */
if (le16_to_cpu(vc_msg->num_tx_q) != num_tx_q ||
le16_to_cpu(vc_msg->num_rx_q) != num_rx_q ||
le16_to_cpu(vc_msg->num_tx_complq) != num_complq ||
- le16_to_cpu(vc_msg->num_rx_bufq) != num_rx_bufq) {
- err = -EINVAL;
- goto rel_lock;
- }
+ le16_to_cpu(vc_msg->num_rx_bufq) != num_rx_bufq)
+ return -EINVAL;
size = struct_size(vc_msg, chunks.chunks,
le16_to_cpu(vc_msg->chunks.num_chunks));
- vport_config->req_qs_chunks = kmemdup(vc_msg, size, GFP_KERNEL);
- if (!vport_config->req_qs_chunks) {
- err = -ENOMEM;
- goto rel_lock;
- }
+ if (reply_sz < size)
+ return -EIO;
-rel_lock:
- mutex_unlock(&((struct idpf_vport *)vport)->vc_buf_lock);
+ vport_config->req_qs_chunks = kmemdup(vc_msg, size, GFP_KERNEL);
+ if (!vport_config->req_qs_chunks)
+ return -ENOMEM;
- return err;
+ return 0;
}
/**
--
2.40.1
^ permalink raw reply related [flat|nested] 44+ messages in thread
* [Intel-wired-lan] [PATCH v4 03/10 iwl-next] idpf: refactor queue related virtchnl messages
@ 2024-02-06 3:37 ` Alan Brady
0 siblings, 0 replies; 44+ messages in thread
From: Alan Brady @ 2024-02-06 3:37 UTC (permalink / raw)
To: intel-wired-lan
Cc: willemdebruijn.kernel, netdev, aleksander.lobakin, Alan Brady,
przemyslaw.kitszel, igor.bagnucki
This reworks queue specific virtchnl messages to use the added
transaction API. It is fairly mechanical and generally makes the
functions using it more simple. Functions using transaction API no
longer need to take the vc_buf_lock since it's not using it anymore.
After filling out an idpf_vc_xn_params struct, idpf_vc_xn_exec takes
care of the send and recv handling.
This also converts those functions where appropriate to use
auto-variables instead of manually calling kfree. This greatly
simplifies the memory alloc paths and makes them less prone memory
leaks.
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Reviewed-by: Igor Bagnucki <igor.bagnucki@intel.com>
Signed-off-by: Alan Brady <alan.brady@intel.com>
---
drivers/net/ethernet/intel/idpf/idpf.h | 2 +-
.../net/ethernet/intel/idpf/idpf_virtchnl.c | 393 ++++++------------
2 files changed, 136 insertions(+), 259 deletions(-)
diff --git a/drivers/net/ethernet/intel/idpf/idpf.h b/drivers/net/ethernet/intel/idpf/idpf.h
index 92103c3cf5a0..393f6e46012f 100644
--- a/drivers/net/ethernet/intel/idpf/idpf.h
+++ b/drivers/net/ethernet/intel/idpf/idpf.h
@@ -648,7 +648,7 @@ struct idpf_vector_lifo {
struct idpf_vport_config {
struct idpf_vport_user_config_data user_config;
struct idpf_vport_max_q max_q;
- void *req_qs_chunks;
+ struct virtchnl2_add_queues *req_qs_chunks;
spinlock_t mac_filter_list_lock;
DECLARE_BITMAP(flags, IDPF_VPORT_CONFIG_FLAGS_NBITS);
};
diff --git a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
index 2f8d44f3ce69..193c284719ae 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
@@ -807,47 +807,15 @@ int idpf_recv_mb_msg(struct idpf_adapter *adapter, u32 op,
case VIRTCHNL2_OP_ENABLE_VPORT:
case VIRTCHNL2_OP_DISABLE_VPORT:
case VIRTCHNL2_OP_DESTROY_VPORT:
- err = idpf_vc_xn_forward_reply(adapter, &ctlq_msg);
- break;
case VIRTCHNL2_OP_CONFIG_TX_QUEUES:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_CONFIG_TXQ,
- IDPF_VC_CONFIG_TXQ_ERR);
- break;
case VIRTCHNL2_OP_CONFIG_RX_QUEUES:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_CONFIG_RXQ,
- IDPF_VC_CONFIG_RXQ_ERR);
- break;
case VIRTCHNL2_OP_ENABLE_QUEUES:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_ENA_QUEUES,
- IDPF_VC_ENA_QUEUES_ERR);
- break;
case VIRTCHNL2_OP_DISABLE_QUEUES:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_DIS_QUEUES,
- IDPF_VC_DIS_QUEUES_ERR);
- break;
case VIRTCHNL2_OP_ADD_QUEUES:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_ADD_QUEUES,
- IDPF_VC_ADD_QUEUES_ERR);
- break;
case VIRTCHNL2_OP_DEL_QUEUES:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_DEL_QUEUES,
- IDPF_VC_DEL_QUEUES_ERR);
- break;
case VIRTCHNL2_OP_MAP_QUEUE_VECTOR:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_MAP_IRQ,
- IDPF_VC_MAP_IRQ_ERR);
- break;
case VIRTCHNL2_OP_UNMAP_QUEUE_VECTOR:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_UNMAP_IRQ,
- IDPF_VC_UNMAP_IRQ_ERR);
+ err = idpf_vc_xn_forward_reply(adapter, &ctlq_msg);
break;
case VIRTCHNL2_OP_GET_STATS:
idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
@@ -1782,11 +1750,13 @@ int idpf_send_disable_vport_msg(struct idpf_vport *vport)
*/
static int idpf_send_config_tx_queues_msg(struct idpf_vport *vport)
{
- struct virtchnl2_config_tx_queues *ctq;
+ struct virtchnl2_config_tx_queues *ctq __free(kfree) = NULL;
+ struct virtchnl2_txq_info *qi __free(kfree) = NULL;
+ struct idpf_vc_xn_params xn_params = {};
u32 config_sz, chunk_sz, buf_sz;
int totqs, num_msgs, num_chunks;
- struct virtchnl2_txq_info *qi;
- int err = 0, i, k = 0;
+ ssize_t reply_sz;
+ int i, k = 0;
totqs = vport->num_txq + vport->num_complq;
qi = kcalloc(totqs, sizeof(struct virtchnl2_txq_info), GFP_KERNEL);
@@ -1847,10 +1817,8 @@ static int idpf_send_config_tx_queues_msg(struct idpf_vport *vport)
}
/* Make sure accounting agrees */
- if (k != totqs) {
- err = -EINVAL;
- goto error;
- }
+ if (k != totqs)
+ return -EINVAL;
/* Chunk up the queue contexts into multiple messages to avoid
* sending a control queue message buffer that is too large
@@ -1864,12 +1832,11 @@ static int idpf_send_config_tx_queues_msg(struct idpf_vport *vport)
buf_sz = struct_size(ctq, qinfo, num_chunks);
ctq = kzalloc(buf_sz, GFP_KERNEL);
- if (!ctq) {
- err = -ENOMEM;
- goto error;
- }
+ if (!ctq)
+ return -ENOMEM;
- mutex_lock(&vport->vc_buf_lock);
+ xn_params.vc_op = VIRTCHNL2_OP_CONFIG_TX_QUEUES;
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
for (i = 0, k = 0; i < num_msgs; i++) {
memset(ctq, 0, buf_sz);
@@ -1877,17 +1844,11 @@ static int idpf_send_config_tx_queues_msg(struct idpf_vport *vport)
ctq->num_qinfo = cpu_to_le16(num_chunks);
memcpy(ctq->qinfo, &qi[k], chunk_sz * num_chunks);
- err = idpf_send_mb_msg(vport->adapter,
- VIRTCHNL2_OP_CONFIG_TX_QUEUES,
- buf_sz, (u8 *)ctq, 0);
- if (err)
- goto mbx_error;
-
- err = idpf_wait_for_event(vport->adapter, vport,
- IDPF_VC_CONFIG_TXQ,
- IDPF_VC_CONFIG_TXQ_ERR);
- if (err)
- goto mbx_error;
+ xn_params.send_buf.iov_base = ctq;
+ xn_params.send_buf.iov_len = buf_sz;
+ reply_sz = idpf_vc_xn_exec(vport->adapter, xn_params);
+ if (reply_sz < 0)
+ return reply_sz;
k += num_chunks;
totqs -= num_chunks;
@@ -1896,13 +1857,7 @@ static int idpf_send_config_tx_queues_msg(struct idpf_vport *vport)
buf_sz = struct_size(ctq, qinfo, num_chunks);
}
-mbx_error:
- mutex_unlock(&vport->vc_buf_lock);
- kfree(ctq);
-error:
- kfree(qi);
-
- return err;
+ return 0;
}
/**
@@ -1914,11 +1869,13 @@ static int idpf_send_config_tx_queues_msg(struct idpf_vport *vport)
*/
static int idpf_send_config_rx_queues_msg(struct idpf_vport *vport)
{
- struct virtchnl2_config_rx_queues *crq;
+ struct virtchnl2_config_rx_queues *crq __free(kfree) = NULL;
+ struct virtchnl2_rxq_info *qi __free(kfree) = NULL;
+ struct idpf_vc_xn_params xn_params = {};
u32 config_sz, chunk_sz, buf_sz;
int totqs, num_msgs, num_chunks;
- struct virtchnl2_rxq_info *qi;
- int err = 0, i, k = 0;
+ ssize_t reply_sz;
+ int i, k = 0;
totqs = vport->num_rxq + vport->num_bufq;
qi = kcalloc(totqs, sizeof(struct virtchnl2_rxq_info), GFP_KERNEL);
@@ -1999,10 +1956,8 @@ static int idpf_send_config_rx_queues_msg(struct idpf_vport *vport)
}
/* Make sure accounting agrees */
- if (k != totqs) {
- err = -EINVAL;
- goto error;
- }
+ if (k != totqs)
+ return -EINVAL;
/* Chunk up the queue contexts into multiple messages to avoid
* sending a control queue message buffer that is too large
@@ -2016,12 +1971,11 @@ static int idpf_send_config_rx_queues_msg(struct idpf_vport *vport)
buf_sz = struct_size(crq, qinfo, num_chunks);
crq = kzalloc(buf_sz, GFP_KERNEL);
- if (!crq) {
- err = -ENOMEM;
- goto error;
- }
+ if (!crq)
+ return -ENOMEM;
- mutex_lock(&vport->vc_buf_lock);
+ xn_params.vc_op = VIRTCHNL2_OP_CONFIG_RX_QUEUES;
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
for (i = 0, k = 0; i < num_msgs; i++) {
memset(crq, 0, buf_sz);
@@ -2029,17 +1983,11 @@ static int idpf_send_config_rx_queues_msg(struct idpf_vport *vport)
crq->num_qinfo = cpu_to_le16(num_chunks);
memcpy(crq->qinfo, &qi[k], chunk_sz * num_chunks);
- err = idpf_send_mb_msg(vport->adapter,
- VIRTCHNL2_OP_CONFIG_RX_QUEUES,
- buf_sz, (u8 *)crq, 0);
- if (err)
- goto mbx_error;
-
- err = idpf_wait_for_event(vport->adapter, vport,
- IDPF_VC_CONFIG_RXQ,
- IDPF_VC_CONFIG_RXQ_ERR);
- if (err)
- goto mbx_error;
+ xn_params.send_buf.iov_base = crq;
+ xn_params.send_buf.iov_len = buf_sz;
+ reply_sz = idpf_vc_xn_exec(vport->adapter, xn_params);
+ if (reply_sz < 0)
+ return reply_sz;
k += num_chunks;
totqs -= num_chunks;
@@ -2048,42 +1996,28 @@ static int idpf_send_config_rx_queues_msg(struct idpf_vport *vport)
buf_sz = struct_size(crq, qinfo, num_chunks);
}
-mbx_error:
- mutex_unlock(&vport->vc_buf_lock);
- kfree(crq);
-error:
- kfree(qi);
-
- return err;
+ return 0;
}
/**
* idpf_send_ena_dis_queues_msg - Send virtchnl enable or disable
* queues message
* @vport: virtual port data structure
- * @vc_op: virtchnl op code to send
+ * @ena: if true enable, false disable
*
* Send enable or disable queues virtchnl message. Returns 0 on success,
* negative on failure.
*/
-static int idpf_send_ena_dis_queues_msg(struct idpf_vport *vport, u32 vc_op)
+static int idpf_send_ena_dis_queues_msg(struct idpf_vport *vport, bool ena)
{
+ struct virtchnl2_del_ena_dis_queues *eq __free(kfree) = NULL;
+ struct virtchnl2_queue_chunk *qc __free(kfree) = NULL;
u32 num_msgs, num_chunks, num_txq, num_rxq, num_q;
- struct idpf_adapter *adapter = vport->adapter;
- struct virtchnl2_del_ena_dis_queues *eq;
+ struct idpf_vc_xn_params xn_params = {};
struct virtchnl2_queue_chunks *qcs;
- struct virtchnl2_queue_chunk *qc;
u32 config_sz, chunk_sz, buf_sz;
- int i, j, k = 0, err = 0;
-
- /* validate virtchnl op */
- switch (vc_op) {
- case VIRTCHNL2_OP_ENABLE_QUEUES:
- case VIRTCHNL2_OP_DISABLE_QUEUES:
- break;
- default:
- return -EINVAL;
- }
+ ssize_t reply_sz;
+ int i, j, k = 0;
num_txq = vport->num_txq + vport->num_complq;
num_rxq = vport->num_rxq + vport->num_bufq;
@@ -2102,10 +2036,8 @@ static int idpf_send_ena_dis_queues_msg(struct idpf_vport *vport, u32 vc_op)
qc[k].num_queues = cpu_to_le32(IDPF_NUMQ_PER_CHUNK);
}
}
- if (vport->num_txq != k) {
- err = -EINVAL;
- goto error;
- }
+ if (vport->num_txq != k)
+ return -EINVAL;
if (!idpf_is_queue_model_split(vport->txq_model))
goto setup_rx;
@@ -2117,10 +2049,8 @@ static int idpf_send_ena_dis_queues_msg(struct idpf_vport *vport, u32 vc_op)
qc[k].start_queue_id = cpu_to_le32(tx_qgrp->complq->q_id);
qc[k].num_queues = cpu_to_le32(IDPF_NUMQ_PER_CHUNK);
}
- if (vport->num_complq != (k - vport->num_txq)) {
- err = -EINVAL;
- goto error;
- }
+ if (vport->num_complq != (k - vport->num_txq))
+ return -EINVAL;
setup_rx:
for (i = 0; i < vport->num_rxq_grp; i++) {
@@ -2146,10 +2076,8 @@ static int idpf_send_ena_dis_queues_msg(struct idpf_vport *vport, u32 vc_op)
qc[k].num_queues = cpu_to_le32(IDPF_NUMQ_PER_CHUNK);
}
}
- if (vport->num_rxq != k - (vport->num_txq + vport->num_complq)) {
- err = -EINVAL;
- goto error;
- }
+ if (vport->num_rxq != k - (vport->num_txq + vport->num_complq))
+ return -EINVAL;
if (!idpf_is_queue_model_split(vport->rxq_model))
goto send_msg;
@@ -2168,10 +2096,8 @@ static int idpf_send_ena_dis_queues_msg(struct idpf_vport *vport, u32 vc_op)
}
if (vport->num_bufq != k - (vport->num_txq +
vport->num_complq +
- vport->num_rxq)) {
- err = -EINVAL;
- goto error;
- }
+ vport->num_rxq))
+ return -EINVAL;
send_msg:
/* Chunk up the queue info into multiple messages */
@@ -2184,12 +2110,16 @@ static int idpf_send_ena_dis_queues_msg(struct idpf_vport *vport, u32 vc_op)
buf_sz = struct_size(eq, chunks.chunks, num_chunks);
eq = kzalloc(buf_sz, GFP_KERNEL);
- if (!eq) {
- err = -ENOMEM;
- goto error;
- }
+ if (!eq)
+ return -ENOMEM;
- mutex_lock(&vport->vc_buf_lock);
+ if (ena) {
+ xn_params.vc_op = VIRTCHNL2_OP_ENABLE_QUEUES;
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
+ } else {
+ xn_params.vc_op = VIRTCHNL2_OP_DISABLE_QUEUES;
+ xn_params.timeout_ms = IDPF_VC_XN_MIN_TIMEOUT_MSEC;
+ }
for (i = 0, k = 0; i < num_msgs; i++) {
memset(eq, 0, buf_sz);
@@ -2198,20 +2128,11 @@ static int idpf_send_ena_dis_queues_msg(struct idpf_vport *vport, u32 vc_op)
qcs = &eq->chunks;
memcpy(qcs->chunks, &qc[k], chunk_sz * num_chunks);
- err = idpf_send_mb_msg(adapter, vc_op, buf_sz, (u8 *)eq, 0);
- if (err)
- goto mbx_error;
-
- if (vc_op == VIRTCHNL2_OP_ENABLE_QUEUES)
- err = idpf_wait_for_event(adapter, vport,
- IDPF_VC_ENA_QUEUES,
- IDPF_VC_ENA_QUEUES_ERR);
- else
- err = idpf_min_wait_for_event(adapter, vport,
- IDPF_VC_DIS_QUEUES,
- IDPF_VC_DIS_QUEUES_ERR);
- if (err)
- goto mbx_error;
+ xn_params.send_buf.iov_base = eq;
+ xn_params.send_buf.iov_len = buf_sz;
+ reply_sz = idpf_vc_xn_exec(vport->adapter, xn_params);
+ if (reply_sz < 0)
+ return reply_sz;
k += num_chunks;
num_q -= num_chunks;
@@ -2220,13 +2141,7 @@ static int idpf_send_ena_dis_queues_msg(struct idpf_vport *vport, u32 vc_op)
buf_sz = struct_size(eq, chunks.chunks, num_chunks);
}
-mbx_error:
- mutex_unlock(&vport->vc_buf_lock);
- kfree(eq);
-error:
- kfree(qc);
-
- return err;
+ return 0;
}
/**
@@ -2240,12 +2155,13 @@ static int idpf_send_ena_dis_queues_msg(struct idpf_vport *vport, u32 vc_op)
*/
int idpf_send_map_unmap_queue_vector_msg(struct idpf_vport *vport, bool map)
{
- struct idpf_adapter *adapter = vport->adapter;
- struct virtchnl2_queue_vector_maps *vqvm;
- struct virtchnl2_queue_vector *vqv;
+ struct virtchnl2_queue_vector_maps *vqvm __free(kfree) = NULL;
+ struct virtchnl2_queue_vector *vqv __free(kfree) = NULL;
+ struct idpf_vc_xn_params xn_params = {};
u32 config_sz, chunk_sz, buf_sz;
u32 num_msgs, num_chunks, num_q;
- int i, j, k = 0, err = 0;
+ ssize_t reply_sz;
+ int i, j, k = 0;
num_q = vport->num_txq + vport->num_rxq;
@@ -2275,10 +2191,8 @@ int idpf_send_map_unmap_queue_vector_msg(struct idpf_vport *vport, bool map)
}
}
- if (vport->num_txq != k) {
- err = -EINVAL;
- goto error;
- }
+ if (vport->num_txq != k)
+ return -EINVAL;
for (i = 0; i < vport->num_rxq_grp; i++) {
struct idpf_rxq_group *rx_qgrp = &vport->rxq_grps[i];
@@ -2305,15 +2219,11 @@ int idpf_send_map_unmap_queue_vector_msg(struct idpf_vport *vport, bool map)
}
if (idpf_is_queue_model_split(vport->txq_model)) {
- if (vport->num_rxq != k - vport->num_complq) {
- err = -EINVAL;
- goto error;
- }
+ if (vport->num_rxq != k - vport->num_complq)
+ return -EINVAL;
} else {
- if (vport->num_rxq != k - vport->num_txq) {
- err = -EINVAL;
- goto error;
- }
+ if (vport->num_rxq != k - vport->num_txq)
+ return -EINVAL;
}
/* Chunk up the vector info into multiple messages */
@@ -2326,39 +2236,28 @@ int idpf_send_map_unmap_queue_vector_msg(struct idpf_vport *vport, bool map)
buf_sz = struct_size(vqvm, qv_maps, num_chunks);
vqvm = kzalloc(buf_sz, GFP_KERNEL);
- if (!vqvm) {
- err = -ENOMEM;
- goto error;
- }
+ if (!vqvm)
+ return -ENOMEM;
- mutex_lock(&vport->vc_buf_lock);
+ if (map) {
+ xn_params.vc_op = VIRTCHNL2_OP_MAP_QUEUE_VECTOR;
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
+ } else {
+ xn_params.vc_op = VIRTCHNL2_OP_UNMAP_QUEUE_VECTOR;
+ xn_params.timeout_ms = IDPF_VC_XN_MIN_TIMEOUT_MSEC;
+ }
for (i = 0, k = 0; i < num_msgs; i++) {
memset(vqvm, 0, buf_sz);
+ xn_params.send_buf.iov_base = vqvm;
+ xn_params.send_buf.iov_len = buf_sz;
vqvm->vport_id = cpu_to_le32(vport->vport_id);
vqvm->num_qv_maps = cpu_to_le16(num_chunks);
memcpy(vqvm->qv_maps, &vqv[k], chunk_sz * num_chunks);
- if (map) {
- err = idpf_send_mb_msg(adapter,
- VIRTCHNL2_OP_MAP_QUEUE_VECTOR,
- buf_sz, (u8 *)vqvm, 0);
- if (!err)
- err = idpf_wait_for_event(adapter, vport,
- IDPF_VC_MAP_IRQ,
- IDPF_VC_MAP_IRQ_ERR);
- } else {
- err = idpf_send_mb_msg(adapter,
- VIRTCHNL2_OP_UNMAP_QUEUE_VECTOR,
- buf_sz, (u8 *)vqvm, 0);
- if (!err)
- err =
- idpf_min_wait_for_event(adapter, vport,
- IDPF_VC_UNMAP_IRQ,
- IDPF_VC_UNMAP_IRQ_ERR);
- }
- if (err)
- goto mbx_error;
+ reply_sz = idpf_vc_xn_exec(vport->adapter, xn_params);
+ if (reply_sz < 0)
+ return reply_sz;
k += num_chunks;
num_q -= num_chunks;
@@ -2367,13 +2266,7 @@ int idpf_send_map_unmap_queue_vector_msg(struct idpf_vport *vport, bool map)
buf_sz = struct_size(vqvm, qv_maps, num_chunks);
}
-mbx_error:
- mutex_unlock(&vport->vc_buf_lock);
- kfree(vqvm);
-error:
- kfree(vqv);
-
- return err;
+ return 0;
}
/**
@@ -2385,7 +2278,7 @@ int idpf_send_map_unmap_queue_vector_msg(struct idpf_vport *vport, bool map)
*/
int idpf_send_enable_queues_msg(struct idpf_vport *vport)
{
- return idpf_send_ena_dis_queues_msg(vport, VIRTCHNL2_OP_ENABLE_QUEUES);
+ return idpf_send_ena_dis_queues_msg(vport, true);
}
/**
@@ -2399,7 +2292,7 @@ int idpf_send_disable_queues_msg(struct idpf_vport *vport)
{
int err, i;
- err = idpf_send_ena_dis_queues_msg(vport, VIRTCHNL2_OP_DISABLE_QUEUES);
+ err = idpf_send_ena_dis_queues_msg(vport, false);
if (err)
return err;
@@ -2445,22 +2338,21 @@ static void idpf_convert_reg_to_queue_chunks(struct virtchnl2_queue_chunk *dchun
*/
int idpf_send_delete_queues_msg(struct idpf_vport *vport)
{
- struct idpf_adapter *adapter = vport->adapter;
+ struct virtchnl2_del_ena_dis_queues *eq __free(kfree) = NULL;
struct virtchnl2_create_vport *vport_params;
struct virtchnl2_queue_reg_chunks *chunks;
- struct virtchnl2_del_ena_dis_queues *eq;
+ struct idpf_vc_xn_params xn_params = {};
struct idpf_vport_config *vport_config;
u16 vport_idx = vport->idx;
- int buf_size, err;
+ ssize_t reply_sz;
u16 num_chunks;
+ int buf_size;
- vport_config = adapter->vport_config[vport_idx];
+ vport_config = vport->adapter->vport_config[vport_idx];
if (vport_config->req_qs_chunks) {
- struct virtchnl2_add_queues *vc_aq =
- (struct virtchnl2_add_queues *)vport_config->req_qs_chunks;
- chunks = &vc_aq->chunks;
+ chunks = &vport_config->req_qs_chunks->chunks;
} else {
- vport_params = adapter->vport_params_recvd[vport_idx];
+ vport_params = vport->adapter->vport_params_recvd[vport_idx];
chunks = &vport_params->chunks;
}
@@ -2477,21 +2369,13 @@ int idpf_send_delete_queues_msg(struct idpf_vport *vport)
idpf_convert_reg_to_queue_chunks(eq->chunks.chunks, chunks->chunks,
num_chunks);
- mutex_lock(&vport->vc_buf_lock);
-
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_DEL_QUEUES,
- buf_size, (u8 *)eq, 0);
- if (err)
- goto rel_lock;
-
- err = idpf_min_wait_for_event(adapter, vport, IDPF_VC_DEL_QUEUES,
- IDPF_VC_DEL_QUEUES_ERR);
-
-rel_lock:
- mutex_unlock(&vport->vc_buf_lock);
- kfree(eq);
+ xn_params.vc_op = VIRTCHNL2_OP_DEL_QUEUES;
+ xn_params.timeout_ms = IDPF_VC_XN_MIN_TIMEOUT_MSEC;
+ xn_params.send_buf.iov_base = eq;
+ xn_params.send_buf.iov_len = buf_size;
+ reply_sz = idpf_vc_xn_exec(vport->adapter, xn_params);
- return err;
+ return reply_sz < 0 ? reply_sz : 0;
}
/**
@@ -2526,14 +2410,21 @@ int idpf_send_config_queues_msg(struct idpf_vport *vport)
int idpf_send_add_queues_msg(const struct idpf_vport *vport, u16 num_tx_q,
u16 num_complq, u16 num_rx_q, u16 num_rx_bufq)
{
- struct idpf_adapter *adapter = vport->adapter;
+ struct virtchnl2_add_queues *vc_msg __free(kfree) = NULL;
+ struct idpf_vc_xn_params xn_params = {};
struct idpf_vport_config *vport_config;
- struct virtchnl2_add_queues aq = { };
- struct virtchnl2_add_queues *vc_msg;
+ struct virtchnl2_add_queues aq = {};
u16 vport_idx = vport->idx;
- int size, err;
+ ssize_t reply_sz;
+ int size;
- vport_config = adapter->vport_config[vport_idx];
+ vc_msg = kzalloc(IDPF_CTLQ_MAX_BUF_LEN, GFP_KERNEL);
+ if (!vc_msg)
+ return -ENOMEM;
+
+ vport_config = vport->adapter->vport_config[vport_idx];
+ kfree(vport_config->req_qs_chunks);
+ vport_config->req_qs_chunks = NULL;
aq.vport_id = cpu_to_le32(vport->vport_id);
aq.num_tx_q = cpu_to_le16(num_tx_q);
@@ -2541,47 +2432,33 @@ int idpf_send_add_queues_msg(const struct idpf_vport *vport, u16 num_tx_q,
aq.num_rx_q = cpu_to_le16(num_rx_q);
aq.num_rx_bufq = cpu_to_le16(num_rx_bufq);
- mutex_lock(&((struct idpf_vport *)vport)->vc_buf_lock);
-
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_ADD_QUEUES,
- sizeof(struct virtchnl2_add_queues), (u8 *)&aq, 0);
- if (err)
- goto rel_lock;
-
- /* We want vport to be const to prevent incidental code changes making
- * changes to the vport config. We're making a special exception here
- * to discard const to use the virtchnl.
- */
- err = idpf_wait_for_event(adapter, (struct idpf_vport *)vport,
- IDPF_VC_ADD_QUEUES, IDPF_VC_ADD_QUEUES_ERR);
- if (err)
- goto rel_lock;
-
- kfree(vport_config->req_qs_chunks);
- vport_config->req_qs_chunks = NULL;
+ xn_params.vc_op = VIRTCHNL2_OP_ADD_QUEUES;
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
+ xn_params.send_buf.iov_base = &aq;
+ xn_params.send_buf.iov_len = sizeof(aq);
+ xn_params.recv_buf.iov_base = vc_msg;
+ xn_params.recv_buf.iov_len = IDPF_CTLQ_MAX_BUF_LEN;
+ reply_sz = idpf_vc_xn_exec(vport->adapter, xn_params);
+ if (reply_sz < 0)
+ return reply_sz;
- vc_msg = (struct virtchnl2_add_queues *)vport->vc_msg;
/* compare vc_msg num queues with vport num queues */
if (le16_to_cpu(vc_msg->num_tx_q) != num_tx_q ||
le16_to_cpu(vc_msg->num_rx_q) != num_rx_q ||
le16_to_cpu(vc_msg->num_tx_complq) != num_complq ||
- le16_to_cpu(vc_msg->num_rx_bufq) != num_rx_bufq) {
- err = -EINVAL;
- goto rel_lock;
- }
+ le16_to_cpu(vc_msg->num_rx_bufq) != num_rx_bufq)
+ return -EINVAL;
size = struct_size(vc_msg, chunks.chunks,
le16_to_cpu(vc_msg->chunks.num_chunks));
- vport_config->req_qs_chunks = kmemdup(vc_msg, size, GFP_KERNEL);
- if (!vport_config->req_qs_chunks) {
- err = -ENOMEM;
- goto rel_lock;
- }
+ if (reply_sz < size)
+ return -EIO;
-rel_lock:
- mutex_unlock(&((struct idpf_vport *)vport)->vc_buf_lock);
+ vport_config->req_qs_chunks = kmemdup(vc_msg, size, GFP_KERNEL);
+ if (!vport_config->req_qs_chunks)
+ return -ENOMEM;
- return err;
+ return 0;
}
/**
--
2.40.1
^ permalink raw reply related [flat|nested] 44+ messages in thread
* [PATCH v4 04/10 iwl-next] idpf: refactor remaining virtchnl messages
2024-02-06 3:37 ` [Intel-wired-lan] " Alan Brady
@ 2024-02-06 3:37 ` Alan Brady
-1 siblings, 0 replies; 44+ messages in thread
From: Alan Brady @ 2024-02-06 3:37 UTC (permalink / raw)
To: intel-wired-lan
Cc: netdev, willemdebruijn.kernel, przemyslaw.kitszel, igor.bagnucki,
aleksander.lobakin, Alan Brady
This takes care of RSS/SRIOV/MAC and other misc virtchnl messages. This
again is mostly mechanical.
In absence of an async_handler for MAC filters, this will simply
generically report any errors from idpf_vc_xn_forward_async. This
maintains the existing behavior. Follow up patch will add an async
handler for MAC filters to remove bad filters from our list.
While we're here we can also make the code much nicer by converting some
variables to auto-variables where appropriate. This makes it cleaner and
less prone to memory leaking.
There's still a bit more cleanup we can do here to remove stuff that's
not being used anymore now; follow-up patches will take care of loose
ends.
Signed-off-by: Alan Brady <alan.brady@intel.com>
---
.../net/ethernet/intel/idpf/idpf_virtchnl.c | 894 ++++++------------
1 file changed, 298 insertions(+), 596 deletions(-)
diff --git a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
index 193c284719ae..6d95daf0637b 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
@@ -3,6 +3,64 @@
#include "idpf.h"
+/**
+ * idpf_vid_to_vport - Translate vport id to vport pointer
+ * @adapter: private data struct
+ * @v_id: vport id to translate
+ *
+ * Returns vport matching v_id, NULL if not found.
+ */
+static
+struct idpf_vport *idpf_vid_to_vport(struct idpf_adapter *adapter, u32 v_id)
+{
+ u16 num_max_vports = idpf_get_max_vports(adapter);
+ int i;
+
+ for (i = 0; i < num_max_vports; i++)
+ if (adapter->vport_ids[i] == v_id)
+ return adapter->vports[i];
+
+ return NULL;
+}
+
+/**
+ * idpf_handle_event_link - Handle link event message
+ * @adapter: private data struct
+ * @v2e: virtchnl event message
+ */
+static void idpf_handle_event_link(struct idpf_adapter *adapter,
+ const struct virtchnl2_event *v2e)
+{
+ struct idpf_netdev_priv *np;
+ struct idpf_vport *vport;
+
+ vport = idpf_vid_to_vport(adapter, le32_to_cpu(v2e->vport_id));
+ if (!vport) {
+ dev_err_ratelimited(&adapter->pdev->dev, "Failed to find vport_id %d for link event\n",
+ v2e->vport_id);
+ return;
+ }
+ np = netdev_priv(vport->netdev);
+
+ vport->link_speed_mbps = le32_to_cpu(v2e->link_speed);
+
+ if (vport->link_up == v2e->link_status)
+ return;
+
+ vport->link_up = v2e->link_status;
+
+ if (np->state != __IDPF_VPORT_UP)
+ return;
+
+ if (vport->link_up) {
+ netif_tx_start_all_queues(vport->netdev);
+ netif_carrier_on(vport->netdev);
+ } else {
+ netif_tx_stop_all_queues(vport->netdev);
+ netif_carrier_off(vport->netdev);
+ }
+}
+
/**
* idpf_recv_event_msg - Receive virtchnl event message
* @vport: virtual port structure
@@ -13,33 +71,24 @@
static void idpf_recv_event_msg(struct idpf_vport *vport,
struct idpf_ctlq_msg *ctlq_msg)
{
- struct idpf_netdev_priv *np = netdev_priv(vport->netdev);
+ int payload_size = ctlq_msg->ctx.indirect.payload->size;
struct virtchnl2_event *v2e;
- bool link_status;
u32 event;
+ if (payload_size < sizeof(*v2e)) {
+ dev_err_ratelimited(&vport->adapter->pdev->dev, "Failed to receive valid payload for event msg (op %d len %d)\n",
+ ctlq_msg->cookie.mbx.chnl_opcode,
+ payload_size);
+ return;
+ }
+
v2e = (struct virtchnl2_event *)ctlq_msg->ctx.indirect.payload->va;
event = le32_to_cpu(v2e->event);
switch (event) {
case VIRTCHNL2_EVENT_LINK_CHANGE:
- vport->link_speed_mbps = le32_to_cpu(v2e->link_speed);
- link_status = v2e->link_status;
-
- if (vport->link_up == link_status)
- break;
-
- vport->link_up = link_status;
- if (np->state == __IDPF_VPORT_UP) {
- if (vport->link_up) {
- netif_carrier_on(vport->netdev);
- netif_tx_start_all_queues(vport->netdev);
- } else {
- netif_tx_stop_all_queues(vport->netdev);
- netif_carrier_off(vport->netdev);
- }
- }
- break;
+ idpf_handle_event_link(vport->adapter, v2e);
+ return;
default:
dev_err(&vport->adapter->pdev->dev,
"Unknown event %d from PF\n", event);
@@ -283,89 +332,6 @@ static int idpf_find_vport(struct idpf_adapter *adapter,
return err;
}
-/**
- * idpf_copy_data_to_vc_buf - Copy the virtchnl response data into the buffer.
- * @adapter: driver specific private structure
- * @vport: virtual port structure
- * @ctlq_msg: msg to copy from
- * @err_enum: err bit to set on error
- *
- * Copies the payload from ctlq_msg into virtchnl buffer. Returns 0 on success,
- * negative on failure.
- */
-static int idpf_copy_data_to_vc_buf(struct idpf_adapter *adapter,
- struct idpf_vport *vport,
- struct idpf_ctlq_msg *ctlq_msg,
- enum idpf_vport_vc_state err_enum)
-{
- if (ctlq_msg->cookie.mbx.chnl_retval) {
- if (vport)
- set_bit(err_enum, vport->vc_state);
- else
- set_bit(err_enum, adapter->vc_state);
-
- return -EINVAL;
- }
-
- if (vport)
- memcpy(vport->vc_msg, ctlq_msg->ctx.indirect.payload->va,
- min_t(int, ctlq_msg->ctx.indirect.payload->size,
- IDPF_CTLQ_MAX_BUF_LEN));
- else
- memcpy(adapter->vc_msg, ctlq_msg->ctx.indirect.payload->va,
- min_t(int, ctlq_msg->ctx.indirect.payload->size,
- IDPF_CTLQ_MAX_BUF_LEN));
-
- return 0;
-}
-
-/**
- * idpf_recv_vchnl_op - helper function with common logic when handling the
- * reception of VIRTCHNL OPs.
- * @adapter: driver specific private structure
- * @vport: virtual port structure
- * @ctlq_msg: msg to copy from
- * @state: state bit used on timeout check
- * @err_state: err bit to set on error
- */
-static void idpf_recv_vchnl_op(struct idpf_adapter *adapter,
- struct idpf_vport *vport,
- struct idpf_ctlq_msg *ctlq_msg,
- enum idpf_vport_vc_state state,
- enum idpf_vport_vc_state err_state)
-{
- wait_queue_head_t *vchnl_wq;
- int err;
-
- if (vport)
- vchnl_wq = &vport->vchnl_wq;
- else
- vchnl_wq = &adapter->vchnl_wq;
-
- err = idpf_copy_data_to_vc_buf(adapter, vport, ctlq_msg, err_state);
- if (wq_has_sleeper(vchnl_wq)) {
- if (vport)
- set_bit(state, vport->vc_state);
- else
- set_bit(state, adapter->vc_state);
-
- wake_up(vchnl_wq);
- } else {
- if (!err) {
- dev_warn(&adapter->pdev->dev, "opcode %d received without waiting thread\n",
- ctlq_msg->cookie.mbx.chnl_opcode);
- } else {
- /* Clear the errors since there is no sleeper to pass
- * them on
- */
- if (vport)
- clear_bit(err_state, vport->vc_state);
- else
- clear_bit(err_state, adapter->vc_state);
- }
- }
-}
-
/* API for virtchnl "transaction" support ("xn" for short).
*
* We are reusing the completion lock to serialize the accesses to the
@@ -760,8 +726,6 @@ int idpf_recv_mb_msg(struct idpf_adapter *adapter, u32 op,
int err;
while (1) {
- struct idpf_vport_config *vport_config;
-
/* Try to get one message */
num_q_msg = 1;
dma_mem = NULL;
@@ -815,112 +779,20 @@ int idpf_recv_mb_msg(struct idpf_adapter *adapter, u32 op,
case VIRTCHNL2_OP_DEL_QUEUES:
case VIRTCHNL2_OP_MAP_QUEUE_VECTOR:
case VIRTCHNL2_OP_UNMAP_QUEUE_VECTOR:
- err = idpf_vc_xn_forward_reply(adapter, &ctlq_msg);
- break;
case VIRTCHNL2_OP_GET_STATS:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_GET_STATS,
- IDPF_VC_GET_STATS_ERR);
- break;
case VIRTCHNL2_OP_GET_RSS_LUT:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_GET_RSS_LUT,
- IDPF_VC_GET_RSS_LUT_ERR);
- break;
case VIRTCHNL2_OP_SET_RSS_LUT:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_SET_RSS_LUT,
- IDPF_VC_SET_RSS_LUT_ERR);
- break;
case VIRTCHNL2_OP_GET_RSS_KEY:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_GET_RSS_KEY,
- IDPF_VC_GET_RSS_KEY_ERR);
- break;
case VIRTCHNL2_OP_SET_RSS_KEY:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_SET_RSS_KEY,
- IDPF_VC_SET_RSS_KEY_ERR);
- break;
case VIRTCHNL2_OP_SET_SRIOV_VFS:
- idpf_recv_vchnl_op(adapter, NULL, &ctlq_msg,
- IDPF_VC_SET_SRIOV_VFS,
- IDPF_VC_SET_SRIOV_VFS_ERR);
- break;
case VIRTCHNL2_OP_ALLOC_VECTORS:
- idpf_recv_vchnl_op(adapter, NULL, &ctlq_msg,
- IDPF_VC_ALLOC_VECTORS,
- IDPF_VC_ALLOC_VECTORS_ERR);
- break;
case VIRTCHNL2_OP_DEALLOC_VECTORS:
- idpf_recv_vchnl_op(adapter, NULL, &ctlq_msg,
- IDPF_VC_DEALLOC_VECTORS,
- IDPF_VC_DEALLOC_VECTORS_ERR);
- break;
case VIRTCHNL2_OP_GET_PTYPE_INFO:
- idpf_recv_vchnl_op(adapter, NULL, &ctlq_msg,
- IDPF_VC_GET_PTYPE_INFO,
- IDPF_VC_GET_PTYPE_INFO_ERR);
- break;
case VIRTCHNL2_OP_LOOPBACK:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_LOOPBACK_STATE,
- IDPF_VC_LOOPBACK_STATE_ERR);
- break;
case VIRTCHNL2_OP_CONFIG_PROMISCUOUS_MODE:
- /* This message can only be sent asynchronously. As
- * such we'll have lost the context in which it was
- * called and thus can only really report if it looks
- * like an error occurred. Don't bother setting ERR bit
- * or waking chnl_wq since no work queue will be waiting
- * to read the message.
- */
- if (ctlq_msg.cookie.mbx.chnl_retval) {
- dev_err(&adapter->pdev->dev, "Failed to set promiscuous mode: %d\n",
- ctlq_msg.cookie.mbx.chnl_retval);
- }
- break;
case VIRTCHNL2_OP_ADD_MAC_ADDR:
- vport_config = adapter->vport_config[vport->idx];
- if (test_and_clear_bit(IDPF_VPORT_ADD_MAC_REQ,
- vport_config->flags)) {
- /* Message was sent asynchronously. We don't
- * normally print errors here, instead
- * prefer to handle errors in the function
- * calling wait_for_event. However, if
- * asynchronous, the context in which the
- * message was sent is lost. We can't really do
- * anything about at it this point, but we
- * should at a minimum indicate that it looks
- * like something went wrong. Also don't bother
- * setting ERR bit or waking vchnl_wq since no
- * one will be waiting to read the async
- * message.
- */
- if (ctlq_msg.cookie.mbx.chnl_retval)
- dev_err(&adapter->pdev->dev, "Failed to add MAC address: %d\n",
- ctlq_msg.cookie.mbx.chnl_retval);
- break;
- }
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_ADD_MAC_ADDR,
- IDPF_VC_ADD_MAC_ADDR_ERR);
- break;
case VIRTCHNL2_OP_DEL_MAC_ADDR:
- vport_config = adapter->vport_config[vport->idx];
- if (test_and_clear_bit(IDPF_VPORT_DEL_MAC_REQ,
- vport_config->flags)) {
- /* Message was sent asynchronously like the
- * VIRTCHNL2_OP_ADD_MAC_ADDR
- */
- if (ctlq_msg.cookie.mbx.chnl_retval)
- dev_err(&adapter->pdev->dev, "Failed to delete MAC address: %d\n",
- ctlq_msg.cookie.mbx.chnl_retval);
- break;
- }
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_DEL_MAC_ADDR,
- IDPF_VC_DEL_MAC_ADDR_ERR);
+ err = idpf_vc_xn_forward_reply(adapter, &ctlq_msg);
break;
case VIRTCHNL2_OP_EVENT:
idpf_recv_event_msg(vport, &ctlq_msg);
@@ -953,116 +825,6 @@ int idpf_recv_mb_msg(struct idpf_adapter *adapter, u32 op,
return err;
}
-/**
- * __idpf_wait_for_event - wrapper function for wait on virtchannel response
- * @adapter: Driver private data structure
- * @vport: virtual port structure
- * @state: check on state upon timeout
- * @err_check: check if this specific error bit is set
- * @timeout: Max time to wait
- *
- * Checks if state is set upon expiry of timeout. Returns 0 on success,
- * negative on failure.
- */
-static int __idpf_wait_for_event(struct idpf_adapter *adapter,
- struct idpf_vport *vport,
- enum idpf_vport_vc_state state,
- enum idpf_vport_vc_state err_check,
- int timeout)
-{
- int time_to_wait, num_waits;
- wait_queue_head_t *vchnl_wq;
- unsigned long *vc_state;
-
- time_to_wait = ((timeout <= IDPF_MAX_WAIT) ? timeout : IDPF_MAX_WAIT);
- num_waits = ((timeout <= IDPF_MAX_WAIT) ? 1 : timeout / IDPF_MAX_WAIT);
-
- if (vport) {
- vchnl_wq = &vport->vchnl_wq;
- vc_state = vport->vc_state;
- } else {
- vchnl_wq = &adapter->vchnl_wq;
- vc_state = adapter->vc_state;
- }
-
- while (num_waits) {
- int event;
-
- /* If we are here and a reset is detected do not wait but
- * return. Reset timing is out of drivers control. So
- * while we are cleaning resources as part of reset if the
- * underlying HW mailbox is gone, wait on mailbox messages
- * is not meaningful
- */
- if (idpf_is_reset_detected(adapter))
- return 0;
-
- event = wait_event_timeout(*vchnl_wq,
- test_and_clear_bit(state, vc_state),
- msecs_to_jiffies(time_to_wait));
- if (event) {
- if (test_and_clear_bit(err_check, vc_state)) {
- dev_err(&adapter->pdev->dev, "VC response error %s\n",
- idpf_vport_vc_state_str[err_check]);
-
- return -EINVAL;
- }
-
- return 0;
- }
- num_waits--;
- }
-
- /* Timeout occurred */
- dev_err(&adapter->pdev->dev, "VC timeout, state = %s\n",
- idpf_vport_vc_state_str[state]);
-
- return -ETIMEDOUT;
-}
-
-/**
- * idpf_min_wait_for_event - wait for virtchannel response
- * @adapter: Driver private data structure
- * @vport: virtual port structure
- * @state: check on state upon timeout
- * @err_check: check if this specific error bit is set
- *
- * Returns 0 on success, negative on failure.
- */
-static int idpf_min_wait_for_event(struct idpf_adapter *adapter,
- struct idpf_vport *vport,
- enum idpf_vport_vc_state state,
- enum idpf_vport_vc_state err_check)
-{
- return __idpf_wait_for_event(adapter, vport, state, err_check,
- IDPF_WAIT_FOR_EVENT_TIMEO_MIN);
-}
-
-/**
- * idpf_wait_for_event - wait for virtchannel response
- * @adapter: Driver private data structure
- * @vport: virtual port structure
- * @state: check on state upon timeout after 500ms
- * @err_check: check if this specific error bit is set
- *
- * Returns 0 on success, negative on failure.
- */
-static int idpf_wait_for_event(struct idpf_adapter *adapter,
- struct idpf_vport *vport,
- enum idpf_vport_vc_state state,
- enum idpf_vport_vc_state err_check)
-{
- /* Increasing the timeout in __IDPF_INIT_SW flow to consider large
- * number of VF's mailbox message responses. When a message is received
- * on mailbox, this thread is woken up by the idpf_recv_mb_msg before
- * the timeout expires. Only in the error case i.e. if no message is
- * received on mailbox, we wait for the complete timeout which is
- * less likely to happen.
- */
- return __idpf_wait_for_event(adapter, vport, state, err_check,
- IDPF_WAIT_FOR_EVENT_TIMEO);
-}
-
/**
* idpf_wait_for_marker_event - wait for software marker response
* @vport: virtual port data structure
@@ -2470,53 +2232,49 @@ int idpf_send_add_queues_msg(const struct idpf_vport *vport, u16 num_tx_q,
*/
int idpf_send_alloc_vectors_msg(struct idpf_adapter *adapter, u16 num_vectors)
{
- struct virtchnl2_alloc_vectors *alloc_vec, *rcvd_vec;
- struct virtchnl2_alloc_vectors ac = { };
+ struct virtchnl2_alloc_vectors *rcvd_vec __free(kfree) = NULL;
+ struct idpf_vc_xn_params xn_params = {};
+ struct virtchnl2_alloc_vectors ac = {};
+ ssize_t reply_sz;
u16 num_vchunks;
- int size, err;
+ int size;
ac.num_vectors = cpu_to_le16(num_vectors);
- mutex_lock(&adapter->vc_buf_lock);
-
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_ALLOC_VECTORS,
- sizeof(ac), (u8 *)&ac, 0);
- if (err)
- goto rel_lock;
+ rcvd_vec = kzalloc(IDPF_CTLQ_MAX_BUF_LEN, GFP_KERNEL);
+ if (!rcvd_vec)
+ return -ENOMEM;
- err = idpf_wait_for_event(adapter, NULL, IDPF_VC_ALLOC_VECTORS,
- IDPF_VC_ALLOC_VECTORS_ERR);
- if (err)
- goto rel_lock;
+ xn_params.vc_op = VIRTCHNL2_OP_ALLOC_VECTORS;
+ xn_params.send_buf.iov_base = ∾
+ xn_params.send_buf.iov_len = sizeof(ac);
+ xn_params.recv_buf.iov_base = rcvd_vec;
+ xn_params.recv_buf.iov_len = IDPF_CTLQ_MAX_BUF_LEN;
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
+ reply_sz = idpf_vc_xn_exec(adapter, xn_params);
+ if (reply_sz < 0)
+ return reply_sz;
- rcvd_vec = (struct virtchnl2_alloc_vectors *)adapter->vc_msg;
num_vchunks = le16_to_cpu(rcvd_vec->vchunks.num_vchunks);
-
size = struct_size(rcvd_vec, vchunks.vchunks, num_vchunks);
- if (size > sizeof(adapter->vc_msg)) {
- err = -EINVAL;
- goto rel_lock;
- }
+ if (reply_sz < size)
+ return -EIO;
+
+ if (size > IDPF_CTLQ_MAX_BUF_LEN)
+ return -EINVAL;
kfree(adapter->req_vec_chunks);
- adapter->req_vec_chunks = NULL;
- adapter->req_vec_chunks = kmemdup(adapter->vc_msg, size, GFP_KERNEL);
- if (!adapter->req_vec_chunks) {
- err = -ENOMEM;
- goto rel_lock;
- }
+ adapter->req_vec_chunks = kmemdup(rcvd_vec, size, GFP_KERNEL);
+ if (!adapter->req_vec_chunks)
+ return -ENOMEM;
- alloc_vec = adapter->req_vec_chunks;
- if (le16_to_cpu(alloc_vec->num_vectors) < num_vectors) {
+ if (le16_to_cpu(adapter->req_vec_chunks->num_vectors) < num_vectors) {
kfree(adapter->req_vec_chunks);
adapter->req_vec_chunks = NULL;
- err = -EINVAL;
+ return -EINVAL;
}
-rel_lock:
- mutex_unlock(&adapter->vc_buf_lock);
-
- return err;
+ return 0;
}
/**
@@ -2529,29 +2287,24 @@ int idpf_send_dealloc_vectors_msg(struct idpf_adapter *adapter)
{
struct virtchnl2_alloc_vectors *ac = adapter->req_vec_chunks;
struct virtchnl2_vector_chunks *vcs = &ac->vchunks;
- int buf_size, err;
+ struct idpf_vc_xn_params xn_params = {};
+ ssize_t reply_sz;
+ int buf_size;
buf_size = struct_size(vcs, vchunks, le16_to_cpu(vcs->num_vchunks));
- mutex_lock(&adapter->vc_buf_lock);
-
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_DEALLOC_VECTORS, buf_size,
- (u8 *)vcs, 0);
- if (err)
- goto rel_lock;
-
- err = idpf_min_wait_for_event(adapter, NULL, IDPF_VC_DEALLOC_VECTORS,
- IDPF_VC_DEALLOC_VECTORS_ERR);
- if (err)
- goto rel_lock;
+ xn_params.vc_op = VIRTCHNL2_OP_DEALLOC_VECTORS;
+ xn_params.send_buf.iov_base = vcs;
+ xn_params.send_buf.iov_len = buf_size;
+ xn_params.timeout_ms = IDPF_VC_XN_MIN_TIMEOUT_MSEC;
+ reply_sz = idpf_vc_xn_exec(adapter, xn_params);
+ if (reply_sz < 0)
+ return reply_sz;
kfree(adapter->req_vec_chunks);
adapter->req_vec_chunks = NULL;
-rel_lock:
- mutex_unlock(&adapter->vc_buf_lock);
-
- return err;
+ return 0;
}
/**
@@ -2574,25 +2327,18 @@ static int idpf_get_max_vfs(struct idpf_adapter *adapter)
*/
int idpf_send_set_sriov_vfs_msg(struct idpf_adapter *adapter, u16 num_vfs)
{
- struct virtchnl2_sriov_vfs_info svi = { };
- int err;
+ struct virtchnl2_sriov_vfs_info svi = {};
+ struct idpf_vc_xn_params xn_params = {};
+ ssize_t reply_sz;
svi.num_vfs = cpu_to_le16(num_vfs);
+ xn_params.vc_op = VIRTCHNL2_OP_SET_SRIOV_VFS;
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
+ xn_params.send_buf.iov_base = &svi;
+ xn_params.send_buf.iov_len = sizeof(svi);
+ reply_sz = idpf_vc_xn_exec(adapter, xn_params);
- mutex_lock(&adapter->vc_buf_lock);
-
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_SET_SRIOV_VFS,
- sizeof(svi), (u8 *)&svi, 0);
- if (err)
- goto rel_lock;
-
- err = idpf_wait_for_event(adapter, NULL, IDPF_VC_SET_SRIOV_VFS,
- IDPF_VC_SET_SRIOV_VFS_ERR);
-
-rel_lock:
- mutex_unlock(&adapter->vc_buf_lock);
-
- return err;
+ return reply_sz < 0 ? reply_sz : 0;
}
/**
@@ -2605,10 +2351,10 @@ int idpf_send_get_stats_msg(struct idpf_vport *vport)
{
struct idpf_netdev_priv *np = netdev_priv(vport->netdev);
struct rtnl_link_stats64 *netstats = &np->netstats;
- struct idpf_adapter *adapter = vport->adapter;
- struct virtchnl2_vport_stats stats_msg = { };
- struct virtchnl2_vport_stats *stats;
- int err;
+ struct virtchnl2_vport_stats stats_msg = {};
+ struct idpf_vc_xn_params xn_params = {};
+ ssize_t reply_sz;
+
/* Don't send get_stats message if the link is down */
if (np->state <= __IDPF_VPORT_DOWN)
@@ -2616,46 +2362,38 @@ int idpf_send_get_stats_msg(struct idpf_vport *vport)
stats_msg.vport_id = cpu_to_le32(vport->vport_id);
- mutex_lock(&vport->vc_buf_lock);
-
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_GET_STATS,
- sizeof(struct virtchnl2_vport_stats),
- (u8 *)&stats_msg, 0);
- if (err)
- goto rel_lock;
-
- err = idpf_wait_for_event(adapter, vport, IDPF_VC_GET_STATS,
- IDPF_VC_GET_STATS_ERR);
- if (err)
- goto rel_lock;
+ xn_params.vc_op = VIRTCHNL2_OP_GET_STATS;
+ xn_params.send_buf.iov_base = &stats_msg;
+ xn_params.send_buf.iov_len = sizeof(stats_msg);
+ xn_params.recv_buf = xn_params.send_buf;
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
- stats = (struct virtchnl2_vport_stats *)vport->vc_msg;
+ reply_sz = idpf_vc_xn_exec(vport->adapter, xn_params);
+ if (reply_sz < 0)
+ return reply_sz;
+ if (reply_sz < sizeof(stats_msg))
+ return -EIO;
spin_lock_bh(&np->stats_lock);
- netstats->rx_packets = le64_to_cpu(stats->rx_unicast) +
- le64_to_cpu(stats->rx_multicast) +
- le64_to_cpu(stats->rx_broadcast);
- netstats->rx_bytes = le64_to_cpu(stats->rx_bytes);
- netstats->rx_dropped = le64_to_cpu(stats->rx_discards);
- netstats->rx_over_errors = le64_to_cpu(stats->rx_overflow_drop);
- netstats->rx_length_errors = le64_to_cpu(stats->rx_invalid_frame_length);
-
- netstats->tx_packets = le64_to_cpu(stats->tx_unicast) +
- le64_to_cpu(stats->tx_multicast) +
- le64_to_cpu(stats->tx_broadcast);
- netstats->tx_bytes = le64_to_cpu(stats->tx_bytes);
- netstats->tx_errors = le64_to_cpu(stats->tx_errors);
- netstats->tx_dropped = le64_to_cpu(stats->tx_discards);
-
- vport->port_stats.vport_stats = *stats;
+ netstats->rx_packets = le64_to_cpu(stats_msg.rx_unicast) +
+ le64_to_cpu(stats_msg.rx_multicast) +
+ le64_to_cpu(stats_msg.rx_broadcast);
+ netstats->tx_packets = le64_to_cpu(stats_msg.tx_unicast) +
+ le64_to_cpu(stats_msg.tx_multicast) +
+ le64_to_cpu(stats_msg.tx_broadcast);
+ netstats->rx_bytes = le64_to_cpu(stats_msg.rx_bytes);
+ netstats->tx_bytes = le64_to_cpu(stats_msg.tx_bytes);
+ netstats->rx_errors = le64_to_cpu(stats_msg.rx_errors);
+ netstats->tx_errors = le64_to_cpu(stats_msg.tx_errors);
+ netstats->rx_dropped = le64_to_cpu(stats_msg.rx_discards);
+ netstats->tx_dropped = le64_to_cpu(stats_msg.tx_discards);
+
+ vport->port_stats.vport_stats = stats_msg;
spin_unlock_bh(&np->stats_lock);
-rel_lock:
- mutex_unlock(&vport->vc_buf_lock);
-
- return err;
+ return 0;
}
/**
@@ -2667,70 +2405,70 @@ int idpf_send_get_stats_msg(struct idpf_vport *vport)
*/
int idpf_send_get_set_rss_lut_msg(struct idpf_vport *vport, bool get)
{
- struct idpf_adapter *adapter = vport->adapter;
- struct virtchnl2_rss_lut *recv_rl;
+ struct virtchnl2_rss_lut *recv_rl __free(kfree) = NULL;
+ struct virtchnl2_rss_lut *rl __free(kfree) = NULL;
+ struct idpf_vc_xn_params xn_params = {};
struct idpf_rss_data *rss_data;
- struct virtchnl2_rss_lut *rl;
int buf_size, lut_buf_size;
- int i, err;
+ ssize_t reply_sz;
+ int i;
- rss_data = &adapter->vport_config[vport->idx]->user_config.rss_data;
+ rss_data =
+ &vport->adapter->vport_config[vport->idx]->user_config.rss_data;
buf_size = struct_size(rl, lut, rss_data->rss_lut_size);
rl = kzalloc(buf_size, GFP_KERNEL);
if (!rl)
return -ENOMEM;
rl->vport_id = cpu_to_le32(vport->vport_id);
- mutex_lock(&vport->vc_buf_lock);
- if (!get) {
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
+ xn_params.send_buf.iov_base = rl;
+ xn_params.send_buf.iov_len = buf_size;
+
+ if (get) {
+ recv_rl = kzalloc(IDPF_CTLQ_MAX_BUF_LEN, GFP_KERNEL);
+ if (!recv_rl)
+ return -ENOMEM;
+ xn_params.vc_op = VIRTCHNL2_OP_GET_RSS_LUT;
+ xn_params.recv_buf.iov_base = recv_rl;
+ xn_params.recv_buf.iov_len = IDPF_CTLQ_MAX_BUF_LEN;
+ } else {
rl->lut_entries = cpu_to_le16(rss_data->rss_lut_size);
for (i = 0; i < rss_data->rss_lut_size; i++)
rl->lut[i] = cpu_to_le32(rss_data->rss_lut[i]);
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_SET_RSS_LUT,
- buf_size, (u8 *)rl, 0);
- if (err)
- goto free_mem;
-
- err = idpf_wait_for_event(adapter, vport, IDPF_VC_SET_RSS_LUT,
- IDPF_VC_SET_RSS_LUT_ERR);
-
- goto free_mem;
+ xn_params.vc_op = VIRTCHNL2_OP_SET_RSS_LUT;
}
+ reply_sz = idpf_vc_xn_exec(vport->adapter, xn_params);
+ if (reply_sz < 0)
+ return reply_sz;
+ if (!get)
+ return 0;
+ if (reply_sz < sizeof(struct virtchnl2_rss_lut))
+ return -EIO;
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_GET_RSS_LUT,
- buf_size, (u8 *)rl, 0);
- if (err)
- goto free_mem;
-
- err = idpf_wait_for_event(adapter, vport, IDPF_VC_GET_RSS_LUT,
- IDPF_VC_GET_RSS_LUT_ERR);
- if (err)
- goto free_mem;
+ lut_buf_size = le16_to_cpu(recv_rl->lut_entries) * sizeof(u32);
+ if (reply_sz < lut_buf_size)
+ return -EIO;
- recv_rl = (struct virtchnl2_rss_lut *)vport->vc_msg;
+ /* size didn't change, we can reuse existing lut buf */
if (rss_data->rss_lut_size == le16_to_cpu(recv_rl->lut_entries))
goto do_memcpy;
rss_data->rss_lut_size = le16_to_cpu(recv_rl->lut_entries);
kfree(rss_data->rss_lut);
- lut_buf_size = rss_data->rss_lut_size * sizeof(u32);
rss_data->rss_lut = kzalloc(lut_buf_size, GFP_KERNEL);
if (!rss_data->rss_lut) {
rss_data->rss_lut_size = 0;
- err = -ENOMEM;
- goto free_mem;
+ return -ENOMEM;
}
do_memcpy:
- memcpy(rss_data->rss_lut, vport->vc_msg, rss_data->rss_lut_size);
-free_mem:
- mutex_unlock(&vport->vc_buf_lock);
- kfree(rl);
+ memcpy(rss_data->rss_lut, recv_rl->lut, rss_data->rss_lut_size);
- return err;
+ return 0;
}
/**
@@ -2742,68 +2480,70 @@ int idpf_send_get_set_rss_lut_msg(struct idpf_vport *vport, bool get)
*/
int idpf_send_get_set_rss_key_msg(struct idpf_vport *vport, bool get)
{
- struct idpf_adapter *adapter = vport->adapter;
- struct virtchnl2_rss_key *recv_rk;
+ struct virtchnl2_rss_key *recv_rk __free(kfree) = NULL;
+ struct virtchnl2_rss_key *rk __free(kfree) = NULL;
+ struct idpf_vc_xn_params xn_params = {};
struct idpf_rss_data *rss_data;
- struct virtchnl2_rss_key *rk;
- int i, buf_size, err;
+ ssize_t reply_sz;
+ int i, buf_size;
+ u16 key_size;
- rss_data = &adapter->vport_config[vport->idx]->user_config.rss_data;
+ rss_data =
+ &vport->adapter->vport_config[vport->idx]->user_config.rss_data;
buf_size = struct_size(rk, key_flex, rss_data->rss_key_size);
rk = kzalloc(buf_size, GFP_KERNEL);
if (!rk)
return -ENOMEM;
rk->vport_id = cpu_to_le32(vport->vport_id);
- mutex_lock(&vport->vc_buf_lock);
-
+ xn_params.send_buf.iov_base = rk;
+ xn_params.send_buf.iov_len = buf_size;
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
if (get) {
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_GET_RSS_KEY,
- buf_size, (u8 *)rk, 0);
- if (err)
- goto error;
+ recv_rk = kzalloc(IDPF_CTLQ_MAX_BUF_LEN, GFP_KERNEL);
+ if (!recv_rk)
+ return -ENOMEM;
- err = idpf_wait_for_event(adapter, vport, IDPF_VC_GET_RSS_KEY,
- IDPF_VC_GET_RSS_KEY_ERR);
- if (err)
- goto error;
-
- recv_rk = (struct virtchnl2_rss_key *)vport->vc_msg;
- if (rss_data->rss_key_size !=
- le16_to_cpu(recv_rk->key_len)) {
- rss_data->rss_key_size =
- min_t(u16, NETDEV_RSS_KEY_LEN,
- le16_to_cpu(recv_rk->key_len));
- kfree(rss_data->rss_key);
- rss_data->rss_key = kzalloc(rss_data->rss_key_size,
- GFP_KERNEL);
- if (!rss_data->rss_key) {
- rss_data->rss_key_size = 0;
- err = -ENOMEM;
- goto error;
- }
- }
- memcpy(rss_data->rss_key, recv_rk->key_flex,
- rss_data->rss_key_size);
+ xn_params.vc_op = VIRTCHNL2_OP_GET_RSS_KEY;
+ xn_params.recv_buf.iov_base = recv_rk;
+ xn_params.recv_buf.iov_len = IDPF_CTLQ_MAX_BUF_LEN;
} else {
rk->key_len = cpu_to_le16(rss_data->rss_key_size);
for (i = 0; i < rss_data->rss_key_size; i++)
rk->key_flex[i] = rss_data->rss_key[i];
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_SET_RSS_KEY,
- buf_size, (u8 *)rk, 0);
- if (err)
- goto error;
+ xn_params.vc_op = VIRTCHNL2_OP_SET_RSS_KEY;
+ }
+
+ reply_sz = idpf_vc_xn_exec(vport->adapter, xn_params);
+ if (reply_sz < 0)
+ return reply_sz;
+ if (!get)
+ return 0;
+ if (reply_sz < sizeof(struct virtchnl2_rss_key))
+ return -EIO;
- err = idpf_wait_for_event(adapter, vport, IDPF_VC_SET_RSS_KEY,
- IDPF_VC_SET_RSS_KEY_ERR);
+ key_size = min_t(u16, NETDEV_RSS_KEY_LEN,
+ le16_to_cpu(recv_rk->key_len));
+ if (reply_sz < key_size)
+ return -EIO;
+
+ /* key len didn't change, reuse existing buf */
+ if (rss_data->rss_key_size == key_size)
+ goto do_memcpy;
+
+ rss_data->rss_key_size = key_size;
+ kfree(rss_data->rss_key);
+ rss_data->rss_key = kzalloc(key_size, GFP_KERNEL);
+ if (!rss_data->rss_key) {
+ rss_data->rss_key_size = 0;
+ return -ENOMEM;
}
-error:
- mutex_unlock(&vport->vc_buf_lock);
- kfree(rk);
+do_memcpy:
+ memcpy(rss_data->rss_key, recv_rk->key_flex, rss_data->rss_key_size);
- return err;
+ return 0;
}
/**
@@ -2855,13 +2595,15 @@ static void idpf_fill_ptype_lookup(struct idpf_rx_ptype_decoded *ptype,
*/
int idpf_send_get_rx_ptype_msg(struct idpf_vport *vport)
{
+ struct virtchnl2_get_ptype_info *get_ptype_info __free(kfree) = NULL;
+ struct virtchnl2_get_ptype_info *ptype_info __free(kfree) = NULL;
struct idpf_rx_ptype_decoded *ptype_lkup = vport->rx_ptype_lkup;
- struct virtchnl2_get_ptype_info get_ptype_info;
int max_ptype, ptypes_recvd = 0, ptype_offset;
struct idpf_adapter *adapter = vport->adapter;
- struct virtchnl2_get_ptype_info *ptype_info;
+ struct idpf_vc_xn_params xn_params = {};
u16 next_ptype_id = 0;
- int err = 0, i, j, k;
+ ssize_t reply_sz;
+ int i, j, k;
if (idpf_is_queue_model_split(vport->rxq_model))
max_ptype = IDPF_RX_MAX_PTYPE;
@@ -2870,43 +2612,44 @@ int idpf_send_get_rx_ptype_msg(struct idpf_vport *vport)
memset(vport->rx_ptype_lkup, 0, sizeof(vport->rx_ptype_lkup));
+ get_ptype_info = kzalloc(sizeof(*get_ptype_info), GFP_KERNEL);
+ if (!get_ptype_info)
+ return -ENOMEM;
+
ptype_info = kzalloc(IDPF_CTLQ_MAX_BUF_LEN, GFP_KERNEL);
if (!ptype_info)
return -ENOMEM;
- mutex_lock(&adapter->vc_buf_lock);
+ xn_params.vc_op = VIRTCHNL2_OP_GET_PTYPE_INFO;
+ xn_params.send_buf.iov_base = get_ptype_info;
+ xn_params.send_buf.iov_len = sizeof(*get_ptype_info);
+ xn_params.recv_buf.iov_base = ptype_info;
+ xn_params.recv_buf.iov_len = IDPF_CTLQ_MAX_BUF_LEN;
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
while (next_ptype_id < max_ptype) {
- get_ptype_info.start_ptype_id = cpu_to_le16(next_ptype_id);
+ get_ptype_info->start_ptype_id = cpu_to_le16(next_ptype_id);
if ((next_ptype_id + IDPF_RX_MAX_PTYPES_PER_BUF) > max_ptype)
- get_ptype_info.num_ptypes =
+ get_ptype_info->num_ptypes =
cpu_to_le16(max_ptype - next_ptype_id);
else
- get_ptype_info.num_ptypes =
+ get_ptype_info->num_ptypes =
cpu_to_le16(IDPF_RX_MAX_PTYPES_PER_BUF);
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_GET_PTYPE_INFO,
- sizeof(struct virtchnl2_get_ptype_info),
- (u8 *)&get_ptype_info, 0);
- if (err)
- goto vc_buf_unlock;
-
- err = idpf_wait_for_event(adapter, NULL, IDPF_VC_GET_PTYPE_INFO,
- IDPF_VC_GET_PTYPE_INFO_ERR);
- if (err)
- goto vc_buf_unlock;
+ reply_sz = idpf_vc_xn_exec(adapter, xn_params);
+ if (reply_sz < 0)
+ return reply_sz;
- memcpy(ptype_info, adapter->vc_msg, IDPF_CTLQ_MAX_BUF_LEN);
+ if (reply_sz < IDPF_CTLQ_MAX_BUF_LEN)
+ return -EIO;
ptypes_recvd += le16_to_cpu(ptype_info->num_ptypes);
- if (ptypes_recvd > max_ptype) {
- err = -EINVAL;
- goto vc_buf_unlock;
- }
+ if (ptypes_recvd > max_ptype)
+ return -EINVAL;
- next_ptype_id = le16_to_cpu(get_ptype_info.start_ptype_id) +
- le16_to_cpu(get_ptype_info.num_ptypes);
+ next_ptype_id = le16_to_cpu(get_ptype_info->start_ptype_id) +
+ le16_to_cpu(get_ptype_info->num_ptypes);
ptype_offset = IDPF_RX_PTYPE_HDR_SZ;
@@ -2919,17 +2662,13 @@ int idpf_send_get_rx_ptype_msg(struct idpf_vport *vport)
((u8 *)ptype_info + ptype_offset);
ptype_offset += IDPF_GET_PTYPE_SIZE(ptype);
- if (ptype_offset > IDPF_CTLQ_MAX_BUF_LEN) {
- err = -EINVAL;
- goto vc_buf_unlock;
- }
+ if (ptype_offset > IDPF_CTLQ_MAX_BUF_LEN)
+ return -EINVAL;
/* 0xFFFF indicates end of ptypes */
if (le16_to_cpu(ptype->ptype_id_10) ==
- IDPF_INVALID_PTYPE_ID) {
- err = 0;
- goto vc_buf_unlock;
- }
+ IDPF_INVALID_PTYPE_ID)
+ return 0;
if (idpf_is_queue_model_split(vport->rxq_model))
k = le16_to_cpu(ptype->ptype_id_10);
@@ -3057,11 +2796,7 @@ int idpf_send_get_rx_ptype_msg(struct idpf_vport *vport)
}
}
-vc_buf_unlock:
- mutex_unlock(&adapter->vc_buf_lock);
- kfree(ptype_info);
-
- return err;
+ return 0;
}
/**
@@ -3073,27 +2808,20 @@ int idpf_send_get_rx_ptype_msg(struct idpf_vport *vport)
*/
int idpf_send_ena_dis_loopback_msg(struct idpf_vport *vport)
{
+ struct idpf_vc_xn_params xn_params = {};
struct virtchnl2_loopback loopback;
- int err;
+ ssize_t reply_sz;
loopback.vport_id = cpu_to_le32(vport->vport_id);
loopback.enable = idpf_is_feature_ena(vport, NETIF_F_LOOPBACK);
- mutex_lock(&vport->vc_buf_lock);
-
- err = idpf_send_mb_msg(vport->adapter, VIRTCHNL2_OP_LOOPBACK,
- sizeof(loopback), (u8 *)&loopback, 0);
- if (err)
- goto rel_lock;
-
- err = idpf_wait_for_event(vport->adapter, vport,
- IDPF_VC_LOOPBACK_STATE,
- IDPF_VC_LOOPBACK_STATE_ERR);
-
-rel_lock:
- mutex_unlock(&vport->vc_buf_lock);
+ xn_params.vc_op = VIRTCHNL2_OP_LOOPBACK;
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
+ xn_params.send_buf.iov_base = &loopback;
+ xn_params.send_buf.iov_len = sizeof(loopback);
+ reply_sz = idpf_vc_xn_exec(vport->adapter, xn_params);
- return err;
+ return reply_sz < 0 ? reply_sz : 0;
}
/**
@@ -3836,17 +3564,20 @@ int idpf_add_del_mac_filters(struct idpf_vport *vport,
struct idpf_netdev_priv *np,
bool add, bool async)
{
- struct virtchnl2_mac_addr_list *ma_list = NULL;
+ struct virtchnl2_mac_addr_list *ma_list __free(kfree) = NULL;
+ struct virtchnl2_mac_addr *mac_addr __free(kfree) = NULL;
struct idpf_adapter *adapter = np->adapter;
+ struct idpf_vc_xn_params xn_params = {};
struct idpf_vport_config *vport_config;
- enum idpf_vport_config_flags mac_flag;
- struct pci_dev *pdev = adapter->pdev;
- enum idpf_vport_vc_state vc, vc_err;
- struct virtchnl2_mac_addr *mac_addr;
- struct idpf_mac_filter *f, *tmp;
u32 num_msgs, total_filters = 0;
- int i = 0, k, err = 0;
- u32 vop;
+ struct idpf_mac_filter *f;
+ ssize_t reply_sz;
+ int i = 0, k;
+
+ xn_params.vc_op = add ? VIRTCHNL2_OP_ADD_MAC_ADDR :
+ VIRTCHNL2_OP_DEL_MAC_ADDR;
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
+ xn_params.async = async;
vport_config = adapter->vport_config[np->vport_idx];
spin_lock_bh(&vport_config->mac_filter_list_lock);
@@ -3870,13 +3601,13 @@ int idpf_add_del_mac_filters(struct idpf_vport *vport,
mac_addr = kcalloc(total_filters, sizeof(struct virtchnl2_mac_addr),
GFP_ATOMIC);
if (!mac_addr) {
- err = -ENOMEM;
spin_unlock_bh(&vport_config->mac_filter_list_lock);
- goto error;
+
+ return -ENOMEM;
}
- list_for_each_entry_safe(f, tmp, &vport_config->user_config.mac_filter_list,
- list) {
+ list_for_each_entry(f, &vport_config->user_config.mac_filter_list,
+ list) {
if (add && f->add) {
ether_addr_copy(mac_addr[i].addr, f->macaddr);
i++;
@@ -3895,26 +3626,11 @@ int idpf_add_del_mac_filters(struct idpf_vport *vport,
spin_unlock_bh(&vport_config->mac_filter_list_lock);
- if (add) {
- vop = VIRTCHNL2_OP_ADD_MAC_ADDR;
- vc = IDPF_VC_ADD_MAC_ADDR;
- vc_err = IDPF_VC_ADD_MAC_ADDR_ERR;
- mac_flag = IDPF_VPORT_ADD_MAC_REQ;
- } else {
- vop = VIRTCHNL2_OP_DEL_MAC_ADDR;
- vc = IDPF_VC_DEL_MAC_ADDR;
- vc_err = IDPF_VC_DEL_MAC_ADDR_ERR;
- mac_flag = IDPF_VPORT_DEL_MAC_REQ;
- }
-
/* Chunk up the filters into multiple messages to avoid
* sending a control queue message buffer that is too large
*/
num_msgs = DIV_ROUND_UP(total_filters, IDPF_NUM_FILTERS_PER_MSG);
- if (!async)
- mutex_lock(&vport->vc_buf_lock);
-
for (i = 0, k = 0; i < num_msgs; i++) {
u32 entries_size, buf_size, num_entries;
@@ -3926,10 +3642,8 @@ int idpf_add_del_mac_filters(struct idpf_vport *vport,
if (!ma_list || num_entries != IDPF_NUM_FILTERS_PER_MSG) {
kfree(ma_list);
ma_list = kzalloc(buf_size, GFP_ATOMIC);
- if (!ma_list) {
- err = -ENOMEM;
- goto list_prep_error;
- }
+ if (!ma_list)
+ return -ENOMEM;
} else {
memset(ma_list, 0, buf_size);
}
@@ -3938,34 +3652,17 @@ int idpf_add_del_mac_filters(struct idpf_vport *vport,
ma_list->num_mac_addr = cpu_to_le16(num_entries);
memcpy(ma_list->mac_addr_list, &mac_addr[k], entries_size);
- if (async)
- set_bit(mac_flag, vport_config->flags);
-
- err = idpf_send_mb_msg(adapter, vop, buf_size, (u8 *)ma_list, 0);
- if (err)
- goto mbx_error;
-
- if (!async) {
- err = idpf_wait_for_event(adapter, vport, vc, vc_err);
- if (err)
- goto mbx_error;
- }
+ xn_params.send_buf.iov_base = ma_list;
+ xn_params.send_buf.iov_len = buf_size;
+ reply_sz = idpf_vc_xn_exec(adapter, xn_params);
+ if (reply_sz < 0)
+ return reply_sz;
k += num_entries;
total_filters -= num_entries;
}
-mbx_error:
- if (!async)
- mutex_unlock(&vport->vc_buf_lock);
- kfree(ma_list);
-list_prep_error:
- kfree(mac_addr);
-error:
- if (err)
- dev_err(&pdev->dev, "Failed to add or del mac filters %d", err);
-
- return err;
+ return 0;
}
/**
@@ -3982,9 +3679,10 @@ int idpf_set_promiscuous(struct idpf_adapter *adapter,
struct idpf_vport_user_config_data *config_data,
u32 vport_id)
{
+ struct idpf_vc_xn_params xn_params = {};
struct virtchnl2_promisc_info vpi;
+ ssize_t reply_sz;
u16 flags = 0;
- int err;
if (test_bit(__IDPF_PROMISC_UC, config_data->user_flags))
flags |= VIRTCHNL2_UNICAST_PROMISC;
@@ -3994,9 +3692,13 @@ int idpf_set_promiscuous(struct idpf_adapter *adapter,
vpi.vport_id = cpu_to_le32(vport_id);
vpi.flags = cpu_to_le16(flags);
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_CONFIG_PROMISCUOUS_MODE,
- sizeof(struct virtchnl2_promisc_info),
- (u8 *)&vpi, 0);
+ xn_params.vc_op = VIRTCHNL2_OP_CONFIG_PROMISCUOUS_MODE;
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
+ xn_params.send_buf.iov_base = &vpi;
+ xn_params.send_buf.iov_len = sizeof(vpi);
+ /* setting promiscuous is only ever done asynchronously */
+ xn_params.async = true;
+ reply_sz = idpf_vc_xn_exec(adapter, xn_params);
- return err;
+ return reply_sz < 0 ? reply_sz : 0;
}
--
2.40.1
^ permalink raw reply related [flat|nested] 44+ messages in thread
* [Intel-wired-lan] [PATCH v4 04/10 iwl-next] idpf: refactor remaining virtchnl messages
@ 2024-02-06 3:37 ` Alan Brady
0 siblings, 0 replies; 44+ messages in thread
From: Alan Brady @ 2024-02-06 3:37 UTC (permalink / raw)
To: intel-wired-lan
Cc: willemdebruijn.kernel, netdev, aleksander.lobakin, Alan Brady,
przemyslaw.kitszel, igor.bagnucki
This takes care of RSS/SRIOV/MAC and other misc virtchnl messages. This
again is mostly mechanical.
In absence of an async_handler for MAC filters, this will simply
generically report any errors from idpf_vc_xn_forward_async. This
maintains the existing behavior. Follow up patch will add an async
handler for MAC filters to remove bad filters from our list.
While we're here we can also make the code much nicer by converting some
variables to auto-variables where appropriate. This makes it cleaner and
less prone to memory leaking.
There's still a bit more cleanup we can do here to remove stuff that's
not being used anymore now; follow-up patches will take care of loose
ends.
Signed-off-by: Alan Brady <alan.brady@intel.com>
---
.../net/ethernet/intel/idpf/idpf_virtchnl.c | 894 ++++++------------
1 file changed, 298 insertions(+), 596 deletions(-)
diff --git a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
index 193c284719ae..6d95daf0637b 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
@@ -3,6 +3,64 @@
#include "idpf.h"
+/**
+ * idpf_vid_to_vport - Translate vport id to vport pointer
+ * @adapter: private data struct
+ * @v_id: vport id to translate
+ *
+ * Returns vport matching v_id, NULL if not found.
+ */
+static
+struct idpf_vport *idpf_vid_to_vport(struct idpf_adapter *adapter, u32 v_id)
+{
+ u16 num_max_vports = idpf_get_max_vports(adapter);
+ int i;
+
+ for (i = 0; i < num_max_vports; i++)
+ if (adapter->vport_ids[i] == v_id)
+ return adapter->vports[i];
+
+ return NULL;
+}
+
+/**
+ * idpf_handle_event_link - Handle link event message
+ * @adapter: private data struct
+ * @v2e: virtchnl event message
+ */
+static void idpf_handle_event_link(struct idpf_adapter *adapter,
+ const struct virtchnl2_event *v2e)
+{
+ struct idpf_netdev_priv *np;
+ struct idpf_vport *vport;
+
+ vport = idpf_vid_to_vport(adapter, le32_to_cpu(v2e->vport_id));
+ if (!vport) {
+ dev_err_ratelimited(&adapter->pdev->dev, "Failed to find vport_id %d for link event\n",
+ v2e->vport_id);
+ return;
+ }
+ np = netdev_priv(vport->netdev);
+
+ vport->link_speed_mbps = le32_to_cpu(v2e->link_speed);
+
+ if (vport->link_up == v2e->link_status)
+ return;
+
+ vport->link_up = v2e->link_status;
+
+ if (np->state != __IDPF_VPORT_UP)
+ return;
+
+ if (vport->link_up) {
+ netif_tx_start_all_queues(vport->netdev);
+ netif_carrier_on(vport->netdev);
+ } else {
+ netif_tx_stop_all_queues(vport->netdev);
+ netif_carrier_off(vport->netdev);
+ }
+}
+
/**
* idpf_recv_event_msg - Receive virtchnl event message
* @vport: virtual port structure
@@ -13,33 +71,24 @@
static void idpf_recv_event_msg(struct idpf_vport *vport,
struct idpf_ctlq_msg *ctlq_msg)
{
- struct idpf_netdev_priv *np = netdev_priv(vport->netdev);
+ int payload_size = ctlq_msg->ctx.indirect.payload->size;
struct virtchnl2_event *v2e;
- bool link_status;
u32 event;
+ if (payload_size < sizeof(*v2e)) {
+ dev_err_ratelimited(&vport->adapter->pdev->dev, "Failed to receive valid payload for event msg (op %d len %d)\n",
+ ctlq_msg->cookie.mbx.chnl_opcode,
+ payload_size);
+ return;
+ }
+
v2e = (struct virtchnl2_event *)ctlq_msg->ctx.indirect.payload->va;
event = le32_to_cpu(v2e->event);
switch (event) {
case VIRTCHNL2_EVENT_LINK_CHANGE:
- vport->link_speed_mbps = le32_to_cpu(v2e->link_speed);
- link_status = v2e->link_status;
-
- if (vport->link_up == link_status)
- break;
-
- vport->link_up = link_status;
- if (np->state == __IDPF_VPORT_UP) {
- if (vport->link_up) {
- netif_carrier_on(vport->netdev);
- netif_tx_start_all_queues(vport->netdev);
- } else {
- netif_tx_stop_all_queues(vport->netdev);
- netif_carrier_off(vport->netdev);
- }
- }
- break;
+ idpf_handle_event_link(vport->adapter, v2e);
+ return;
default:
dev_err(&vport->adapter->pdev->dev,
"Unknown event %d from PF\n", event);
@@ -283,89 +332,6 @@ static int idpf_find_vport(struct idpf_adapter *adapter,
return err;
}
-/**
- * idpf_copy_data_to_vc_buf - Copy the virtchnl response data into the buffer.
- * @adapter: driver specific private structure
- * @vport: virtual port structure
- * @ctlq_msg: msg to copy from
- * @err_enum: err bit to set on error
- *
- * Copies the payload from ctlq_msg into virtchnl buffer. Returns 0 on success,
- * negative on failure.
- */
-static int idpf_copy_data_to_vc_buf(struct idpf_adapter *adapter,
- struct idpf_vport *vport,
- struct idpf_ctlq_msg *ctlq_msg,
- enum idpf_vport_vc_state err_enum)
-{
- if (ctlq_msg->cookie.mbx.chnl_retval) {
- if (vport)
- set_bit(err_enum, vport->vc_state);
- else
- set_bit(err_enum, adapter->vc_state);
-
- return -EINVAL;
- }
-
- if (vport)
- memcpy(vport->vc_msg, ctlq_msg->ctx.indirect.payload->va,
- min_t(int, ctlq_msg->ctx.indirect.payload->size,
- IDPF_CTLQ_MAX_BUF_LEN));
- else
- memcpy(adapter->vc_msg, ctlq_msg->ctx.indirect.payload->va,
- min_t(int, ctlq_msg->ctx.indirect.payload->size,
- IDPF_CTLQ_MAX_BUF_LEN));
-
- return 0;
-}
-
-/**
- * idpf_recv_vchnl_op - helper function with common logic when handling the
- * reception of VIRTCHNL OPs.
- * @adapter: driver specific private structure
- * @vport: virtual port structure
- * @ctlq_msg: msg to copy from
- * @state: state bit used on timeout check
- * @err_state: err bit to set on error
- */
-static void idpf_recv_vchnl_op(struct idpf_adapter *adapter,
- struct idpf_vport *vport,
- struct idpf_ctlq_msg *ctlq_msg,
- enum idpf_vport_vc_state state,
- enum idpf_vport_vc_state err_state)
-{
- wait_queue_head_t *vchnl_wq;
- int err;
-
- if (vport)
- vchnl_wq = &vport->vchnl_wq;
- else
- vchnl_wq = &adapter->vchnl_wq;
-
- err = idpf_copy_data_to_vc_buf(adapter, vport, ctlq_msg, err_state);
- if (wq_has_sleeper(vchnl_wq)) {
- if (vport)
- set_bit(state, vport->vc_state);
- else
- set_bit(state, adapter->vc_state);
-
- wake_up(vchnl_wq);
- } else {
- if (!err) {
- dev_warn(&adapter->pdev->dev, "opcode %d received without waiting thread\n",
- ctlq_msg->cookie.mbx.chnl_opcode);
- } else {
- /* Clear the errors since there is no sleeper to pass
- * them on
- */
- if (vport)
- clear_bit(err_state, vport->vc_state);
- else
- clear_bit(err_state, adapter->vc_state);
- }
- }
-}
-
/* API for virtchnl "transaction" support ("xn" for short).
*
* We are reusing the completion lock to serialize the accesses to the
@@ -760,8 +726,6 @@ int idpf_recv_mb_msg(struct idpf_adapter *adapter, u32 op,
int err;
while (1) {
- struct idpf_vport_config *vport_config;
-
/* Try to get one message */
num_q_msg = 1;
dma_mem = NULL;
@@ -815,112 +779,20 @@ int idpf_recv_mb_msg(struct idpf_adapter *adapter, u32 op,
case VIRTCHNL2_OP_DEL_QUEUES:
case VIRTCHNL2_OP_MAP_QUEUE_VECTOR:
case VIRTCHNL2_OP_UNMAP_QUEUE_VECTOR:
- err = idpf_vc_xn_forward_reply(adapter, &ctlq_msg);
- break;
case VIRTCHNL2_OP_GET_STATS:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_GET_STATS,
- IDPF_VC_GET_STATS_ERR);
- break;
case VIRTCHNL2_OP_GET_RSS_LUT:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_GET_RSS_LUT,
- IDPF_VC_GET_RSS_LUT_ERR);
- break;
case VIRTCHNL2_OP_SET_RSS_LUT:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_SET_RSS_LUT,
- IDPF_VC_SET_RSS_LUT_ERR);
- break;
case VIRTCHNL2_OP_GET_RSS_KEY:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_GET_RSS_KEY,
- IDPF_VC_GET_RSS_KEY_ERR);
- break;
case VIRTCHNL2_OP_SET_RSS_KEY:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_SET_RSS_KEY,
- IDPF_VC_SET_RSS_KEY_ERR);
- break;
case VIRTCHNL2_OP_SET_SRIOV_VFS:
- idpf_recv_vchnl_op(adapter, NULL, &ctlq_msg,
- IDPF_VC_SET_SRIOV_VFS,
- IDPF_VC_SET_SRIOV_VFS_ERR);
- break;
case VIRTCHNL2_OP_ALLOC_VECTORS:
- idpf_recv_vchnl_op(adapter, NULL, &ctlq_msg,
- IDPF_VC_ALLOC_VECTORS,
- IDPF_VC_ALLOC_VECTORS_ERR);
- break;
case VIRTCHNL2_OP_DEALLOC_VECTORS:
- idpf_recv_vchnl_op(adapter, NULL, &ctlq_msg,
- IDPF_VC_DEALLOC_VECTORS,
- IDPF_VC_DEALLOC_VECTORS_ERR);
- break;
case VIRTCHNL2_OP_GET_PTYPE_INFO:
- idpf_recv_vchnl_op(adapter, NULL, &ctlq_msg,
- IDPF_VC_GET_PTYPE_INFO,
- IDPF_VC_GET_PTYPE_INFO_ERR);
- break;
case VIRTCHNL2_OP_LOOPBACK:
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_LOOPBACK_STATE,
- IDPF_VC_LOOPBACK_STATE_ERR);
- break;
case VIRTCHNL2_OP_CONFIG_PROMISCUOUS_MODE:
- /* This message can only be sent asynchronously. As
- * such we'll have lost the context in which it was
- * called and thus can only really report if it looks
- * like an error occurred. Don't bother setting ERR bit
- * or waking chnl_wq since no work queue will be waiting
- * to read the message.
- */
- if (ctlq_msg.cookie.mbx.chnl_retval) {
- dev_err(&adapter->pdev->dev, "Failed to set promiscuous mode: %d\n",
- ctlq_msg.cookie.mbx.chnl_retval);
- }
- break;
case VIRTCHNL2_OP_ADD_MAC_ADDR:
- vport_config = adapter->vport_config[vport->idx];
- if (test_and_clear_bit(IDPF_VPORT_ADD_MAC_REQ,
- vport_config->flags)) {
- /* Message was sent asynchronously. We don't
- * normally print errors here, instead
- * prefer to handle errors in the function
- * calling wait_for_event. However, if
- * asynchronous, the context in which the
- * message was sent is lost. We can't really do
- * anything about at it this point, but we
- * should at a minimum indicate that it looks
- * like something went wrong. Also don't bother
- * setting ERR bit or waking vchnl_wq since no
- * one will be waiting to read the async
- * message.
- */
- if (ctlq_msg.cookie.mbx.chnl_retval)
- dev_err(&adapter->pdev->dev, "Failed to add MAC address: %d\n",
- ctlq_msg.cookie.mbx.chnl_retval);
- break;
- }
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_ADD_MAC_ADDR,
- IDPF_VC_ADD_MAC_ADDR_ERR);
- break;
case VIRTCHNL2_OP_DEL_MAC_ADDR:
- vport_config = adapter->vport_config[vport->idx];
- if (test_and_clear_bit(IDPF_VPORT_DEL_MAC_REQ,
- vport_config->flags)) {
- /* Message was sent asynchronously like the
- * VIRTCHNL2_OP_ADD_MAC_ADDR
- */
- if (ctlq_msg.cookie.mbx.chnl_retval)
- dev_err(&adapter->pdev->dev, "Failed to delete MAC address: %d\n",
- ctlq_msg.cookie.mbx.chnl_retval);
- break;
- }
- idpf_recv_vchnl_op(adapter, vport, &ctlq_msg,
- IDPF_VC_DEL_MAC_ADDR,
- IDPF_VC_DEL_MAC_ADDR_ERR);
+ err = idpf_vc_xn_forward_reply(adapter, &ctlq_msg);
break;
case VIRTCHNL2_OP_EVENT:
idpf_recv_event_msg(vport, &ctlq_msg);
@@ -953,116 +825,6 @@ int idpf_recv_mb_msg(struct idpf_adapter *adapter, u32 op,
return err;
}
-/**
- * __idpf_wait_for_event - wrapper function for wait on virtchannel response
- * @adapter: Driver private data structure
- * @vport: virtual port structure
- * @state: check on state upon timeout
- * @err_check: check if this specific error bit is set
- * @timeout: Max time to wait
- *
- * Checks if state is set upon expiry of timeout. Returns 0 on success,
- * negative on failure.
- */
-static int __idpf_wait_for_event(struct idpf_adapter *adapter,
- struct idpf_vport *vport,
- enum idpf_vport_vc_state state,
- enum idpf_vport_vc_state err_check,
- int timeout)
-{
- int time_to_wait, num_waits;
- wait_queue_head_t *vchnl_wq;
- unsigned long *vc_state;
-
- time_to_wait = ((timeout <= IDPF_MAX_WAIT) ? timeout : IDPF_MAX_WAIT);
- num_waits = ((timeout <= IDPF_MAX_WAIT) ? 1 : timeout / IDPF_MAX_WAIT);
-
- if (vport) {
- vchnl_wq = &vport->vchnl_wq;
- vc_state = vport->vc_state;
- } else {
- vchnl_wq = &adapter->vchnl_wq;
- vc_state = adapter->vc_state;
- }
-
- while (num_waits) {
- int event;
-
- /* If we are here and a reset is detected do not wait but
- * return. Reset timing is out of drivers control. So
- * while we are cleaning resources as part of reset if the
- * underlying HW mailbox is gone, wait on mailbox messages
- * is not meaningful
- */
- if (idpf_is_reset_detected(adapter))
- return 0;
-
- event = wait_event_timeout(*vchnl_wq,
- test_and_clear_bit(state, vc_state),
- msecs_to_jiffies(time_to_wait));
- if (event) {
- if (test_and_clear_bit(err_check, vc_state)) {
- dev_err(&adapter->pdev->dev, "VC response error %s\n",
- idpf_vport_vc_state_str[err_check]);
-
- return -EINVAL;
- }
-
- return 0;
- }
- num_waits--;
- }
-
- /* Timeout occurred */
- dev_err(&adapter->pdev->dev, "VC timeout, state = %s\n",
- idpf_vport_vc_state_str[state]);
-
- return -ETIMEDOUT;
-}
-
-/**
- * idpf_min_wait_for_event - wait for virtchannel response
- * @adapter: Driver private data structure
- * @vport: virtual port structure
- * @state: check on state upon timeout
- * @err_check: check if this specific error bit is set
- *
- * Returns 0 on success, negative on failure.
- */
-static int idpf_min_wait_for_event(struct idpf_adapter *adapter,
- struct idpf_vport *vport,
- enum idpf_vport_vc_state state,
- enum idpf_vport_vc_state err_check)
-{
- return __idpf_wait_for_event(adapter, vport, state, err_check,
- IDPF_WAIT_FOR_EVENT_TIMEO_MIN);
-}
-
-/**
- * idpf_wait_for_event - wait for virtchannel response
- * @adapter: Driver private data structure
- * @vport: virtual port structure
- * @state: check on state upon timeout after 500ms
- * @err_check: check if this specific error bit is set
- *
- * Returns 0 on success, negative on failure.
- */
-static int idpf_wait_for_event(struct idpf_adapter *adapter,
- struct idpf_vport *vport,
- enum idpf_vport_vc_state state,
- enum idpf_vport_vc_state err_check)
-{
- /* Increasing the timeout in __IDPF_INIT_SW flow to consider large
- * number of VF's mailbox message responses. When a message is received
- * on mailbox, this thread is woken up by the idpf_recv_mb_msg before
- * the timeout expires. Only in the error case i.e. if no message is
- * received on mailbox, we wait for the complete timeout which is
- * less likely to happen.
- */
- return __idpf_wait_for_event(adapter, vport, state, err_check,
- IDPF_WAIT_FOR_EVENT_TIMEO);
-}
-
/**
* idpf_wait_for_marker_event - wait for software marker response
* @vport: virtual port data structure
@@ -2470,53 +2232,49 @@ int idpf_send_add_queues_msg(const struct idpf_vport *vport, u16 num_tx_q,
*/
int idpf_send_alloc_vectors_msg(struct idpf_adapter *adapter, u16 num_vectors)
{
- struct virtchnl2_alloc_vectors *alloc_vec, *rcvd_vec;
- struct virtchnl2_alloc_vectors ac = { };
+ struct virtchnl2_alloc_vectors *rcvd_vec __free(kfree) = NULL;
+ struct idpf_vc_xn_params xn_params = {};
+ struct virtchnl2_alloc_vectors ac = {};
+ ssize_t reply_sz;
u16 num_vchunks;
- int size, err;
+ int size;
ac.num_vectors = cpu_to_le16(num_vectors);
- mutex_lock(&adapter->vc_buf_lock);
-
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_ALLOC_VECTORS,
- sizeof(ac), (u8 *)&ac, 0);
- if (err)
- goto rel_lock;
+ rcvd_vec = kzalloc(IDPF_CTLQ_MAX_BUF_LEN, GFP_KERNEL);
+ if (!rcvd_vec)
+ return -ENOMEM;
- err = idpf_wait_for_event(adapter, NULL, IDPF_VC_ALLOC_VECTORS,
- IDPF_VC_ALLOC_VECTORS_ERR);
- if (err)
- goto rel_lock;
+ xn_params.vc_op = VIRTCHNL2_OP_ALLOC_VECTORS;
+ xn_params.send_buf.iov_base = ∾
+ xn_params.send_buf.iov_len = sizeof(ac);
+ xn_params.recv_buf.iov_base = rcvd_vec;
+ xn_params.recv_buf.iov_len = IDPF_CTLQ_MAX_BUF_LEN;
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
+ reply_sz = idpf_vc_xn_exec(adapter, xn_params);
+ if (reply_sz < 0)
+ return reply_sz;
- rcvd_vec = (struct virtchnl2_alloc_vectors *)adapter->vc_msg;
num_vchunks = le16_to_cpu(rcvd_vec->vchunks.num_vchunks);
-
size = struct_size(rcvd_vec, vchunks.vchunks, num_vchunks);
- if (size > sizeof(adapter->vc_msg)) {
- err = -EINVAL;
- goto rel_lock;
- }
+ if (reply_sz < size)
+ return -EIO;
+
+ if (size > IDPF_CTLQ_MAX_BUF_LEN)
+ return -EINVAL;
kfree(adapter->req_vec_chunks);
- adapter->req_vec_chunks = NULL;
- adapter->req_vec_chunks = kmemdup(adapter->vc_msg, size, GFP_KERNEL);
- if (!adapter->req_vec_chunks) {
- err = -ENOMEM;
- goto rel_lock;
- }
+ adapter->req_vec_chunks = kmemdup(rcvd_vec, size, GFP_KERNEL);
+ if (!adapter->req_vec_chunks)
+ return -ENOMEM;
- alloc_vec = adapter->req_vec_chunks;
- if (le16_to_cpu(alloc_vec->num_vectors) < num_vectors) {
+ if (le16_to_cpu(adapter->req_vec_chunks->num_vectors) < num_vectors) {
kfree(adapter->req_vec_chunks);
adapter->req_vec_chunks = NULL;
- err = -EINVAL;
+ return -EINVAL;
}
-rel_lock:
- mutex_unlock(&adapter->vc_buf_lock);
-
- return err;
+ return 0;
}
/**
@@ -2529,29 +2287,24 @@ int idpf_send_dealloc_vectors_msg(struct idpf_adapter *adapter)
{
struct virtchnl2_alloc_vectors *ac = adapter->req_vec_chunks;
struct virtchnl2_vector_chunks *vcs = &ac->vchunks;
- int buf_size, err;
+ struct idpf_vc_xn_params xn_params = {};
+ ssize_t reply_sz;
+ int buf_size;
buf_size = struct_size(vcs, vchunks, le16_to_cpu(vcs->num_vchunks));
- mutex_lock(&adapter->vc_buf_lock);
-
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_DEALLOC_VECTORS, buf_size,
- (u8 *)vcs, 0);
- if (err)
- goto rel_lock;
-
- err = idpf_min_wait_for_event(adapter, NULL, IDPF_VC_DEALLOC_VECTORS,
- IDPF_VC_DEALLOC_VECTORS_ERR);
- if (err)
- goto rel_lock;
+ xn_params.vc_op = VIRTCHNL2_OP_DEALLOC_VECTORS;
+ xn_params.send_buf.iov_base = vcs;
+ xn_params.send_buf.iov_len = buf_size;
+ xn_params.timeout_ms = IDPF_VC_XN_MIN_TIMEOUT_MSEC;
+ reply_sz = idpf_vc_xn_exec(adapter, xn_params);
+ if (reply_sz < 0)
+ return reply_sz;
kfree(adapter->req_vec_chunks);
adapter->req_vec_chunks = NULL;
-rel_lock:
- mutex_unlock(&adapter->vc_buf_lock);
-
- return err;
+ return 0;
}
/**
@@ -2574,25 +2327,18 @@ static int idpf_get_max_vfs(struct idpf_adapter *adapter)
*/
int idpf_send_set_sriov_vfs_msg(struct idpf_adapter *adapter, u16 num_vfs)
{
- struct virtchnl2_sriov_vfs_info svi = { };
- int err;
+ struct virtchnl2_sriov_vfs_info svi = {};
+ struct idpf_vc_xn_params xn_params = {};
+ ssize_t reply_sz;
svi.num_vfs = cpu_to_le16(num_vfs);
+ xn_params.vc_op = VIRTCHNL2_OP_SET_SRIOV_VFS;
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
+ xn_params.send_buf.iov_base = &svi;
+ xn_params.send_buf.iov_len = sizeof(svi);
+ reply_sz = idpf_vc_xn_exec(adapter, xn_params);
- mutex_lock(&adapter->vc_buf_lock);
-
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_SET_SRIOV_VFS,
- sizeof(svi), (u8 *)&svi, 0);
- if (err)
- goto rel_lock;
-
- err = idpf_wait_for_event(adapter, NULL, IDPF_VC_SET_SRIOV_VFS,
- IDPF_VC_SET_SRIOV_VFS_ERR);
-
-rel_lock:
- mutex_unlock(&adapter->vc_buf_lock);
-
- return err;
+ return reply_sz < 0 ? reply_sz : 0;
}
/**
@@ -2605,10 +2351,10 @@ int idpf_send_get_stats_msg(struct idpf_vport *vport)
{
struct idpf_netdev_priv *np = netdev_priv(vport->netdev);
struct rtnl_link_stats64 *netstats = &np->netstats;
- struct idpf_adapter *adapter = vport->adapter;
- struct virtchnl2_vport_stats stats_msg = { };
- struct virtchnl2_vport_stats *stats;
- int err;
+ struct virtchnl2_vport_stats stats_msg = {};
+ struct idpf_vc_xn_params xn_params = {};
+ ssize_t reply_sz;
+
/* Don't send get_stats message if the link is down */
if (np->state <= __IDPF_VPORT_DOWN)
@@ -2616,46 +2362,38 @@ int idpf_send_get_stats_msg(struct idpf_vport *vport)
stats_msg.vport_id = cpu_to_le32(vport->vport_id);
- mutex_lock(&vport->vc_buf_lock);
-
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_GET_STATS,
- sizeof(struct virtchnl2_vport_stats),
- (u8 *)&stats_msg, 0);
- if (err)
- goto rel_lock;
-
- err = idpf_wait_for_event(adapter, vport, IDPF_VC_GET_STATS,
- IDPF_VC_GET_STATS_ERR);
- if (err)
- goto rel_lock;
+ xn_params.vc_op = VIRTCHNL2_OP_GET_STATS;
+ xn_params.send_buf.iov_base = &stats_msg;
+ xn_params.send_buf.iov_len = sizeof(stats_msg);
+ xn_params.recv_buf = xn_params.send_buf;
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
- stats = (struct virtchnl2_vport_stats *)vport->vc_msg;
+ reply_sz = idpf_vc_xn_exec(vport->adapter, xn_params);
+ if (reply_sz < 0)
+ return reply_sz;
+ if (reply_sz < sizeof(stats_msg))
+ return -EIO;
spin_lock_bh(&np->stats_lock);
- netstats->rx_packets = le64_to_cpu(stats->rx_unicast) +
- le64_to_cpu(stats->rx_multicast) +
- le64_to_cpu(stats->rx_broadcast);
- netstats->rx_bytes = le64_to_cpu(stats->rx_bytes);
- netstats->rx_dropped = le64_to_cpu(stats->rx_discards);
- netstats->rx_over_errors = le64_to_cpu(stats->rx_overflow_drop);
- netstats->rx_length_errors = le64_to_cpu(stats->rx_invalid_frame_length);
-
- netstats->tx_packets = le64_to_cpu(stats->tx_unicast) +
- le64_to_cpu(stats->tx_multicast) +
- le64_to_cpu(stats->tx_broadcast);
- netstats->tx_bytes = le64_to_cpu(stats->tx_bytes);
- netstats->tx_errors = le64_to_cpu(stats->tx_errors);
- netstats->tx_dropped = le64_to_cpu(stats->tx_discards);
-
- vport->port_stats.vport_stats = *stats;
+ netstats->rx_packets = le64_to_cpu(stats_msg.rx_unicast) +
+ le64_to_cpu(stats_msg.rx_multicast) +
+ le64_to_cpu(stats_msg.rx_broadcast);
+ netstats->tx_packets = le64_to_cpu(stats_msg.tx_unicast) +
+ le64_to_cpu(stats_msg.tx_multicast) +
+ le64_to_cpu(stats_msg.tx_broadcast);
+ netstats->rx_bytes = le64_to_cpu(stats_msg.rx_bytes);
+ netstats->tx_bytes = le64_to_cpu(stats_msg.tx_bytes);
+ netstats->rx_errors = le64_to_cpu(stats_msg.rx_errors);
+ netstats->tx_errors = le64_to_cpu(stats_msg.tx_errors);
+ netstats->rx_dropped = le64_to_cpu(stats_msg.rx_discards);
+ netstats->tx_dropped = le64_to_cpu(stats_msg.tx_discards);
+
+ vport->port_stats.vport_stats = stats_msg;
spin_unlock_bh(&np->stats_lock);
-rel_lock:
- mutex_unlock(&vport->vc_buf_lock);
-
- return err;
+ return 0;
}
/**
@@ -2667,70 +2405,70 @@ int idpf_send_get_stats_msg(struct idpf_vport *vport)
*/
int idpf_send_get_set_rss_lut_msg(struct idpf_vport *vport, bool get)
{
- struct idpf_adapter *adapter = vport->adapter;
- struct virtchnl2_rss_lut *recv_rl;
+ struct virtchnl2_rss_lut *recv_rl __free(kfree) = NULL;
+ struct virtchnl2_rss_lut *rl __free(kfree) = NULL;
+ struct idpf_vc_xn_params xn_params = {};
struct idpf_rss_data *rss_data;
- struct virtchnl2_rss_lut *rl;
int buf_size, lut_buf_size;
- int i, err;
+ ssize_t reply_sz;
+ int i;
- rss_data = &adapter->vport_config[vport->idx]->user_config.rss_data;
+ rss_data =
+ &vport->adapter->vport_config[vport->idx]->user_config.rss_data;
buf_size = struct_size(rl, lut, rss_data->rss_lut_size);
rl = kzalloc(buf_size, GFP_KERNEL);
if (!rl)
return -ENOMEM;
rl->vport_id = cpu_to_le32(vport->vport_id);
- mutex_lock(&vport->vc_buf_lock);
- if (!get) {
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
+ xn_params.send_buf.iov_base = rl;
+ xn_params.send_buf.iov_len = buf_size;
+
+ if (get) {
+ recv_rl = kzalloc(IDPF_CTLQ_MAX_BUF_LEN, GFP_KERNEL);
+ if (!recv_rl)
+ return -ENOMEM;
+ xn_params.vc_op = VIRTCHNL2_OP_GET_RSS_LUT;
+ xn_params.recv_buf.iov_base = recv_rl;
+ xn_params.recv_buf.iov_len = IDPF_CTLQ_MAX_BUF_LEN;
+ } else {
rl->lut_entries = cpu_to_le16(rss_data->rss_lut_size);
for (i = 0; i < rss_data->rss_lut_size; i++)
rl->lut[i] = cpu_to_le32(rss_data->rss_lut[i]);
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_SET_RSS_LUT,
- buf_size, (u8 *)rl, 0);
- if (err)
- goto free_mem;
-
- err = idpf_wait_for_event(adapter, vport, IDPF_VC_SET_RSS_LUT,
- IDPF_VC_SET_RSS_LUT_ERR);
-
- goto free_mem;
+ xn_params.vc_op = VIRTCHNL2_OP_SET_RSS_LUT;
}
+ reply_sz = idpf_vc_xn_exec(vport->adapter, xn_params);
+ if (reply_sz < 0)
+ return reply_sz;
+ if (!get)
+ return 0;
+ if (reply_sz < sizeof(struct virtchnl2_rss_lut))
+ return -EIO;
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_GET_RSS_LUT,
- buf_size, (u8 *)rl, 0);
- if (err)
- goto free_mem;
-
- err = idpf_wait_for_event(adapter, vport, IDPF_VC_GET_RSS_LUT,
- IDPF_VC_GET_RSS_LUT_ERR);
- if (err)
- goto free_mem;
+ lut_buf_size = le16_to_cpu(recv_rl->lut_entries) * sizeof(u32);
+ if (reply_sz < lut_buf_size)
+ return -EIO;
- recv_rl = (struct virtchnl2_rss_lut *)vport->vc_msg;
+ /* size didn't change, we can reuse existing lut buf */
if (rss_data->rss_lut_size == le16_to_cpu(recv_rl->lut_entries))
goto do_memcpy;
rss_data->rss_lut_size = le16_to_cpu(recv_rl->lut_entries);
kfree(rss_data->rss_lut);
- lut_buf_size = rss_data->rss_lut_size * sizeof(u32);
rss_data->rss_lut = kzalloc(lut_buf_size, GFP_KERNEL);
if (!rss_data->rss_lut) {
rss_data->rss_lut_size = 0;
- err = -ENOMEM;
- goto free_mem;
+ return -ENOMEM;
}
do_memcpy:
- memcpy(rss_data->rss_lut, vport->vc_msg, rss_data->rss_lut_size);
-free_mem:
- mutex_unlock(&vport->vc_buf_lock);
- kfree(rl);
+ memcpy(rss_data->rss_lut, recv_rl->lut, rss_data->rss_lut_size);
- return err;
+ return 0;
}
/**
@@ -2742,68 +2480,70 @@ int idpf_send_get_set_rss_lut_msg(struct idpf_vport *vport, bool get)
*/
int idpf_send_get_set_rss_key_msg(struct idpf_vport *vport, bool get)
{
- struct idpf_adapter *adapter = vport->adapter;
- struct virtchnl2_rss_key *recv_rk;
+ struct virtchnl2_rss_key *recv_rk __free(kfree) = NULL;
+ struct virtchnl2_rss_key *rk __free(kfree) = NULL;
+ struct idpf_vc_xn_params xn_params = {};
struct idpf_rss_data *rss_data;
- struct virtchnl2_rss_key *rk;
- int i, buf_size, err;
+ ssize_t reply_sz;
+ int i, buf_size;
+ u16 key_size;
- rss_data = &adapter->vport_config[vport->idx]->user_config.rss_data;
+ rss_data =
+ &vport->adapter->vport_config[vport->idx]->user_config.rss_data;
buf_size = struct_size(rk, key_flex, rss_data->rss_key_size);
rk = kzalloc(buf_size, GFP_KERNEL);
if (!rk)
return -ENOMEM;
rk->vport_id = cpu_to_le32(vport->vport_id);
- mutex_lock(&vport->vc_buf_lock);
-
+ xn_params.send_buf.iov_base = rk;
+ xn_params.send_buf.iov_len = buf_size;
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
if (get) {
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_GET_RSS_KEY,
- buf_size, (u8 *)rk, 0);
- if (err)
- goto error;
+ recv_rk = kzalloc(IDPF_CTLQ_MAX_BUF_LEN, GFP_KERNEL);
+ if (!recv_rk)
+ return -ENOMEM;
- err = idpf_wait_for_event(adapter, vport, IDPF_VC_GET_RSS_KEY,
- IDPF_VC_GET_RSS_KEY_ERR);
- if (err)
- goto error;
-
- recv_rk = (struct virtchnl2_rss_key *)vport->vc_msg;
- if (rss_data->rss_key_size !=
- le16_to_cpu(recv_rk->key_len)) {
- rss_data->rss_key_size =
- min_t(u16, NETDEV_RSS_KEY_LEN,
- le16_to_cpu(recv_rk->key_len));
- kfree(rss_data->rss_key);
- rss_data->rss_key = kzalloc(rss_data->rss_key_size,
- GFP_KERNEL);
- if (!rss_data->rss_key) {
- rss_data->rss_key_size = 0;
- err = -ENOMEM;
- goto error;
- }
- }
- memcpy(rss_data->rss_key, recv_rk->key_flex,
- rss_data->rss_key_size);
+ xn_params.vc_op = VIRTCHNL2_OP_GET_RSS_KEY;
+ xn_params.recv_buf.iov_base = recv_rk;
+ xn_params.recv_buf.iov_len = IDPF_CTLQ_MAX_BUF_LEN;
} else {
rk->key_len = cpu_to_le16(rss_data->rss_key_size);
for (i = 0; i < rss_data->rss_key_size; i++)
rk->key_flex[i] = rss_data->rss_key[i];
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_SET_RSS_KEY,
- buf_size, (u8 *)rk, 0);
- if (err)
- goto error;
+ xn_params.vc_op = VIRTCHNL2_OP_SET_RSS_KEY;
+ }
+
+ reply_sz = idpf_vc_xn_exec(vport->adapter, xn_params);
+ if (reply_sz < 0)
+ return reply_sz;
+ if (!get)
+ return 0;
+ if (reply_sz < sizeof(struct virtchnl2_rss_key))
+ return -EIO;
- err = idpf_wait_for_event(adapter, vport, IDPF_VC_SET_RSS_KEY,
- IDPF_VC_SET_RSS_KEY_ERR);
+ key_size = min_t(u16, NETDEV_RSS_KEY_LEN,
+ le16_to_cpu(recv_rk->key_len));
+ if (reply_sz < key_size)
+ return -EIO;
+
+ /* key len didn't change, reuse existing buf */
+ if (rss_data->rss_key_size == key_size)
+ goto do_memcpy;
+
+ rss_data->rss_key_size = key_size;
+ kfree(rss_data->rss_key);
+ rss_data->rss_key = kzalloc(key_size, GFP_KERNEL);
+ if (!rss_data->rss_key) {
+ rss_data->rss_key_size = 0;
+ return -ENOMEM;
}
-error:
- mutex_unlock(&vport->vc_buf_lock);
- kfree(rk);
+do_memcpy:
+ memcpy(rss_data->rss_key, recv_rk->key_flex, rss_data->rss_key_size);
- return err;
+ return 0;
}
/**
@@ -2855,13 +2595,15 @@ static void idpf_fill_ptype_lookup(struct idpf_rx_ptype_decoded *ptype,
*/
int idpf_send_get_rx_ptype_msg(struct idpf_vport *vport)
{
+ struct virtchnl2_get_ptype_info *get_ptype_info __free(kfree) = NULL;
+ struct virtchnl2_get_ptype_info *ptype_info __free(kfree) = NULL;
struct idpf_rx_ptype_decoded *ptype_lkup = vport->rx_ptype_lkup;
- struct virtchnl2_get_ptype_info get_ptype_info;
int max_ptype, ptypes_recvd = 0, ptype_offset;
struct idpf_adapter *adapter = vport->adapter;
- struct virtchnl2_get_ptype_info *ptype_info;
+ struct idpf_vc_xn_params xn_params = {};
u16 next_ptype_id = 0;
- int err = 0, i, j, k;
+ ssize_t reply_sz;
+ int i, j, k;
if (idpf_is_queue_model_split(vport->rxq_model))
max_ptype = IDPF_RX_MAX_PTYPE;
@@ -2870,43 +2612,44 @@ int idpf_send_get_rx_ptype_msg(struct idpf_vport *vport)
memset(vport->rx_ptype_lkup, 0, sizeof(vport->rx_ptype_lkup));
+ get_ptype_info = kzalloc(sizeof(*get_ptype_info), GFP_KERNEL);
+ if (!get_ptype_info)
+ return -ENOMEM;
+
ptype_info = kzalloc(IDPF_CTLQ_MAX_BUF_LEN, GFP_KERNEL);
if (!ptype_info)
return -ENOMEM;
- mutex_lock(&adapter->vc_buf_lock);
+ xn_params.vc_op = VIRTCHNL2_OP_GET_PTYPE_INFO;
+ xn_params.send_buf.iov_base = get_ptype_info;
+ xn_params.send_buf.iov_len = sizeof(*get_ptype_info);
+ xn_params.recv_buf.iov_base = ptype_info;
+ xn_params.recv_buf.iov_len = IDPF_CTLQ_MAX_BUF_LEN;
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
while (next_ptype_id < max_ptype) {
- get_ptype_info.start_ptype_id = cpu_to_le16(next_ptype_id);
+ get_ptype_info->start_ptype_id = cpu_to_le16(next_ptype_id);
if ((next_ptype_id + IDPF_RX_MAX_PTYPES_PER_BUF) > max_ptype)
- get_ptype_info.num_ptypes =
+ get_ptype_info->num_ptypes =
cpu_to_le16(max_ptype - next_ptype_id);
else
- get_ptype_info.num_ptypes =
+ get_ptype_info->num_ptypes =
cpu_to_le16(IDPF_RX_MAX_PTYPES_PER_BUF);
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_GET_PTYPE_INFO,
- sizeof(struct virtchnl2_get_ptype_info),
- (u8 *)&get_ptype_info, 0);
- if (err)
- goto vc_buf_unlock;
-
- err = idpf_wait_for_event(adapter, NULL, IDPF_VC_GET_PTYPE_INFO,
- IDPF_VC_GET_PTYPE_INFO_ERR);
- if (err)
- goto vc_buf_unlock;
+ reply_sz = idpf_vc_xn_exec(adapter, xn_params);
+ if (reply_sz < 0)
+ return reply_sz;
- memcpy(ptype_info, adapter->vc_msg, IDPF_CTLQ_MAX_BUF_LEN);
+ if (reply_sz < IDPF_CTLQ_MAX_BUF_LEN)
+ return -EIO;
ptypes_recvd += le16_to_cpu(ptype_info->num_ptypes);
- if (ptypes_recvd > max_ptype) {
- err = -EINVAL;
- goto vc_buf_unlock;
- }
+ if (ptypes_recvd > max_ptype)
+ return -EINVAL;
- next_ptype_id = le16_to_cpu(get_ptype_info.start_ptype_id) +
- le16_to_cpu(get_ptype_info.num_ptypes);
+ next_ptype_id = le16_to_cpu(get_ptype_info->start_ptype_id) +
+ le16_to_cpu(get_ptype_info->num_ptypes);
ptype_offset = IDPF_RX_PTYPE_HDR_SZ;
@@ -2919,17 +2662,13 @@ int idpf_send_get_rx_ptype_msg(struct idpf_vport *vport)
((u8 *)ptype_info + ptype_offset);
ptype_offset += IDPF_GET_PTYPE_SIZE(ptype);
- if (ptype_offset > IDPF_CTLQ_MAX_BUF_LEN) {
- err = -EINVAL;
- goto vc_buf_unlock;
- }
+ if (ptype_offset > IDPF_CTLQ_MAX_BUF_LEN)
+ return -EINVAL;
/* 0xFFFF indicates end of ptypes */
if (le16_to_cpu(ptype->ptype_id_10) ==
- IDPF_INVALID_PTYPE_ID) {
- err = 0;
- goto vc_buf_unlock;
- }
+ IDPF_INVALID_PTYPE_ID)
+ return 0;
if (idpf_is_queue_model_split(vport->rxq_model))
k = le16_to_cpu(ptype->ptype_id_10);
@@ -3057,11 +2796,7 @@ int idpf_send_get_rx_ptype_msg(struct idpf_vport *vport)
}
}
-vc_buf_unlock:
- mutex_unlock(&adapter->vc_buf_lock);
- kfree(ptype_info);
-
- return err;
+ return 0;
}
/**
@@ -3073,27 +2808,20 @@ int idpf_send_get_rx_ptype_msg(struct idpf_vport *vport)
*/
int idpf_send_ena_dis_loopback_msg(struct idpf_vport *vport)
{
+ struct idpf_vc_xn_params xn_params = {};
struct virtchnl2_loopback loopback;
- int err;
+ ssize_t reply_sz;
loopback.vport_id = cpu_to_le32(vport->vport_id);
loopback.enable = idpf_is_feature_ena(vport, NETIF_F_LOOPBACK);
- mutex_lock(&vport->vc_buf_lock);
-
- err = idpf_send_mb_msg(vport->adapter, VIRTCHNL2_OP_LOOPBACK,
- sizeof(loopback), (u8 *)&loopback, 0);
- if (err)
- goto rel_lock;
-
- err = idpf_wait_for_event(vport->adapter, vport,
- IDPF_VC_LOOPBACK_STATE,
- IDPF_VC_LOOPBACK_STATE_ERR);
-
-rel_lock:
- mutex_unlock(&vport->vc_buf_lock);
+ xn_params.vc_op = VIRTCHNL2_OP_LOOPBACK;
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
+ xn_params.send_buf.iov_base = &loopback;
+ xn_params.send_buf.iov_len = sizeof(loopback);
+ reply_sz = idpf_vc_xn_exec(vport->adapter, xn_params);
- return err;
+ return reply_sz < 0 ? reply_sz : 0;
}
/**
@@ -3836,17 +3564,20 @@ int idpf_add_del_mac_filters(struct idpf_vport *vport,
struct idpf_netdev_priv *np,
bool add, bool async)
{
- struct virtchnl2_mac_addr_list *ma_list = NULL;
+ struct virtchnl2_mac_addr_list *ma_list __free(kfree) = NULL;
+ struct virtchnl2_mac_addr *mac_addr __free(kfree) = NULL;
struct idpf_adapter *adapter = np->adapter;
+ struct idpf_vc_xn_params xn_params = {};
struct idpf_vport_config *vport_config;
- enum idpf_vport_config_flags mac_flag;
- struct pci_dev *pdev = adapter->pdev;
- enum idpf_vport_vc_state vc, vc_err;
- struct virtchnl2_mac_addr *mac_addr;
- struct idpf_mac_filter *f, *tmp;
u32 num_msgs, total_filters = 0;
- int i = 0, k, err = 0;
- u32 vop;
+ struct idpf_mac_filter *f;
+ ssize_t reply_sz;
+ int i = 0, k;
+
+ xn_params.vc_op = add ? VIRTCHNL2_OP_ADD_MAC_ADDR :
+ VIRTCHNL2_OP_DEL_MAC_ADDR;
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
+ xn_params.async = async;
vport_config = adapter->vport_config[np->vport_idx];
spin_lock_bh(&vport_config->mac_filter_list_lock);
@@ -3870,13 +3601,13 @@ int idpf_add_del_mac_filters(struct idpf_vport *vport,
mac_addr = kcalloc(total_filters, sizeof(struct virtchnl2_mac_addr),
GFP_ATOMIC);
if (!mac_addr) {
- err = -ENOMEM;
spin_unlock_bh(&vport_config->mac_filter_list_lock);
- goto error;
+
+ return -ENOMEM;
}
- list_for_each_entry_safe(f, tmp, &vport_config->user_config.mac_filter_list,
- list) {
+ list_for_each_entry(f, &vport_config->user_config.mac_filter_list,
+ list) {
if (add && f->add) {
ether_addr_copy(mac_addr[i].addr, f->macaddr);
i++;
@@ -3895,26 +3626,11 @@ int idpf_add_del_mac_filters(struct idpf_vport *vport,
spin_unlock_bh(&vport_config->mac_filter_list_lock);
- if (add) {
- vop = VIRTCHNL2_OP_ADD_MAC_ADDR;
- vc = IDPF_VC_ADD_MAC_ADDR;
- vc_err = IDPF_VC_ADD_MAC_ADDR_ERR;
- mac_flag = IDPF_VPORT_ADD_MAC_REQ;
- } else {
- vop = VIRTCHNL2_OP_DEL_MAC_ADDR;
- vc = IDPF_VC_DEL_MAC_ADDR;
- vc_err = IDPF_VC_DEL_MAC_ADDR_ERR;
- mac_flag = IDPF_VPORT_DEL_MAC_REQ;
- }
-
/* Chunk up the filters into multiple messages to avoid
* sending a control queue message buffer that is too large
*/
num_msgs = DIV_ROUND_UP(total_filters, IDPF_NUM_FILTERS_PER_MSG);
- if (!async)
- mutex_lock(&vport->vc_buf_lock);
-
for (i = 0, k = 0; i < num_msgs; i++) {
u32 entries_size, buf_size, num_entries;
@@ -3926,10 +3642,8 @@ int idpf_add_del_mac_filters(struct idpf_vport *vport,
if (!ma_list || num_entries != IDPF_NUM_FILTERS_PER_MSG) {
kfree(ma_list);
ma_list = kzalloc(buf_size, GFP_ATOMIC);
- if (!ma_list) {
- err = -ENOMEM;
- goto list_prep_error;
- }
+ if (!ma_list)
+ return -ENOMEM;
} else {
memset(ma_list, 0, buf_size);
}
@@ -3938,34 +3652,17 @@ int idpf_add_del_mac_filters(struct idpf_vport *vport,
ma_list->num_mac_addr = cpu_to_le16(num_entries);
memcpy(ma_list->mac_addr_list, &mac_addr[k], entries_size);
- if (async)
- set_bit(mac_flag, vport_config->flags);
-
- err = idpf_send_mb_msg(adapter, vop, buf_size, (u8 *)ma_list, 0);
- if (err)
- goto mbx_error;
-
- if (!async) {
- err = idpf_wait_for_event(adapter, vport, vc, vc_err);
- if (err)
- goto mbx_error;
- }
+ xn_params.send_buf.iov_base = ma_list;
+ xn_params.send_buf.iov_len = buf_size;
+ reply_sz = idpf_vc_xn_exec(adapter, xn_params);
+ if (reply_sz < 0)
+ return reply_sz;
k += num_entries;
total_filters -= num_entries;
}
-mbx_error:
- if (!async)
- mutex_unlock(&vport->vc_buf_lock);
- kfree(ma_list);
-list_prep_error:
- kfree(mac_addr);
-error:
- if (err)
- dev_err(&pdev->dev, "Failed to add or del mac filters %d", err);
-
- return err;
+ return 0;
}
/**
@@ -3982,9 +3679,10 @@ int idpf_set_promiscuous(struct idpf_adapter *adapter,
struct idpf_vport_user_config_data *config_data,
u32 vport_id)
{
+ struct idpf_vc_xn_params xn_params = {};
struct virtchnl2_promisc_info vpi;
+ ssize_t reply_sz;
u16 flags = 0;
- int err;
if (test_bit(__IDPF_PROMISC_UC, config_data->user_flags))
flags |= VIRTCHNL2_UNICAST_PROMISC;
@@ -3994,9 +3692,13 @@ int idpf_set_promiscuous(struct idpf_adapter *adapter,
vpi.vport_id = cpu_to_le32(vport_id);
vpi.flags = cpu_to_le16(flags);
- err = idpf_send_mb_msg(adapter, VIRTCHNL2_OP_CONFIG_PROMISCUOUS_MODE,
- sizeof(struct virtchnl2_promisc_info),
- (u8 *)&vpi, 0);
+ xn_params.vc_op = VIRTCHNL2_OP_CONFIG_PROMISCUOUS_MODE;
+ xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
+ xn_params.send_buf.iov_base = &vpi;
+ xn_params.send_buf.iov_len = sizeof(vpi);
+ /* setting promiscuous is only ever done asynchronously */
+ xn_params.async = true;
+ reply_sz = idpf_vc_xn_exec(adapter, xn_params);
- return err;
+ return reply_sz < 0 ? reply_sz : 0;
}
--
2.40.1
^ permalink raw reply related [flat|nested] 44+ messages in thread
* [PATCH v4 05/10 iwl-next] idpf: add async_handler for MAC filter messages
2024-02-06 3:37 ` [Intel-wired-lan] " Alan Brady
@ 2024-02-06 3:37 ` Alan Brady
-1 siblings, 0 replies; 44+ messages in thread
From: Alan Brady @ 2024-02-06 3:37 UTC (permalink / raw)
To: intel-wired-lan
Cc: netdev, willemdebruijn.kernel, przemyslaw.kitszel, igor.bagnucki,
aleksander.lobakin, Alan Brady
There are situations where the driver needs to add a MAC filter but
we're explicitly not allowed to sleep so we can wait for a virtchnl
message to complete.
This adds an async_handler for asynchronously sent messages for MAC
filters so that we can better handle if there's an error of some kind.
If success we don't need to do anything else, but if we failed to
program the new filter we really should remove it from our list of MAC
filters. If we don't remove bad filters, what I expect to happen is
after a reset of some kind we try to program the MAC filter again and it
fails again. This is clearly wrong and I would expect to be confusing
for the user.
It could also be the failure is for a delete MAC filter message but
those filters get deleted regardless. Not much we can do about a delete
failure.
Signed-off-by: Alan Brady <alan.brady@intel.com>
---
.../net/ethernet/intel/idpf/idpf_virtchnl.c | 70 +++++++++++++++++++
1 file changed, 70 insertions(+)
diff --git a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
index 6d95daf0637b..8756e151263d 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
@@ -3551,6 +3551,75 @@ u32 idpf_get_vport_id(struct idpf_vport *vport)
return le32_to_cpu(vport_msg->vport_id);
}
+/**
+ * idpf_mac_filter_async_handler - Async callback for mac filters
+ * @adapter: private data struct
+ * @xn: transaction for message
+ * @ctlq_msg: received message
+ *
+ * In some scenarios driver can't sleep and wait for a reply (e.g.: stack is
+ * holding rtnl_lock) when adding a new mac filter. It puts us in a difficult
+ * situation to deal with errors returned on the reply. The best we can
+ * ultimately do is remove it from our list of mac filters and report the
+ * error.
+ */
+static int idpf_mac_filter_async_handler(struct idpf_adapter *adapter,
+ struct idpf_vc_xn *xn,
+ const struct idpf_ctlq_msg *ctlq_msg)
+{
+ struct virtchnl2_mac_addr_list *ma_list;
+ struct idpf_vport_config *vport_config;
+ struct virtchnl2_mac_addr *mac_addr;
+ struct idpf_mac_filter *f, *tmp;
+ struct list_head *ma_list_head;
+ struct idpf_vport *vport;
+ u16 num_entries;
+ int i;
+
+ /* if success we're done, we're only here if something bad happened */
+ if (!ctlq_msg->cookie.mbx.chnl_retval)
+ return 0;
+
+ /* make sure at least struct is there */
+ if (xn->reply_sz < sizeof(*ma_list))
+ goto invalid_payload;
+
+ ma_list = ctlq_msg->ctx.indirect.payload->va;
+ mac_addr = ma_list->mac_addr_list;
+ num_entries = le16_to_cpu(ma_list->num_mac_addr);
+ /* we should have received a buffer at least this big */
+ if (xn->reply_sz < struct_size(ma_list, mac_addr_list, num_entries))
+ goto invalid_payload;
+
+ vport = idpf_vid_to_vport(adapter, le32_to_cpu(ma_list->vport_id));
+ if (!vport)
+ goto invalid_payload;
+
+ vport_config = adapter->vport_config[le32_to_cpu(ma_list->vport_id)];
+ ma_list_head = &vport_config->user_config.mac_filter_list;
+
+ /* We can't do much to reconcile bad filters at this point, however we
+ * should at least remove them from our list one way or the other so we
+ * have some idea what good filters we have.
+ */
+ spin_lock_bh(&vport_config->mac_filter_list_lock);
+ list_for_each_entry_safe(f, tmp, ma_list_head, list)
+ for (i = 0; i < num_entries; i++)
+ if (ether_addr_equal(mac_addr[i].addr, f->macaddr))
+ list_del(&f->list);
+ spin_unlock_bh(&vport_config->mac_filter_list_lock);
+ dev_err_ratelimited(&adapter->pdev->dev, "Received error sending MAC filter request (op %d)\n",
+ xn->vc_op);
+
+ return 0;
+
+invalid_payload:
+ dev_err_ratelimited(&adapter->pdev->dev, "Received invalid MAC filter payload (op %d) (len %zd)\n",
+ xn->vc_op, xn->reply_sz);
+
+ return -EINVAL;
+}
+
/**
* idpf_add_del_mac_filters - Add/del mac filters
* @vport: Virtual port data structure
@@ -3578,6 +3647,7 @@ int idpf_add_del_mac_filters(struct idpf_vport *vport,
VIRTCHNL2_OP_DEL_MAC_ADDR;
xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
xn_params.async = async;
+ xn_params.async_handler = idpf_mac_filter_async_handler;
vport_config = adapter->vport_config[np->vport_idx];
spin_lock_bh(&vport_config->mac_filter_list_lock);
--
2.40.1
^ permalink raw reply related [flat|nested] 44+ messages in thread
* [Intel-wired-lan] [PATCH v4 05/10 iwl-next] idpf: add async_handler for MAC filter messages
@ 2024-02-06 3:37 ` Alan Brady
0 siblings, 0 replies; 44+ messages in thread
From: Alan Brady @ 2024-02-06 3:37 UTC (permalink / raw)
To: intel-wired-lan
Cc: willemdebruijn.kernel, netdev, aleksander.lobakin, Alan Brady,
przemyslaw.kitszel, igor.bagnucki
There are situations where the driver needs to add a MAC filter but
we're explicitly not allowed to sleep so we can wait for a virtchnl
message to complete.
This adds an async_handler for asynchronously sent messages for MAC
filters so that we can better handle if there's an error of some kind.
If success we don't need to do anything else, but if we failed to
program the new filter we really should remove it from our list of MAC
filters. If we don't remove bad filters, what I expect to happen is
after a reset of some kind we try to program the MAC filter again and it
fails again. This is clearly wrong and I would expect to be confusing
for the user.
It could also be the failure is for a delete MAC filter message but
those filters get deleted regardless. Not much we can do about a delete
failure.
Signed-off-by: Alan Brady <alan.brady@intel.com>
---
.../net/ethernet/intel/idpf/idpf_virtchnl.c | 70 +++++++++++++++++++
1 file changed, 70 insertions(+)
diff --git a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
index 6d95daf0637b..8756e151263d 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
@@ -3551,6 +3551,75 @@ u32 idpf_get_vport_id(struct idpf_vport *vport)
return le32_to_cpu(vport_msg->vport_id);
}
+/**
+ * idpf_mac_filter_async_handler - Async callback for mac filters
+ * @adapter: private data struct
+ * @xn: transaction for message
+ * @ctlq_msg: received message
+ *
+ * In some scenarios driver can't sleep and wait for a reply (e.g.: stack is
+ * holding rtnl_lock) when adding a new mac filter. It puts us in a difficult
+ * situation to deal with errors returned on the reply. The best we can
+ * ultimately do is remove it from our list of mac filters and report the
+ * error.
+ */
+static int idpf_mac_filter_async_handler(struct idpf_adapter *adapter,
+ struct idpf_vc_xn *xn,
+ const struct idpf_ctlq_msg *ctlq_msg)
+{
+ struct virtchnl2_mac_addr_list *ma_list;
+ struct idpf_vport_config *vport_config;
+ struct virtchnl2_mac_addr *mac_addr;
+ struct idpf_mac_filter *f, *tmp;
+ struct list_head *ma_list_head;
+ struct idpf_vport *vport;
+ u16 num_entries;
+ int i;
+
+ /* if success we're done, we're only here if something bad happened */
+ if (!ctlq_msg->cookie.mbx.chnl_retval)
+ return 0;
+
+ /* make sure at least struct is there */
+ if (xn->reply_sz < sizeof(*ma_list))
+ goto invalid_payload;
+
+ ma_list = ctlq_msg->ctx.indirect.payload->va;
+ mac_addr = ma_list->mac_addr_list;
+ num_entries = le16_to_cpu(ma_list->num_mac_addr);
+ /* we should have received a buffer at least this big */
+ if (xn->reply_sz < struct_size(ma_list, mac_addr_list, num_entries))
+ goto invalid_payload;
+
+ vport = idpf_vid_to_vport(adapter, le32_to_cpu(ma_list->vport_id));
+ if (!vport)
+ goto invalid_payload;
+
+ vport_config = adapter->vport_config[le32_to_cpu(ma_list->vport_id)];
+ ma_list_head = &vport_config->user_config.mac_filter_list;
+
+ /* We can't do much to reconcile bad filters at this point, however we
+ * should at least remove them from our list one way or the other so we
+ * have some idea what good filters we have.
+ */
+ spin_lock_bh(&vport_config->mac_filter_list_lock);
+ list_for_each_entry_safe(f, tmp, ma_list_head, list)
+ for (i = 0; i < num_entries; i++)
+ if (ether_addr_equal(mac_addr[i].addr, f->macaddr))
+ list_del(&f->list);
+ spin_unlock_bh(&vport_config->mac_filter_list_lock);
+ dev_err_ratelimited(&adapter->pdev->dev, "Received error sending MAC filter request (op %d)\n",
+ xn->vc_op);
+
+ return 0;
+
+invalid_payload:
+ dev_err_ratelimited(&adapter->pdev->dev, "Received invalid MAC filter payload (op %d) (len %zd)\n",
+ xn->vc_op, xn->reply_sz);
+
+ return -EINVAL;
+}
+
/**
* idpf_add_del_mac_filters - Add/del mac filters
* @vport: Virtual port data structure
@@ -3578,6 +3647,7 @@ int idpf_add_del_mac_filters(struct idpf_vport *vport,
VIRTCHNL2_OP_DEL_MAC_ADDR;
xn_params.timeout_ms = IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC;
xn_params.async = async;
+ xn_params.async_handler = idpf_mac_filter_async_handler;
vport_config = adapter->vport_config[np->vport_idx];
spin_lock_bh(&vport_config->mac_filter_list_lock);
--
2.40.1
^ permalink raw reply related [flat|nested] 44+ messages in thread
* [PATCH v4 06/10 iwl-next] idpf: refactor idpf_recv_mb_msg
2024-02-06 3:37 ` [Intel-wired-lan] " Alan Brady
@ 2024-02-06 3:38 ` Alan Brady
-1 siblings, 0 replies; 44+ messages in thread
From: Alan Brady @ 2024-02-06 3:38 UTC (permalink / raw)
To: intel-wired-lan
Cc: netdev, willemdebruijn.kernel, przemyslaw.kitszel, igor.bagnucki,
aleksander.lobakin, Alan Brady
Now that all the messages are using the transaction API, we can rework
idpf_recv_mb_msg quite a lot to simplify it. Due to this, we remove
idpf_find_vport as no longer used and alter idpf_recv_event_msg
slightly.
Signed-off-by: Alan Brady <alan.brady@intel.com>
---
drivers/net/ethernet/intel/idpf/idpf.h | 3 +-
drivers/net/ethernet/intel/idpf/idpf_lib.c | 2 +-
.../net/ethernet/intel/idpf/idpf_virtchnl.c | 254 +++---------------
3 files changed, 37 insertions(+), 222 deletions(-)
diff --git a/drivers/net/ethernet/intel/idpf/idpf.h b/drivers/net/ethernet/intel/idpf/idpf.h
index 393f6e46012f..2d5449b9288a 100644
--- a/drivers/net/ethernet/intel/idpf/idpf.h
+++ b/drivers/net/ethernet/intel/idpf/idpf.h
@@ -1031,8 +1031,7 @@ int idpf_send_get_stats_msg(struct idpf_vport *vport);
int idpf_get_vec_ids(struct idpf_adapter *adapter,
u16 *vecids, int num_vecids,
struct virtchnl2_vector_chunks *chunks);
-int idpf_recv_mb_msg(struct idpf_adapter *adapter, u32 op,
- void *msg, int msg_size);
+int idpf_recv_mb_msg(struct idpf_adapter *adapter);
int idpf_send_mb_msg(struct idpf_adapter *adapter, u32 op,
u16 msg_size, u8 *msg, u16 cookie);
void idpf_set_ethtool_ops(struct net_device *netdev);
diff --git a/drivers/net/ethernet/intel/idpf/idpf_lib.c b/drivers/net/ethernet/intel/idpf/idpf_lib.c
index 6b4a7408246c..8135a6b4a71e 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_lib.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_lib.c
@@ -1253,7 +1253,7 @@ void idpf_mbx_task(struct work_struct *work)
queue_delayed_work(adapter->mbx_wq, &adapter->mbx_task,
msecs_to_jiffies(300));
- idpf_recv_mb_msg(adapter, VIRTCHNL2_OP_UNKNOWN, NULL, 0);
+ idpf_recv_mb_msg(adapter);
}
/**
diff --git a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
index 8756e151263d..5a68a3908cf0 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
@@ -63,12 +63,12 @@ static void idpf_handle_event_link(struct idpf_adapter *adapter,
/**
* idpf_recv_event_msg - Receive virtchnl event message
- * @vport: virtual port structure
+ * @adapter: Driver specific private structure
* @ctlq_msg: message to copy from
*
* Receive virtchnl event message
*/
-static void idpf_recv_event_msg(struct idpf_vport *vport,
+static void idpf_recv_event_msg(struct idpf_adapter *adapter,
struct idpf_ctlq_msg *ctlq_msg)
{
int payload_size = ctlq_msg->ctx.indirect.payload->size;
@@ -76,7 +76,7 @@ static void idpf_recv_event_msg(struct idpf_vport *vport,
u32 event;
if (payload_size < sizeof(*v2e)) {
- dev_err_ratelimited(&vport->adapter->pdev->dev, "Failed to receive valid payload for event msg (op %d len %d)\n",
+ dev_err_ratelimited(&adapter->pdev->dev, "Failed to receive valid payload for event msg (op %d len %d)\n",
ctlq_msg->cookie.mbx.chnl_opcode,
payload_size);
return;
@@ -87,10 +87,10 @@ static void idpf_recv_event_msg(struct idpf_vport *vport,
switch (event) {
case VIRTCHNL2_EVENT_LINK_CHANGE:
- idpf_handle_event_link(vport->adapter, v2e);
+ idpf_handle_event_link(adapter, v2e);
return;
default:
- dev_err(&vport->adapter->pdev->dev,
+ dev_err(&adapter->pdev->dev,
"Unknown event %d from PF\n", event);
break;
}
@@ -213,125 +213,6 @@ int idpf_send_mb_msg(struct idpf_adapter *adapter, u32 op,
return err;
}
-/**
- * idpf_find_vport - Find vport pointer from control queue message
- * @adapter: driver specific private structure
- * @vport: address of vport pointer to copy the vport from adapters vport list
- * @ctlq_msg: control queue message
- *
- * Return 0 on success, error value on failure. Also this function does check
- * for the opcodes which expect to receive payload and return error value if
- * it is not the case.
- */
-static int idpf_find_vport(struct idpf_adapter *adapter,
- struct idpf_vport **vport,
- struct idpf_ctlq_msg *ctlq_msg)
-{
- bool no_op = false, vid_found = false;
- int i, err = 0;
- char *vc_msg;
- u32 v_id;
-
- vc_msg = kcalloc(IDPF_CTLQ_MAX_BUF_LEN, sizeof(char), GFP_KERNEL);
- if (!vc_msg)
- return -ENOMEM;
-
- if (ctlq_msg->data_len) {
- size_t payload_size = ctlq_msg->ctx.indirect.payload->size;
-
- if (!payload_size) {
- dev_err(&adapter->pdev->dev, "Failed to receive payload buffer\n");
- kfree(vc_msg);
-
- return -EINVAL;
- }
-
- memcpy(vc_msg, ctlq_msg->ctx.indirect.payload->va,
- min_t(size_t, payload_size, IDPF_CTLQ_MAX_BUF_LEN));
- }
-
- switch (ctlq_msg->cookie.mbx.chnl_opcode) {
- case VIRTCHNL2_OP_VERSION:
- case VIRTCHNL2_OP_GET_CAPS:
- case VIRTCHNL2_OP_CREATE_VPORT:
- case VIRTCHNL2_OP_SET_SRIOV_VFS:
- case VIRTCHNL2_OP_ALLOC_VECTORS:
- case VIRTCHNL2_OP_DEALLOC_VECTORS:
- case VIRTCHNL2_OP_GET_PTYPE_INFO:
- goto free_vc_msg;
- case VIRTCHNL2_OP_ENABLE_VPORT:
- case VIRTCHNL2_OP_DISABLE_VPORT:
- case VIRTCHNL2_OP_DESTROY_VPORT:
- v_id = le32_to_cpu(((struct virtchnl2_vport *)vc_msg)->vport_id);
- break;
- case VIRTCHNL2_OP_CONFIG_TX_QUEUES:
- v_id = le32_to_cpu(((struct virtchnl2_config_tx_queues *)vc_msg)->vport_id);
- break;
- case VIRTCHNL2_OP_CONFIG_RX_QUEUES:
- v_id = le32_to_cpu(((struct virtchnl2_config_rx_queues *)vc_msg)->vport_id);
- break;
- case VIRTCHNL2_OP_ENABLE_QUEUES:
- case VIRTCHNL2_OP_DISABLE_QUEUES:
- case VIRTCHNL2_OP_DEL_QUEUES:
- v_id = le32_to_cpu(((struct virtchnl2_del_ena_dis_queues *)vc_msg)->vport_id);
- break;
- case VIRTCHNL2_OP_ADD_QUEUES:
- v_id = le32_to_cpu(((struct virtchnl2_add_queues *)vc_msg)->vport_id);
- break;
- case VIRTCHNL2_OP_MAP_QUEUE_VECTOR:
- case VIRTCHNL2_OP_UNMAP_QUEUE_VECTOR:
- v_id = le32_to_cpu(((struct virtchnl2_queue_vector_maps *)vc_msg)->vport_id);
- break;
- case VIRTCHNL2_OP_GET_STATS:
- v_id = le32_to_cpu(((struct virtchnl2_vport_stats *)vc_msg)->vport_id);
- break;
- case VIRTCHNL2_OP_GET_RSS_LUT:
- case VIRTCHNL2_OP_SET_RSS_LUT:
- v_id = le32_to_cpu(((struct virtchnl2_rss_lut *)vc_msg)->vport_id);
- break;
- case VIRTCHNL2_OP_GET_RSS_KEY:
- case VIRTCHNL2_OP_SET_RSS_KEY:
- v_id = le32_to_cpu(((struct virtchnl2_rss_key *)vc_msg)->vport_id);
- break;
- case VIRTCHNL2_OP_EVENT:
- v_id = le32_to_cpu(((struct virtchnl2_event *)vc_msg)->vport_id);
- break;
- case VIRTCHNL2_OP_LOOPBACK:
- v_id = le32_to_cpu(((struct virtchnl2_loopback *)vc_msg)->vport_id);
- break;
- case VIRTCHNL2_OP_CONFIG_PROMISCUOUS_MODE:
- v_id = le32_to_cpu(((struct virtchnl2_promisc_info *)vc_msg)->vport_id);
- break;
- case VIRTCHNL2_OP_ADD_MAC_ADDR:
- case VIRTCHNL2_OP_DEL_MAC_ADDR:
- v_id = le32_to_cpu(((struct virtchnl2_mac_addr_list *)vc_msg)->vport_id);
- break;
- default:
- no_op = true;
- break;
- }
-
- if (no_op)
- goto free_vc_msg;
-
- for (i = 0; i < idpf_get_max_vports(adapter); i++) {
- if (adapter->vport_ids[i] == v_id) {
- vid_found = true;
- break;
- }
- }
-
- if (vid_found)
- *vport = adapter->vports[i];
- else
- err = -EINVAL;
-
-free_vc_msg:
- kfree(vc_msg);
-
- return err;
-}
-
/* API for virtchnl "transaction" support ("xn" for short).
*
* We are reusing the completion lock to serialize the accesses to the
@@ -707,118 +588,53 @@ idpf_vc_xn_forward_reply(struct idpf_adapter *adapter,
/**
* idpf_recv_mb_msg - Receive message over mailbox
* @adapter: Driver specific private structure
- * @op: virtchannel operation code
- * @msg: Received message holding buffer
- * @msg_size: message size
*
* Will receive control queue message and posts the receive buffer. Returns 0
* on success and negative on failure.
*/
-int idpf_recv_mb_msg(struct idpf_adapter *adapter, u32 op,
- void *msg, int msg_size)
+int idpf_recv_mb_msg(struct idpf_adapter *adapter)
{
- struct idpf_vport *vport = NULL;
struct idpf_ctlq_msg ctlq_msg;
struct idpf_dma_mem *dma_mem;
- bool work_done = false;
- int num_retry = 2000;
- u16 num_q_msg;
- int err;
+ int post_err, err;
+ u16 num_recv;
while (1) {
- /* Try to get one message */
- num_q_msg = 1;
- dma_mem = NULL;
- err = idpf_ctlq_recv(adapter->hw.arq, &num_q_msg, &ctlq_msg);
- /* If no message then decide if we have to retry based on
- * opcode
- */
- if (err || !num_q_msg) {
- /* Increasing num_retry to consider the delayed
- * responses because of large number of VF's mailbox
- * messages. If the mailbox message is received from
- * the other side, we come out of the sleep cycle
- * immediately else we wait for more time.
- */
- if (!op || !num_retry--)
- break;
- if (test_bit(IDPF_REMOVE_IN_PROG, adapter->flags)) {
- err = -EIO;
- break;
- }
- msleep(20);
- continue;
- }
-
- /* If we are here a message is received. Check if we are looking
- * for a specific message based on opcode. If it is different
- * ignore and post buffers
+ /* This will get <= num_recv messages and output how many
+ * actually received on num_recv.
*/
- if (op && ctlq_msg.cookie.mbx.chnl_opcode != op)
- goto post_buffs;
-
- err = idpf_find_vport(adapter, &vport, &ctlq_msg);
- if (err)
- goto post_buffs;
-
- /* All conditions are met. Either a message requested is
- * received or we received a message to be processed
- */
- switch (ctlq_msg.cookie.mbx.chnl_opcode) {
- case VIRTCHNL2_OP_VERSION:
- case VIRTCHNL2_OP_GET_CAPS:
- case VIRTCHNL2_OP_CREATE_VPORT:
- case VIRTCHNL2_OP_ENABLE_VPORT:
- case VIRTCHNL2_OP_DISABLE_VPORT:
- case VIRTCHNL2_OP_DESTROY_VPORT:
- case VIRTCHNL2_OP_CONFIG_TX_QUEUES:
- case VIRTCHNL2_OP_CONFIG_RX_QUEUES:
- case VIRTCHNL2_OP_ENABLE_QUEUES:
- case VIRTCHNL2_OP_DISABLE_QUEUES:
- case VIRTCHNL2_OP_ADD_QUEUES:
- case VIRTCHNL2_OP_DEL_QUEUES:
- case VIRTCHNL2_OP_MAP_QUEUE_VECTOR:
- case VIRTCHNL2_OP_UNMAP_QUEUE_VECTOR:
- case VIRTCHNL2_OP_GET_STATS:
- case VIRTCHNL2_OP_GET_RSS_LUT:
- case VIRTCHNL2_OP_SET_RSS_LUT:
- case VIRTCHNL2_OP_GET_RSS_KEY:
- case VIRTCHNL2_OP_SET_RSS_KEY:
- case VIRTCHNL2_OP_SET_SRIOV_VFS:
- case VIRTCHNL2_OP_ALLOC_VECTORS:
- case VIRTCHNL2_OP_DEALLOC_VECTORS:
- case VIRTCHNL2_OP_GET_PTYPE_INFO:
- case VIRTCHNL2_OP_LOOPBACK:
- case VIRTCHNL2_OP_CONFIG_PROMISCUOUS_MODE:
- case VIRTCHNL2_OP_ADD_MAC_ADDR:
- case VIRTCHNL2_OP_DEL_MAC_ADDR:
- err = idpf_vc_xn_forward_reply(adapter, &ctlq_msg);
- break;
- case VIRTCHNL2_OP_EVENT:
- idpf_recv_event_msg(vport, &ctlq_msg);
+ num_recv = 1;
+ err = idpf_ctlq_recv(adapter->hw.arq, &num_recv, &ctlq_msg);
+ if (err || !num_recv)
break;
- default:
- dev_warn(&adapter->pdev->dev,
- "Unhandled virtchnl response %d\n",
- ctlq_msg.cookie.mbx.chnl_opcode);
- break;
- }
-post_buffs:
- if (ctlq_msg.data_len)
+ if (ctlq_msg.data_len) {
dma_mem = ctlq_msg.ctx.indirect.payload;
+ } else {
+ dma_mem = NULL;
+ num_recv = 0;
+ }
+
+ if (ctlq_msg.cookie.mbx.chnl_opcode == VIRTCHNL2_OP_EVENT)
+ idpf_recv_event_msg(adapter, &ctlq_msg);
else
- num_q_msg = 0;
+ err = idpf_vc_xn_forward_reply(adapter, &ctlq_msg);
+
+ post_err = idpf_ctlq_post_rx_buffs(&adapter->hw,
+ adapter->hw.arq,
+ &num_recv, &dma_mem);
- err = idpf_ctlq_post_rx_buffs(&adapter->hw, adapter->hw.arq,
- &num_q_msg, &dma_mem);
/* If post failed clear the only buffer we supplied */
- if (err && dma_mem)
- dma_free_coherent(&adapter->pdev->dev, dma_mem->size,
- dma_mem->va, dma_mem->pa);
+ if (post_err) {
+ if (dma_mem)
+ dmam_free_coherent(&adapter->pdev->dev,
+ dma_mem->size, dma_mem->va,
+ dma_mem->pa);
+ break;
+ }
- /* Applies only if we are looking for a specific opcode */
- if (work_done)
+ /* virtchnl trying to shutdown, stop cleaning */
+ if (err == -ENXIO)
break;
}
--
2.40.1
^ permalink raw reply related [flat|nested] 44+ messages in thread
* [Intel-wired-lan] [PATCH v4 06/10 iwl-next] idpf: refactor idpf_recv_mb_msg
@ 2024-02-06 3:38 ` Alan Brady
0 siblings, 0 replies; 44+ messages in thread
From: Alan Brady @ 2024-02-06 3:38 UTC (permalink / raw)
To: intel-wired-lan
Cc: willemdebruijn.kernel, netdev, aleksander.lobakin, Alan Brady,
przemyslaw.kitszel, igor.bagnucki
Now that all the messages are using the transaction API, we can rework
idpf_recv_mb_msg quite a lot to simplify it. Due to this, we remove
idpf_find_vport as no longer used and alter idpf_recv_event_msg
slightly.
Signed-off-by: Alan Brady <alan.brady@intel.com>
---
drivers/net/ethernet/intel/idpf/idpf.h | 3 +-
drivers/net/ethernet/intel/idpf/idpf_lib.c | 2 +-
.../net/ethernet/intel/idpf/idpf_virtchnl.c | 254 +++---------------
3 files changed, 37 insertions(+), 222 deletions(-)
diff --git a/drivers/net/ethernet/intel/idpf/idpf.h b/drivers/net/ethernet/intel/idpf/idpf.h
index 393f6e46012f..2d5449b9288a 100644
--- a/drivers/net/ethernet/intel/idpf/idpf.h
+++ b/drivers/net/ethernet/intel/idpf/idpf.h
@@ -1031,8 +1031,7 @@ int idpf_send_get_stats_msg(struct idpf_vport *vport);
int idpf_get_vec_ids(struct idpf_adapter *adapter,
u16 *vecids, int num_vecids,
struct virtchnl2_vector_chunks *chunks);
-int idpf_recv_mb_msg(struct idpf_adapter *adapter, u32 op,
- void *msg, int msg_size);
+int idpf_recv_mb_msg(struct idpf_adapter *adapter);
int idpf_send_mb_msg(struct idpf_adapter *adapter, u32 op,
u16 msg_size, u8 *msg, u16 cookie);
void idpf_set_ethtool_ops(struct net_device *netdev);
diff --git a/drivers/net/ethernet/intel/idpf/idpf_lib.c b/drivers/net/ethernet/intel/idpf/idpf_lib.c
index 6b4a7408246c..8135a6b4a71e 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_lib.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_lib.c
@@ -1253,7 +1253,7 @@ void idpf_mbx_task(struct work_struct *work)
queue_delayed_work(adapter->mbx_wq, &adapter->mbx_task,
msecs_to_jiffies(300));
- idpf_recv_mb_msg(adapter, VIRTCHNL2_OP_UNKNOWN, NULL, 0);
+ idpf_recv_mb_msg(adapter);
}
/**
diff --git a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
index 8756e151263d..5a68a3908cf0 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
@@ -63,12 +63,12 @@ static void idpf_handle_event_link(struct idpf_adapter *adapter,
/**
* idpf_recv_event_msg - Receive virtchnl event message
- * @vport: virtual port structure
+ * @adapter: Driver specific private structure
* @ctlq_msg: message to copy from
*
* Receive virtchnl event message
*/
-static void idpf_recv_event_msg(struct idpf_vport *vport,
+static void idpf_recv_event_msg(struct idpf_adapter *adapter,
struct idpf_ctlq_msg *ctlq_msg)
{
int payload_size = ctlq_msg->ctx.indirect.payload->size;
@@ -76,7 +76,7 @@ static void idpf_recv_event_msg(struct idpf_vport *vport,
u32 event;
if (payload_size < sizeof(*v2e)) {
- dev_err_ratelimited(&vport->adapter->pdev->dev, "Failed to receive valid payload for event msg (op %d len %d)\n",
+ dev_err_ratelimited(&adapter->pdev->dev, "Failed to receive valid payload for event msg (op %d len %d)\n",
ctlq_msg->cookie.mbx.chnl_opcode,
payload_size);
return;
@@ -87,10 +87,10 @@ static void idpf_recv_event_msg(struct idpf_vport *vport,
switch (event) {
case VIRTCHNL2_EVENT_LINK_CHANGE:
- idpf_handle_event_link(vport->adapter, v2e);
+ idpf_handle_event_link(adapter, v2e);
return;
default:
- dev_err(&vport->adapter->pdev->dev,
+ dev_err(&adapter->pdev->dev,
"Unknown event %d from PF\n", event);
break;
}
@@ -213,125 +213,6 @@ int idpf_send_mb_msg(struct idpf_adapter *adapter, u32 op,
return err;
}
-/**
- * idpf_find_vport - Find vport pointer from control queue message
- * @adapter: driver specific private structure
- * @vport: address of vport pointer to copy the vport from adapters vport list
- * @ctlq_msg: control queue message
- *
- * Return 0 on success, error value on failure. Also this function does check
- * for the opcodes which expect to receive payload and return error value if
- * it is not the case.
- */
-static int idpf_find_vport(struct idpf_adapter *adapter,
- struct idpf_vport **vport,
- struct idpf_ctlq_msg *ctlq_msg)
-{
- bool no_op = false, vid_found = false;
- int i, err = 0;
- char *vc_msg;
- u32 v_id;
-
- vc_msg = kcalloc(IDPF_CTLQ_MAX_BUF_LEN, sizeof(char), GFP_KERNEL);
- if (!vc_msg)
- return -ENOMEM;
-
- if (ctlq_msg->data_len) {
- size_t payload_size = ctlq_msg->ctx.indirect.payload->size;
-
- if (!payload_size) {
- dev_err(&adapter->pdev->dev, "Failed to receive payload buffer\n");
- kfree(vc_msg);
-
- return -EINVAL;
- }
-
- memcpy(vc_msg, ctlq_msg->ctx.indirect.payload->va,
- min_t(size_t, payload_size, IDPF_CTLQ_MAX_BUF_LEN));
- }
-
- switch (ctlq_msg->cookie.mbx.chnl_opcode) {
- case VIRTCHNL2_OP_VERSION:
- case VIRTCHNL2_OP_GET_CAPS:
- case VIRTCHNL2_OP_CREATE_VPORT:
- case VIRTCHNL2_OP_SET_SRIOV_VFS:
- case VIRTCHNL2_OP_ALLOC_VECTORS:
- case VIRTCHNL2_OP_DEALLOC_VECTORS:
- case VIRTCHNL2_OP_GET_PTYPE_INFO:
- goto free_vc_msg;
- case VIRTCHNL2_OP_ENABLE_VPORT:
- case VIRTCHNL2_OP_DISABLE_VPORT:
- case VIRTCHNL2_OP_DESTROY_VPORT:
- v_id = le32_to_cpu(((struct virtchnl2_vport *)vc_msg)->vport_id);
- break;
- case VIRTCHNL2_OP_CONFIG_TX_QUEUES:
- v_id = le32_to_cpu(((struct virtchnl2_config_tx_queues *)vc_msg)->vport_id);
- break;
- case VIRTCHNL2_OP_CONFIG_RX_QUEUES:
- v_id = le32_to_cpu(((struct virtchnl2_config_rx_queues *)vc_msg)->vport_id);
- break;
- case VIRTCHNL2_OP_ENABLE_QUEUES:
- case VIRTCHNL2_OP_DISABLE_QUEUES:
- case VIRTCHNL2_OP_DEL_QUEUES:
- v_id = le32_to_cpu(((struct virtchnl2_del_ena_dis_queues *)vc_msg)->vport_id);
- break;
- case VIRTCHNL2_OP_ADD_QUEUES:
- v_id = le32_to_cpu(((struct virtchnl2_add_queues *)vc_msg)->vport_id);
- break;
- case VIRTCHNL2_OP_MAP_QUEUE_VECTOR:
- case VIRTCHNL2_OP_UNMAP_QUEUE_VECTOR:
- v_id = le32_to_cpu(((struct virtchnl2_queue_vector_maps *)vc_msg)->vport_id);
- break;
- case VIRTCHNL2_OP_GET_STATS:
- v_id = le32_to_cpu(((struct virtchnl2_vport_stats *)vc_msg)->vport_id);
- break;
- case VIRTCHNL2_OP_GET_RSS_LUT:
- case VIRTCHNL2_OP_SET_RSS_LUT:
- v_id = le32_to_cpu(((struct virtchnl2_rss_lut *)vc_msg)->vport_id);
- break;
- case VIRTCHNL2_OP_GET_RSS_KEY:
- case VIRTCHNL2_OP_SET_RSS_KEY:
- v_id = le32_to_cpu(((struct virtchnl2_rss_key *)vc_msg)->vport_id);
- break;
- case VIRTCHNL2_OP_EVENT:
- v_id = le32_to_cpu(((struct virtchnl2_event *)vc_msg)->vport_id);
- break;
- case VIRTCHNL2_OP_LOOPBACK:
- v_id = le32_to_cpu(((struct virtchnl2_loopback *)vc_msg)->vport_id);
- break;
- case VIRTCHNL2_OP_CONFIG_PROMISCUOUS_MODE:
- v_id = le32_to_cpu(((struct virtchnl2_promisc_info *)vc_msg)->vport_id);
- break;
- case VIRTCHNL2_OP_ADD_MAC_ADDR:
- case VIRTCHNL2_OP_DEL_MAC_ADDR:
- v_id = le32_to_cpu(((struct virtchnl2_mac_addr_list *)vc_msg)->vport_id);
- break;
- default:
- no_op = true;
- break;
- }
-
- if (no_op)
- goto free_vc_msg;
-
- for (i = 0; i < idpf_get_max_vports(adapter); i++) {
- if (adapter->vport_ids[i] == v_id) {
- vid_found = true;
- break;
- }
- }
-
- if (vid_found)
- *vport = adapter->vports[i];
- else
- err = -EINVAL;
-
-free_vc_msg:
- kfree(vc_msg);
-
- return err;
-}
-
/* API for virtchnl "transaction" support ("xn" for short).
*
* We are reusing the completion lock to serialize the accesses to the
@@ -707,118 +588,53 @@ idpf_vc_xn_forward_reply(struct idpf_adapter *adapter,
/**
* idpf_recv_mb_msg - Receive message over mailbox
* @adapter: Driver specific private structure
- * @op: virtchannel operation code
- * @msg: Received message holding buffer
- * @msg_size: message size
*
* Will receive control queue message and posts the receive buffer. Returns 0
* on success and negative on failure.
*/
-int idpf_recv_mb_msg(struct idpf_adapter *adapter, u32 op,
- void *msg, int msg_size)
+int idpf_recv_mb_msg(struct idpf_adapter *adapter)
{
- struct idpf_vport *vport = NULL;
struct idpf_ctlq_msg ctlq_msg;
struct idpf_dma_mem *dma_mem;
- bool work_done = false;
- int num_retry = 2000;
- u16 num_q_msg;
- int err;
+ int post_err, err;
+ u16 num_recv;
while (1) {
- /* Try to get one message */
- num_q_msg = 1;
- dma_mem = NULL;
- err = idpf_ctlq_recv(adapter->hw.arq, &num_q_msg, &ctlq_msg);
- /* If no message then decide if we have to retry based on
- * opcode
- */
- if (err || !num_q_msg) {
- /* Increasing num_retry to consider the delayed
- * responses because of large number of VF's mailbox
- * messages. If the mailbox message is received from
- * the other side, we come out of the sleep cycle
- * immediately else we wait for more time.
- */
- if (!op || !num_retry--)
- break;
- if (test_bit(IDPF_REMOVE_IN_PROG, adapter->flags)) {
- err = -EIO;
- break;
- }
- msleep(20);
- continue;
- }
-
- /* If we are here a message is received. Check if we are looking
- * for a specific message based on opcode. If it is different
- * ignore and post buffers
+ /* This will get <= num_recv messages and output how many
+ * actually received on num_recv.
*/
- if (op && ctlq_msg.cookie.mbx.chnl_opcode != op)
- goto post_buffs;
-
- err = idpf_find_vport(adapter, &vport, &ctlq_msg);
- if (err)
- goto post_buffs;
-
- /* All conditions are met. Either a message requested is
- * received or we received a message to be processed
- */
- switch (ctlq_msg.cookie.mbx.chnl_opcode) {
- case VIRTCHNL2_OP_VERSION:
- case VIRTCHNL2_OP_GET_CAPS:
- case VIRTCHNL2_OP_CREATE_VPORT:
- case VIRTCHNL2_OP_ENABLE_VPORT:
- case VIRTCHNL2_OP_DISABLE_VPORT:
- case VIRTCHNL2_OP_DESTROY_VPORT:
- case VIRTCHNL2_OP_CONFIG_TX_QUEUES:
- case VIRTCHNL2_OP_CONFIG_RX_QUEUES:
- case VIRTCHNL2_OP_ENABLE_QUEUES:
- case VIRTCHNL2_OP_DISABLE_QUEUES:
- case VIRTCHNL2_OP_ADD_QUEUES:
- case VIRTCHNL2_OP_DEL_QUEUES:
- case VIRTCHNL2_OP_MAP_QUEUE_VECTOR:
- case VIRTCHNL2_OP_UNMAP_QUEUE_VECTOR:
- case VIRTCHNL2_OP_GET_STATS:
- case VIRTCHNL2_OP_GET_RSS_LUT:
- case VIRTCHNL2_OP_SET_RSS_LUT:
- case VIRTCHNL2_OP_GET_RSS_KEY:
- case VIRTCHNL2_OP_SET_RSS_KEY:
- case VIRTCHNL2_OP_SET_SRIOV_VFS:
- case VIRTCHNL2_OP_ALLOC_VECTORS:
- case VIRTCHNL2_OP_DEALLOC_VECTORS:
- case VIRTCHNL2_OP_GET_PTYPE_INFO:
- case VIRTCHNL2_OP_LOOPBACK:
- case VIRTCHNL2_OP_CONFIG_PROMISCUOUS_MODE:
- case VIRTCHNL2_OP_ADD_MAC_ADDR:
- case VIRTCHNL2_OP_DEL_MAC_ADDR:
- err = idpf_vc_xn_forward_reply(adapter, &ctlq_msg);
- break;
- case VIRTCHNL2_OP_EVENT:
- idpf_recv_event_msg(vport, &ctlq_msg);
+ num_recv = 1;
+ err = idpf_ctlq_recv(adapter->hw.arq, &num_recv, &ctlq_msg);
+ if (err || !num_recv)
break;
- default:
- dev_warn(&adapter->pdev->dev,
- "Unhandled virtchnl response %d\n",
- ctlq_msg.cookie.mbx.chnl_opcode);
- break;
- }
-post_buffs:
- if (ctlq_msg.data_len)
+ if (ctlq_msg.data_len) {
dma_mem = ctlq_msg.ctx.indirect.payload;
+ } else {
+ dma_mem = NULL;
+ num_recv = 0;
+ }
+
+ if (ctlq_msg.cookie.mbx.chnl_opcode == VIRTCHNL2_OP_EVENT)
+ idpf_recv_event_msg(adapter, &ctlq_msg);
else
- num_q_msg = 0;
+ err = idpf_vc_xn_forward_reply(adapter, &ctlq_msg);
+
+ post_err = idpf_ctlq_post_rx_buffs(&adapter->hw,
+ adapter->hw.arq,
+ &num_recv, &dma_mem);
- err = idpf_ctlq_post_rx_buffs(&adapter->hw, adapter->hw.arq,
- &num_q_msg, &dma_mem);
/* If post failed clear the only buffer we supplied */
- if (err && dma_mem)
- dma_free_coherent(&adapter->pdev->dev, dma_mem->size,
- dma_mem->va, dma_mem->pa);
+ if (post_err) {
+ if (dma_mem)
+ dmam_free_coherent(&adapter->pdev->dev,
+ dma_mem->size, dma_mem->va,
+ dma_mem->pa);
+ break;
+ }
- /* Applies only if we are looking for a specific opcode */
- if (work_done)
+ /* virtchnl trying to shutdown, stop cleaning */
+ if (err == -ENXIO)
break;
}
--
2.40.1
^ permalink raw reply related [flat|nested] 44+ messages in thread
* [PATCH v4 07/10 iwl-next] idpf: cleanup virtchnl cruft
2024-02-06 3:37 ` [Intel-wired-lan] " Alan Brady
@ 2024-02-06 3:38 ` Alan Brady
-1 siblings, 0 replies; 44+ messages in thread
From: Alan Brady @ 2024-02-06 3:38 UTC (permalink / raw)
To: intel-wired-lan
Cc: netdev, willemdebruijn.kernel, przemyslaw.kitszel, igor.bagnucki,
aleksander.lobakin, Alan Brady
We can now remove a bunch of gross code we don't need anymore like the
vc state bits and vc_buf_lock since everything is using transaction API
now.
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Reviewed-by: Igor Bagnucki <igor.bagnucki@intel.com>
Signed-off-by: Alan Brady <alan.brady@intel.com>
---
drivers/net/ethernet/intel/idpf/idpf.h | 86 -------------------
drivers/net/ethernet/intel/idpf/idpf_lib.c | 25 +-----
drivers/net/ethernet/intel/idpf/idpf_main.c | 2 -
.../net/ethernet/intel/idpf/idpf_virtchnl.c | 13 ---
4 files changed, 2 insertions(+), 124 deletions(-)
diff --git a/drivers/net/ethernet/intel/idpf/idpf.h b/drivers/net/ethernet/intel/idpf/idpf.h
index 2d5449b9288a..0793173bb36d 100644
--- a/drivers/net/ethernet/intel/idpf/idpf.h
+++ b/drivers/net/ethernet/intel/idpf/idpf.h
@@ -37,8 +37,6 @@ struct idpf_vport_max_q;
#define IDPF_MB_MAX_ERR 20
#define IDPF_NUM_CHUNKS_PER_MSG(struct_sz, chunk_sz) \
((IDPF_CTLQ_MAX_BUF_LEN - (struct_sz)) / (chunk_sz))
-#define IDPF_WAIT_FOR_EVENT_TIMEO_MIN 2000
-#define IDPF_WAIT_FOR_EVENT_TIMEO 60000
#define IDPF_VC_XN_MIN_TIMEOUT_MSEC 2000
#define IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC (60 * 1000)
#define IDPF_VC_XN_IDX_M GENMASK(7, 0)
@@ -212,71 +210,6 @@ struct idpf_dev_ops {
struct idpf_reg_ops reg_ops;
};
-/* These macros allow us to generate an enum and a matching char * array of
- * stringified enums that are always in sync. Checkpatch issues a bogus warning
- * about this being a complex macro; but it's wrong, these are never used as a
- * statement and instead only used to define the enum and array.
- */
-#define IDPF_FOREACH_VPORT_VC_STATE(STATE) \
- STATE(IDPF_VC_CREATE_VPORT) \
- STATE(IDPF_VC_CREATE_VPORT_ERR) \
- STATE(IDPF_VC_ENA_VPORT) \
- STATE(IDPF_VC_ENA_VPORT_ERR) \
- STATE(IDPF_VC_DIS_VPORT) \
- STATE(IDPF_VC_DIS_VPORT_ERR) \
- STATE(IDPF_VC_DESTROY_VPORT) \
- STATE(IDPF_VC_DESTROY_VPORT_ERR) \
- STATE(IDPF_VC_CONFIG_TXQ) \
- STATE(IDPF_VC_CONFIG_TXQ_ERR) \
- STATE(IDPF_VC_CONFIG_RXQ) \
- STATE(IDPF_VC_CONFIG_RXQ_ERR) \
- STATE(IDPF_VC_ENA_QUEUES) \
- STATE(IDPF_VC_ENA_QUEUES_ERR) \
- STATE(IDPF_VC_DIS_QUEUES) \
- STATE(IDPF_VC_DIS_QUEUES_ERR) \
- STATE(IDPF_VC_MAP_IRQ) \
- STATE(IDPF_VC_MAP_IRQ_ERR) \
- STATE(IDPF_VC_UNMAP_IRQ) \
- STATE(IDPF_VC_UNMAP_IRQ_ERR) \
- STATE(IDPF_VC_ADD_QUEUES) \
- STATE(IDPF_VC_ADD_QUEUES_ERR) \
- STATE(IDPF_VC_DEL_QUEUES) \
- STATE(IDPF_VC_DEL_QUEUES_ERR) \
- STATE(IDPF_VC_ALLOC_VECTORS) \
- STATE(IDPF_VC_ALLOC_VECTORS_ERR) \
- STATE(IDPF_VC_DEALLOC_VECTORS) \
- STATE(IDPF_VC_DEALLOC_VECTORS_ERR) \
- STATE(IDPF_VC_SET_SRIOV_VFS) \
- STATE(IDPF_VC_SET_SRIOV_VFS_ERR) \
- STATE(IDPF_VC_GET_RSS_LUT) \
- STATE(IDPF_VC_GET_RSS_LUT_ERR) \
- STATE(IDPF_VC_SET_RSS_LUT) \
- STATE(IDPF_VC_SET_RSS_LUT_ERR) \
- STATE(IDPF_VC_GET_RSS_KEY) \
- STATE(IDPF_VC_GET_RSS_KEY_ERR) \
- STATE(IDPF_VC_SET_RSS_KEY) \
- STATE(IDPF_VC_SET_RSS_KEY_ERR) \
- STATE(IDPF_VC_GET_STATS) \
- STATE(IDPF_VC_GET_STATS_ERR) \
- STATE(IDPF_VC_ADD_MAC_ADDR) \
- STATE(IDPF_VC_ADD_MAC_ADDR_ERR) \
- STATE(IDPF_VC_DEL_MAC_ADDR) \
- STATE(IDPF_VC_DEL_MAC_ADDR_ERR) \
- STATE(IDPF_VC_GET_PTYPE_INFO) \
- STATE(IDPF_VC_GET_PTYPE_INFO_ERR) \
- STATE(IDPF_VC_LOOPBACK_STATE) \
- STATE(IDPF_VC_LOOPBACK_STATE_ERR) \
- STATE(IDPF_VC_NBITS)
-
-#define IDPF_GEN_ENUM(ENUM) ENUM,
-#define IDPF_GEN_STRING(STRING) #STRING,
-
-enum idpf_vport_vc_state {
- IDPF_FOREACH_VPORT_VC_STATE(IDPF_GEN_ENUM)
-};
-
-extern const char * const idpf_vport_vc_state_str[];
-
/**
* enum idpf_vport_reset_cause - Vport soft reset causes
* @IDPF_SR_Q_CHANGE: Soft reset queue change
@@ -451,11 +384,7 @@ struct idpf_vc_xn_manager {
* @port_stats: per port csum, header split, and other offload stats
* @link_up: True if link is up
* @link_speed_mbps: Link speed in mbps
- * @vc_msg: Virtchnl message buffer
- * @vc_state: Virtchnl message state
- * @vchnl_wq: Wait queue for virtchnl messages
* @sw_marker_wq: workqueue for marker packets
- * @vc_buf_lock: Lock to protect virtchnl buffer
*/
struct idpf_vport {
u16 num_txq;
@@ -501,12 +430,7 @@ struct idpf_vport {
bool link_up;
u32 link_speed_mbps;
- char vc_msg[IDPF_CTLQ_MAX_BUF_LEN];
- DECLARE_BITMAP(vc_state, IDPF_VC_NBITS);
-
- wait_queue_head_t vchnl_wq;
wait_queue_head_t sw_marker_wq;
- struct mutex vc_buf_lock;
};
/**
@@ -569,15 +493,11 @@ struct idpf_vport_user_config_data {
* enum idpf_vport_config_flags - Vport config flags
* @IDPF_VPORT_REG_NETDEV: Register netdev
* @IDPF_VPORT_UP_REQUESTED: Set if interface up is requested on core reset
- * @IDPF_VPORT_ADD_MAC_REQ: Asynchronous add ether address in flight
- * @IDPF_VPORT_DEL_MAC_REQ: Asynchronous delete ether address in flight
* @IDPF_VPORT_CONFIG_FLAGS_NBITS: Must be last
*/
enum idpf_vport_config_flags {
IDPF_VPORT_REG_NETDEV,
IDPF_VPORT_UP_REQUESTED,
- IDPF_VPORT_ADD_MAC_REQ,
- IDPF_VPORT_DEL_MAC_REQ,
IDPF_VPORT_CONFIG_FLAGS_NBITS,
};
@@ -694,9 +614,6 @@ struct idpf_vport_config {
* @stats_task: Periodic statistics retrieval task
* @stats_wq: Workqueue for statistics task
* @caps: Negotiated capabilities with device
- * @vchnl_wq: Wait queue for virtchnl messages
- * @vc_state: Virtchnl message state
- * @vc_msg: Virtchnl message buffer
* @vcxn_mngr: Virtchnl transaction manager
* @dev_ops: See idpf_dev_ops
* @num_vfs: Number of allocated VFs through sysfs. PF does not directly talk
@@ -754,9 +671,6 @@ struct idpf_adapter {
struct workqueue_struct *stats_wq;
struct virtchnl2_get_capabilities caps;
- wait_queue_head_t vchnl_wq;
- DECLARE_BITMAP(vc_state, IDPF_VC_NBITS);
- char vc_msg[IDPF_CTLQ_MAX_BUF_LEN];
struct idpf_vc_xn_manager vcxn_mngr;
struct idpf_dev_ops dev_ops;
int num_vfs;
diff --git a/drivers/net/ethernet/intel/idpf/idpf_lib.c b/drivers/net/ethernet/intel/idpf/idpf_lib.c
index 8135a6b4a71e..1aae6963628b 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_lib.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_lib.c
@@ -6,10 +6,6 @@
static const struct net_device_ops idpf_netdev_ops_splitq;
static const struct net_device_ops idpf_netdev_ops_singleq;
-const char * const idpf_vport_vc_state_str[] = {
- IDPF_FOREACH_VPORT_VC_STATE(IDPF_GEN_STRING)
-};
-
/**
* idpf_init_vector_stack - Fill the MSIX vector stack with vector index
* @adapter: private data struct
@@ -975,7 +971,6 @@ static void idpf_vport_rel(struct idpf_vport *vport)
struct idpf_rss_data *rss_data;
struct idpf_vport_max_q max_q;
u16 idx = vport->idx;
- int i;
vport_config = adapter->vport_config[vport->idx];
idpf_deinit_rss(vport);
@@ -985,20 +980,6 @@ static void idpf_vport_rel(struct idpf_vport *vport)
idpf_send_destroy_vport_msg(vport);
- /* Set all bits as we dont know on which vc_state the vport vhnl_wq
- * is waiting on and wakeup the virtchnl workqueue even if it is
- * waiting for the response as we are going down
- */
- for (i = 0; i < IDPF_VC_NBITS; i++)
- set_bit(i, vport->vc_state);
- wake_up(&vport->vchnl_wq);
-
- mutex_destroy(&vport->vc_buf_lock);
-
- /* Clear all the bits */
- for (i = 0; i < IDPF_VC_NBITS; i++)
- clear_bit(i, vport->vc_state);
-
/* Release all max queues allocated to the adapter's pool */
max_q.max_rxq = vport_config->max_q.max_rxq;
max_q.max_txq = vport_config->max_q.max_txq;
@@ -1543,9 +1524,7 @@ void idpf_init_task(struct work_struct *work)
vport_config = adapter->vport_config[index];
init_waitqueue_head(&vport->sw_marker_wq);
- init_waitqueue_head(&vport->vchnl_wq);
- mutex_init(&vport->vc_buf_lock);
spin_lock_init(&vport_config->mac_filter_list_lock);
INIT_LIST_HEAD(&vport_config->user_config.mac_filter_list);
@@ -1904,7 +1883,7 @@ int idpf_initiate_soft_reset(struct idpf_vport *vport,
* mess with. Nothing below should use those variables from new_vport
* and should instead always refer to them in vport if they need to.
*/
- memcpy(new_vport, vport, offsetof(struct idpf_vport, vc_state));
+ memcpy(new_vport, vport, offsetof(struct idpf_vport, link_speed_mbps));
/* Adjust resource parameters prior to reallocating resources */
switch (reset_cause) {
@@ -1953,7 +1932,7 @@ int idpf_initiate_soft_reset(struct idpf_vport *vport,
/* Same comment as above regarding avoiding copying the wait_queues and
* mutexes applies here. We do not want to mess with those if possible.
*/
- memcpy(vport, new_vport, offsetof(struct idpf_vport, vc_state));
+ memcpy(vport, new_vport, offsetof(struct idpf_vport, link_speed_mbps));
/* Since idpf_vport_queues_alloc was called with new_port, the queue
* back pointers are currently pointing to the local new_vport. Reset
diff --git a/drivers/net/ethernet/intel/idpf/idpf_main.c b/drivers/net/ethernet/intel/idpf/idpf_main.c
index c82233b112bd..c6d7bec95d69 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_main.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_main.c
@@ -230,8 +230,6 @@ static int idpf_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
mutex_init(&adapter->queue_lock);
mutex_init(&adapter->vc_buf_lock);
- init_waitqueue_head(&adapter->vchnl_wq);
-
INIT_DELAYED_WORK(&adapter->init_task, idpf_init_task);
INIT_DELAYED_WORK(&adapter->serv_task, idpf_service_task);
INIT_DELAYED_WORK(&adapter->mbx_task, idpf_mbx_task);
diff --git a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
index 5a68a3908cf0..f6e6e3eff29e 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
@@ -2929,28 +2929,15 @@ int idpf_vc_core_init(struct idpf_adapter *adapter)
*/
void idpf_vc_core_deinit(struct idpf_adapter *adapter)
{
- int i;
-
idpf_vc_xn_shutdown(&adapter->vcxn_mngr);
idpf_deinit_task(adapter);
idpf_intr_rel(adapter);
- /* Set all bits as we dont know on which vc_state the vhnl_wq is
- * waiting on and wakeup the virtchnl workqueue even if it is waiting
- * for the response as we are going down
- */
- for (i = 0; i < IDPF_VC_NBITS; i++)
- set_bit(i, adapter->vc_state);
- wake_up(&adapter->vchnl_wq);
cancel_delayed_work_sync(&adapter->serv_task);
cancel_delayed_work_sync(&adapter->mbx_task);
idpf_vport_params_buf_rel(adapter);
- /* Clear all the bits */
- for (i = 0; i < IDPF_VC_NBITS; i++)
- clear_bit(i, adapter->vc_state);
-
kfree(adapter->vports);
adapter->vports = NULL;
}
--
2.40.1
^ permalink raw reply related [flat|nested] 44+ messages in thread
* [Intel-wired-lan] [PATCH v4 07/10 iwl-next] idpf: cleanup virtchnl cruft
@ 2024-02-06 3:38 ` Alan Brady
0 siblings, 0 replies; 44+ messages in thread
From: Alan Brady @ 2024-02-06 3:38 UTC (permalink / raw)
To: intel-wired-lan
Cc: willemdebruijn.kernel, netdev, aleksander.lobakin, Alan Brady,
przemyslaw.kitszel, igor.bagnucki
We can now remove a bunch of gross code we don't need anymore like the
vc state bits and vc_buf_lock since everything is using transaction API
now.
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Reviewed-by: Igor Bagnucki <igor.bagnucki@intel.com>
Signed-off-by: Alan Brady <alan.brady@intel.com>
---
drivers/net/ethernet/intel/idpf/idpf.h | 86 -------------------
drivers/net/ethernet/intel/idpf/idpf_lib.c | 25 +-----
drivers/net/ethernet/intel/idpf/idpf_main.c | 2 -
.../net/ethernet/intel/idpf/idpf_virtchnl.c | 13 ---
4 files changed, 2 insertions(+), 124 deletions(-)
diff --git a/drivers/net/ethernet/intel/idpf/idpf.h b/drivers/net/ethernet/intel/idpf/idpf.h
index 2d5449b9288a..0793173bb36d 100644
--- a/drivers/net/ethernet/intel/idpf/idpf.h
+++ b/drivers/net/ethernet/intel/idpf/idpf.h
@@ -37,8 +37,6 @@ struct idpf_vport_max_q;
#define IDPF_MB_MAX_ERR 20
#define IDPF_NUM_CHUNKS_PER_MSG(struct_sz, chunk_sz) \
((IDPF_CTLQ_MAX_BUF_LEN - (struct_sz)) / (chunk_sz))
-#define IDPF_WAIT_FOR_EVENT_TIMEO_MIN 2000
-#define IDPF_WAIT_FOR_EVENT_TIMEO 60000
#define IDPF_VC_XN_MIN_TIMEOUT_MSEC 2000
#define IDPF_VC_XN_DEFAULT_TIMEOUT_MSEC (60 * 1000)
#define IDPF_VC_XN_IDX_M GENMASK(7, 0)
@@ -212,71 +210,6 @@ struct idpf_dev_ops {
struct idpf_reg_ops reg_ops;
};
-/* These macros allow us to generate an enum and a matching char * array of
- * stringified enums that are always in sync. Checkpatch issues a bogus warning
- * about this being a complex macro; but it's wrong, these are never used as a
- * statement and instead only used to define the enum and array.
- */
-#define IDPF_FOREACH_VPORT_VC_STATE(STATE) \
- STATE(IDPF_VC_CREATE_VPORT) \
- STATE(IDPF_VC_CREATE_VPORT_ERR) \
- STATE(IDPF_VC_ENA_VPORT) \
- STATE(IDPF_VC_ENA_VPORT_ERR) \
- STATE(IDPF_VC_DIS_VPORT) \
- STATE(IDPF_VC_DIS_VPORT_ERR) \
- STATE(IDPF_VC_DESTROY_VPORT) \
- STATE(IDPF_VC_DESTROY_VPORT_ERR) \
- STATE(IDPF_VC_CONFIG_TXQ) \
- STATE(IDPF_VC_CONFIG_TXQ_ERR) \
- STATE(IDPF_VC_CONFIG_RXQ) \
- STATE(IDPF_VC_CONFIG_RXQ_ERR) \
- STATE(IDPF_VC_ENA_QUEUES) \
- STATE(IDPF_VC_ENA_QUEUES_ERR) \
- STATE(IDPF_VC_DIS_QUEUES) \
- STATE(IDPF_VC_DIS_QUEUES_ERR) \
- STATE(IDPF_VC_MAP_IRQ) \
- STATE(IDPF_VC_MAP_IRQ_ERR) \
- STATE(IDPF_VC_UNMAP_IRQ) \
- STATE(IDPF_VC_UNMAP_IRQ_ERR) \
- STATE(IDPF_VC_ADD_QUEUES) \
- STATE(IDPF_VC_ADD_QUEUES_ERR) \
- STATE(IDPF_VC_DEL_QUEUES) \
- STATE(IDPF_VC_DEL_QUEUES_ERR) \
- STATE(IDPF_VC_ALLOC_VECTORS) \
- STATE(IDPF_VC_ALLOC_VECTORS_ERR) \
- STATE(IDPF_VC_DEALLOC_VECTORS) \
- STATE(IDPF_VC_DEALLOC_VECTORS_ERR) \
- STATE(IDPF_VC_SET_SRIOV_VFS) \
- STATE(IDPF_VC_SET_SRIOV_VFS_ERR) \
- STATE(IDPF_VC_GET_RSS_LUT) \
- STATE(IDPF_VC_GET_RSS_LUT_ERR) \
- STATE(IDPF_VC_SET_RSS_LUT) \
- STATE(IDPF_VC_SET_RSS_LUT_ERR) \
- STATE(IDPF_VC_GET_RSS_KEY) \
- STATE(IDPF_VC_GET_RSS_KEY_ERR) \
- STATE(IDPF_VC_SET_RSS_KEY) \
- STATE(IDPF_VC_SET_RSS_KEY_ERR) \
- STATE(IDPF_VC_GET_STATS) \
- STATE(IDPF_VC_GET_STATS_ERR) \
- STATE(IDPF_VC_ADD_MAC_ADDR) \
- STATE(IDPF_VC_ADD_MAC_ADDR_ERR) \
- STATE(IDPF_VC_DEL_MAC_ADDR) \
- STATE(IDPF_VC_DEL_MAC_ADDR_ERR) \
- STATE(IDPF_VC_GET_PTYPE_INFO) \
- STATE(IDPF_VC_GET_PTYPE_INFO_ERR) \
- STATE(IDPF_VC_LOOPBACK_STATE) \
- STATE(IDPF_VC_LOOPBACK_STATE_ERR) \
- STATE(IDPF_VC_NBITS)
-
-#define IDPF_GEN_ENUM(ENUM) ENUM,
-#define IDPF_GEN_STRING(STRING) #STRING,
-
-enum idpf_vport_vc_state {
- IDPF_FOREACH_VPORT_VC_STATE(IDPF_GEN_ENUM)
-};
-
-extern const char * const idpf_vport_vc_state_str[];
-
/**
* enum idpf_vport_reset_cause - Vport soft reset causes
* @IDPF_SR_Q_CHANGE: Soft reset queue change
@@ -451,11 +384,7 @@ struct idpf_vc_xn_manager {
* @port_stats: per port csum, header split, and other offload stats
* @link_up: True if link is up
* @link_speed_mbps: Link speed in mbps
- * @vc_msg: Virtchnl message buffer
- * @vc_state: Virtchnl message state
- * @vchnl_wq: Wait queue for virtchnl messages
* @sw_marker_wq: workqueue for marker packets
- * @vc_buf_lock: Lock to protect virtchnl buffer
*/
struct idpf_vport {
u16 num_txq;
@@ -501,12 +430,7 @@ struct idpf_vport {
bool link_up;
u32 link_speed_mbps;
- char vc_msg[IDPF_CTLQ_MAX_BUF_LEN];
- DECLARE_BITMAP(vc_state, IDPF_VC_NBITS);
-
- wait_queue_head_t vchnl_wq;
wait_queue_head_t sw_marker_wq;
- struct mutex vc_buf_lock;
};
/**
@@ -569,15 +493,11 @@ struct idpf_vport_user_config_data {
* enum idpf_vport_config_flags - Vport config flags
* @IDPF_VPORT_REG_NETDEV: Register netdev
* @IDPF_VPORT_UP_REQUESTED: Set if interface up is requested on core reset
- * @IDPF_VPORT_ADD_MAC_REQ: Asynchronous add ether address in flight
- * @IDPF_VPORT_DEL_MAC_REQ: Asynchronous delete ether address in flight
* @IDPF_VPORT_CONFIG_FLAGS_NBITS: Must be last
*/
enum idpf_vport_config_flags {
IDPF_VPORT_REG_NETDEV,
IDPF_VPORT_UP_REQUESTED,
- IDPF_VPORT_ADD_MAC_REQ,
- IDPF_VPORT_DEL_MAC_REQ,
IDPF_VPORT_CONFIG_FLAGS_NBITS,
};
@@ -694,9 +614,6 @@ struct idpf_vport_config {
* @stats_task: Periodic statistics retrieval task
* @stats_wq: Workqueue for statistics task
* @caps: Negotiated capabilities with device
- * @vchnl_wq: Wait queue for virtchnl messages
- * @vc_state: Virtchnl message state
- * @vc_msg: Virtchnl message buffer
* @vcxn_mngr: Virtchnl transaction manager
* @dev_ops: See idpf_dev_ops
* @num_vfs: Number of allocated VFs through sysfs. PF does not directly talk
@@ -754,9 +671,6 @@ struct idpf_adapter {
struct workqueue_struct *stats_wq;
struct virtchnl2_get_capabilities caps;
- wait_queue_head_t vchnl_wq;
- DECLARE_BITMAP(vc_state, IDPF_VC_NBITS);
- char vc_msg[IDPF_CTLQ_MAX_BUF_LEN];
struct idpf_vc_xn_manager vcxn_mngr;
struct idpf_dev_ops dev_ops;
int num_vfs;
diff --git a/drivers/net/ethernet/intel/idpf/idpf_lib.c b/drivers/net/ethernet/intel/idpf/idpf_lib.c
index 8135a6b4a71e..1aae6963628b 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_lib.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_lib.c
@@ -6,10 +6,6 @@
static const struct net_device_ops idpf_netdev_ops_splitq;
static const struct net_device_ops idpf_netdev_ops_singleq;
-const char * const idpf_vport_vc_state_str[] = {
- IDPF_FOREACH_VPORT_VC_STATE(IDPF_GEN_STRING)
-};
-
/**
* idpf_init_vector_stack - Fill the MSIX vector stack with vector index
* @adapter: private data struct
@@ -975,7 +971,6 @@ static void idpf_vport_rel(struct idpf_vport *vport)
struct idpf_rss_data *rss_data;
struct idpf_vport_max_q max_q;
u16 idx = vport->idx;
- int i;
vport_config = adapter->vport_config[vport->idx];
idpf_deinit_rss(vport);
@@ -985,20 +980,6 @@ static void idpf_vport_rel(struct idpf_vport *vport)
idpf_send_destroy_vport_msg(vport);
- /* Set all bits as we dont know on which vc_state the vport vhnl_wq
- * is waiting on and wakeup the virtchnl workqueue even if it is
- * waiting for the response as we are going down
- */
- for (i = 0; i < IDPF_VC_NBITS; i++)
- set_bit(i, vport->vc_state);
- wake_up(&vport->vchnl_wq);
-
- mutex_destroy(&vport->vc_buf_lock);
-
- /* Clear all the bits */
- for (i = 0; i < IDPF_VC_NBITS; i++)
- clear_bit(i, vport->vc_state);
-
/* Release all max queues allocated to the adapter's pool */
max_q.max_rxq = vport_config->max_q.max_rxq;
max_q.max_txq = vport_config->max_q.max_txq;
@@ -1543,9 +1524,7 @@ void idpf_init_task(struct work_struct *work)
vport_config = adapter->vport_config[index];
init_waitqueue_head(&vport->sw_marker_wq);
- init_waitqueue_head(&vport->vchnl_wq);
- mutex_init(&vport->vc_buf_lock);
spin_lock_init(&vport_config->mac_filter_list_lock);
INIT_LIST_HEAD(&vport_config->user_config.mac_filter_list);
@@ -1904,7 +1883,7 @@ int idpf_initiate_soft_reset(struct idpf_vport *vport,
* mess with. Nothing below should use those variables from new_vport
* and should instead always refer to them in vport if they need to.
*/
- memcpy(new_vport, vport, offsetof(struct idpf_vport, vc_state));
+ memcpy(new_vport, vport, offsetof(struct idpf_vport, link_speed_mbps));
/* Adjust resource parameters prior to reallocating resources */
switch (reset_cause) {
@@ -1953,7 +1932,7 @@ int idpf_initiate_soft_reset(struct idpf_vport *vport,
/* Same comment as above regarding avoiding copying the wait_queues and
* mutexes applies here. We do not want to mess with those if possible.
*/
- memcpy(vport, new_vport, offsetof(struct idpf_vport, vc_state));
+ memcpy(vport, new_vport, offsetof(struct idpf_vport, link_speed_mbps));
/* Since idpf_vport_queues_alloc was called with new_port, the queue
* back pointers are currently pointing to the local new_vport. Reset
diff --git a/drivers/net/ethernet/intel/idpf/idpf_main.c b/drivers/net/ethernet/intel/idpf/idpf_main.c
index c82233b112bd..c6d7bec95d69 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_main.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_main.c
@@ -230,8 +230,6 @@ static int idpf_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
mutex_init(&adapter->queue_lock);
mutex_init(&adapter->vc_buf_lock);
- init_waitqueue_head(&adapter->vchnl_wq);
-
INIT_DELAYED_WORK(&adapter->init_task, idpf_init_task);
INIT_DELAYED_WORK(&adapter->serv_task, idpf_service_task);
INIT_DELAYED_WORK(&adapter->mbx_task, idpf_mbx_task);
diff --git a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
index 5a68a3908cf0..f6e6e3eff29e 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
@@ -2929,28 +2929,15 @@ int idpf_vc_core_init(struct idpf_adapter *adapter)
*/
void idpf_vc_core_deinit(struct idpf_adapter *adapter)
{
- int i;
-
idpf_vc_xn_shutdown(&adapter->vcxn_mngr);
idpf_deinit_task(adapter);
idpf_intr_rel(adapter);
- /* Set all bits as we dont know on which vc_state the vhnl_wq is
- * waiting on and wakeup the virtchnl workqueue even if it is waiting
- * for the response as we are going down
- */
- for (i = 0; i < IDPF_VC_NBITS; i++)
- set_bit(i, adapter->vc_state);
- wake_up(&adapter->vchnl_wq);
cancel_delayed_work_sync(&adapter->serv_task);
cancel_delayed_work_sync(&adapter->mbx_task);
idpf_vport_params_buf_rel(adapter);
- /* Clear all the bits */
- for (i = 0; i < IDPF_VC_NBITS; i++)
- clear_bit(i, adapter->vc_state);
-
kfree(adapter->vports);
adapter->vports = NULL;
}
--
2.40.1
^ permalink raw reply related [flat|nested] 44+ messages in thread
* [PATCH v4 08/10 iwl-next] idpf: prevent deinit uninitialized virtchnl core
2024-02-06 3:37 ` [Intel-wired-lan] " Alan Brady
@ 2024-02-06 3:38 ` Alan Brady
-1 siblings, 0 replies; 44+ messages in thread
From: Alan Brady @ 2024-02-06 3:38 UTC (permalink / raw)
To: intel-wired-lan
Cc: netdev, willemdebruijn.kernel, przemyslaw.kitszel, igor.bagnucki,
aleksander.lobakin, Alan Brady
In idpf_remove we need to tear down the virtchnl core with
idpf_vc_core_deinit so we can free up resources and leave things in a
good state. However, in the case where we failed to establish VC
communications we may not have ever actually successfully initialized
the virtchnl core.
This fixes it by setting a bit once we successfully init the virtchnl
core. Then, in deinit, we'll check for it before going on further,
otherwise we just return. Also clear the bit at the end of deinit so we
know it's gone now.
Signed-off-by: Alan Brady <alan.brady@intel.com>
---
drivers/net/ethernet/intel/idpf/idpf.h | 2 ++
drivers/net/ethernet/intel/idpf/idpf_virtchnl.c | 10 ++++++++--
2 files changed, 10 insertions(+), 2 deletions(-)
diff --git a/drivers/net/ethernet/intel/idpf/idpf.h b/drivers/net/ethernet/intel/idpf/idpf.h
index 0793173bb36d..e7a2fbf9e6cd 100644
--- a/drivers/net/ethernet/intel/idpf/idpf.h
+++ b/drivers/net/ethernet/intel/idpf/idpf.h
@@ -88,6 +88,7 @@ enum idpf_state {
* @IDPF_HR_RESET_IN_PROG: Reset in progress
* @IDPF_REMOVE_IN_PROG: Driver remove in progress
* @IDPF_MB_INTR_MODE: Mailbox in interrupt mode
+ * @IDPF_VC_CORE_INIT: virtchnl core has been init
* @IDPF_FLAGS_NBITS: Must be last
*/
enum idpf_flags {
@@ -96,6 +97,7 @@ enum idpf_flags {
IDPF_HR_RESET_IN_PROG,
IDPF_REMOVE_IN_PROG,
IDPF_MB_INTR_MODE,
+ IDPF_VC_CORE_INIT,
IDPF_FLAGS_NBITS,
};
diff --git a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
index f6e6e3eff29e..2bf67f13079d 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
@@ -2886,7 +2886,9 @@ int idpf_vc_core_init(struct idpf_adapter *adapter)
queue_delayed_work(adapter->init_wq, &adapter->init_task,
msecs_to_jiffies(5 * (adapter->pdev->devfn & 0x07)));
- goto no_err;
+ set_bit(IDPF_VC_CORE_INIT, adapter->flags);
+
+ return 0;
err_intr_req:
cancel_delayed_work_sync(&adapter->serv_task);
@@ -2895,7 +2897,6 @@ int idpf_vc_core_init(struct idpf_adapter *adapter)
err_netdev_alloc:
kfree(adapter->vports);
adapter->vports = NULL;
-no_err:
return err;
init_failed:
@@ -2929,6 +2930,9 @@ int idpf_vc_core_init(struct idpf_adapter *adapter)
*/
void idpf_vc_core_deinit(struct idpf_adapter *adapter)
{
+ if (!test_bit(IDPF_VC_CORE_INIT, adapter->flags))
+ return;
+
idpf_vc_xn_shutdown(&adapter->vcxn_mngr);
idpf_deinit_task(adapter);
idpf_intr_rel(adapter);
@@ -2940,6 +2944,8 @@ void idpf_vc_core_deinit(struct idpf_adapter *adapter)
kfree(adapter->vports);
adapter->vports = NULL;
+
+ clear_bit(IDPF_VC_CORE_INIT, adapter->flags);
}
/**
--
2.40.1
^ permalink raw reply related [flat|nested] 44+ messages in thread
* [Intel-wired-lan] [PATCH v4 08/10 iwl-next] idpf: prevent deinit uninitialized virtchnl core
@ 2024-02-06 3:38 ` Alan Brady
0 siblings, 0 replies; 44+ messages in thread
From: Alan Brady @ 2024-02-06 3:38 UTC (permalink / raw)
To: intel-wired-lan
Cc: willemdebruijn.kernel, netdev, aleksander.lobakin, Alan Brady,
przemyslaw.kitszel, igor.bagnucki
In idpf_remove we need to tear down the virtchnl core with
idpf_vc_core_deinit so we can free up resources and leave things in a
good state. However, in the case where we failed to establish VC
communications we may not have ever actually successfully initialized
the virtchnl core.
This fixes it by setting a bit once we successfully init the virtchnl
core. Then, in deinit, we'll check for it before going on further,
otherwise we just return. Also clear the bit at the end of deinit so we
know it's gone now.
Signed-off-by: Alan Brady <alan.brady@intel.com>
---
drivers/net/ethernet/intel/idpf/idpf.h | 2 ++
drivers/net/ethernet/intel/idpf/idpf_virtchnl.c | 10 ++++++++--
2 files changed, 10 insertions(+), 2 deletions(-)
diff --git a/drivers/net/ethernet/intel/idpf/idpf.h b/drivers/net/ethernet/intel/idpf/idpf.h
index 0793173bb36d..e7a2fbf9e6cd 100644
--- a/drivers/net/ethernet/intel/idpf/idpf.h
+++ b/drivers/net/ethernet/intel/idpf/idpf.h
@@ -88,6 +88,7 @@ enum idpf_state {
* @IDPF_HR_RESET_IN_PROG: Reset in progress
* @IDPF_REMOVE_IN_PROG: Driver remove in progress
* @IDPF_MB_INTR_MODE: Mailbox in interrupt mode
+ * @IDPF_VC_CORE_INIT: virtchnl core has been init
* @IDPF_FLAGS_NBITS: Must be last
*/
enum idpf_flags {
@@ -96,6 +97,7 @@ enum idpf_flags {
IDPF_HR_RESET_IN_PROG,
IDPF_REMOVE_IN_PROG,
IDPF_MB_INTR_MODE,
+ IDPF_VC_CORE_INIT,
IDPF_FLAGS_NBITS,
};
diff --git a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
index f6e6e3eff29e..2bf67f13079d 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
@@ -2886,7 +2886,9 @@ int idpf_vc_core_init(struct idpf_adapter *adapter)
queue_delayed_work(adapter->init_wq, &adapter->init_task,
msecs_to_jiffies(5 * (adapter->pdev->devfn & 0x07)));
- goto no_err;
+ set_bit(IDPF_VC_CORE_INIT, adapter->flags);
+
+ return 0;
err_intr_req:
cancel_delayed_work_sync(&adapter->serv_task);
@@ -2895,7 +2897,6 @@ int idpf_vc_core_init(struct idpf_adapter *adapter)
err_netdev_alloc:
kfree(adapter->vports);
adapter->vports = NULL;
-no_err:
return err;
init_failed:
@@ -2929,6 +2930,9 @@ int idpf_vc_core_init(struct idpf_adapter *adapter)
*/
void idpf_vc_core_deinit(struct idpf_adapter *adapter)
{
+ if (!test_bit(IDPF_VC_CORE_INIT, adapter->flags))
+ return;
+
idpf_vc_xn_shutdown(&adapter->vcxn_mngr);
idpf_deinit_task(adapter);
idpf_intr_rel(adapter);
@@ -2940,6 +2944,8 @@ void idpf_vc_core_deinit(struct idpf_adapter *adapter)
kfree(adapter->vports);
adapter->vports = NULL;
+
+ clear_bit(IDPF_VC_CORE_INIT, adapter->flags);
}
/**
--
2.40.1
^ permalink raw reply related [flat|nested] 44+ messages in thread
* [PATCH v4 09/10 iwl-next] idpf: fix minor controlq issues
2024-02-06 3:37 ` [Intel-wired-lan] " Alan Brady
@ 2024-02-06 3:38 ` Alan Brady
-1 siblings, 0 replies; 44+ messages in thread
From: Alan Brady @ 2024-02-06 3:38 UTC (permalink / raw)
To: intel-wired-lan
Cc: netdev, willemdebruijn.kernel, przemyslaw.kitszel, igor.bagnucki,
aleksander.lobakin, Alan Brady
While we're here improving virtchnl we can include two minor fixes for
the lower level ctrlq flow.
This adds a memory barrier to idpf_post_rx_buffs before we update tail
on the controlq. We should make sure our writes have had a chance to
finish before we tell HW it can touch them.
This also removes some defensive programming in idpf_ctrlq_recv. The
caller should not be using a num_q_msg value of zero or more than the
ring size and it's their responsibility to call functions sanely.
Signed-off-by: Alan Brady <alan.brady@intel.com>
---
drivers/net/ethernet/intel/idpf/idpf_controlq.c | 7 ++-----
1 file changed, 2 insertions(+), 5 deletions(-)
diff --git a/drivers/net/ethernet/intel/idpf/idpf_controlq.c b/drivers/net/ethernet/intel/idpf/idpf_controlq.c
index c7f43d2fcd13..4849590a5591 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_controlq.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_controlq.c
@@ -516,6 +516,8 @@ int idpf_ctlq_post_rx_buffs(struct idpf_hw *hw, struct idpf_ctlq_info *cq,
/* Wrap to end of end ring since current ntp is 0 */
cq->next_to_post = cq->ring_size - 1;
+ dma_wmb();
+
wr32(hw, cq->reg.tail, cq->next_to_post);
}
@@ -546,11 +548,6 @@ int idpf_ctlq_recv(struct idpf_ctlq_info *cq, u16 *num_q_msg,
int err = 0;
u16 i;
- if (*num_q_msg == 0)
- return 0;
- else if (*num_q_msg > cq->ring_size)
- return -EBADR;
-
/* take the lock before we start messing with the ring */
mutex_lock(&cq->cq_lock);
--
2.40.1
^ permalink raw reply related [flat|nested] 44+ messages in thread
* [Intel-wired-lan] [PATCH v4 09/10 iwl-next] idpf: fix minor controlq issues
@ 2024-02-06 3:38 ` Alan Brady
0 siblings, 0 replies; 44+ messages in thread
From: Alan Brady @ 2024-02-06 3:38 UTC (permalink / raw)
To: intel-wired-lan
Cc: willemdebruijn.kernel, netdev, aleksander.lobakin, Alan Brady,
przemyslaw.kitszel, igor.bagnucki
While we're here improving virtchnl we can include two minor fixes for
the lower level ctrlq flow.
This adds a memory barrier to idpf_post_rx_buffs before we update tail
on the controlq. We should make sure our writes have had a chance to
finish before we tell HW it can touch them.
This also removes some defensive programming in idpf_ctrlq_recv. The
caller should not be using a num_q_msg value of zero or more than the
ring size and it's their responsibility to call functions sanely.
Signed-off-by: Alan Brady <alan.brady@intel.com>
---
drivers/net/ethernet/intel/idpf/idpf_controlq.c | 7 ++-----
1 file changed, 2 insertions(+), 5 deletions(-)
diff --git a/drivers/net/ethernet/intel/idpf/idpf_controlq.c b/drivers/net/ethernet/intel/idpf/idpf_controlq.c
index c7f43d2fcd13..4849590a5591 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_controlq.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_controlq.c
@@ -516,6 +516,8 @@ int idpf_ctlq_post_rx_buffs(struct idpf_hw *hw, struct idpf_ctlq_info *cq,
/* Wrap to end of end ring since current ntp is 0 */
cq->next_to_post = cq->ring_size - 1;
+ dma_wmb();
+
wr32(hw, cq->reg.tail, cq->next_to_post);
}
@@ -546,11 +548,6 @@ int idpf_ctlq_recv(struct idpf_ctlq_info *cq, u16 *num_q_msg,
int err = 0;
u16 i;
- if (*num_q_msg == 0)
- return 0;
- else if (*num_q_msg > cq->ring_size)
- return -EBADR;
-
/* take the lock before we start messing with the ring */
mutex_lock(&cq->cq_lock);
--
2.40.1
^ permalink raw reply related [flat|nested] 44+ messages in thread
* [PATCH v4 10/10 iwl-next] idpf: remove dealloc vector msg err in idpf_intr_rel
2024-02-06 3:37 ` [Intel-wired-lan] " Alan Brady
@ 2024-02-06 3:38 ` Alan Brady
-1 siblings, 0 replies; 44+ messages in thread
From: Alan Brady @ 2024-02-06 3:38 UTC (permalink / raw)
To: intel-wired-lan
Cc: netdev, willemdebruijn.kernel, przemyslaw.kitszel, igor.bagnucki,
aleksander.lobakin, Alan Brady
This error message is at best not really helpful and at worst
misleading. If we're here in idpf_intr_rel we're likely trying to do
remove or reset. If we're in reset, this message will fail because we
lose the virtchnl on reset and HW is going to clean up those resources
regardless in that case. If we're in remove and we get an error here,
we're going to reset the device at the end of remove anyway so not a big
deal. Just remove this message it's not useful.
Signed-off-by: Alan Brady <alan.brady@intel.com>
---
drivers/net/ethernet/intel/idpf/idpf_lib.c | 9 +--------
1 file changed, 1 insertion(+), 8 deletions(-)
diff --git a/drivers/net/ethernet/intel/idpf/idpf_lib.c b/drivers/net/ethernet/intel/idpf/idpf_lib.c
index 1aae6963628b..1e30ef98cd68 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_lib.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_lib.c
@@ -78,19 +78,12 @@ static void idpf_mb_intr_rel_irq(struct idpf_adapter *adapter)
*/
void idpf_intr_rel(struct idpf_adapter *adapter)
{
- int err;
-
if (!adapter->msix_entries)
return;
idpf_mb_intr_rel_irq(adapter);
pci_free_irq_vectors(adapter->pdev);
-
- err = idpf_send_dealloc_vectors_msg(adapter);
- if (err)
- dev_err(&adapter->pdev->dev,
- "Failed to deallocate vectors: %d\n", err);
-
+ idpf_send_dealloc_vectors_msg(adapter);
idpf_deinit_vector_stack(adapter);
kfree(adapter->msix_entries);
adapter->msix_entries = NULL;
--
2.40.1
^ permalink raw reply related [flat|nested] 44+ messages in thread
* [Intel-wired-lan] [PATCH v4 10/10 iwl-next] idpf: remove dealloc vector msg err in idpf_intr_rel
@ 2024-02-06 3:38 ` Alan Brady
0 siblings, 0 replies; 44+ messages in thread
From: Alan Brady @ 2024-02-06 3:38 UTC (permalink / raw)
To: intel-wired-lan
Cc: willemdebruijn.kernel, netdev, aleksander.lobakin, Alan Brady,
przemyslaw.kitszel, igor.bagnucki
This error message is at best not really helpful and at worst
misleading. If we're here in idpf_intr_rel we're likely trying to do
remove or reset. If we're in reset, this message will fail because we
lose the virtchnl on reset and HW is going to clean up those resources
regardless in that case. If we're in remove and we get an error here,
we're going to reset the device at the end of remove anyway so not a big
deal. Just remove this message it's not useful.
Signed-off-by: Alan Brady <alan.brady@intel.com>
---
drivers/net/ethernet/intel/idpf/idpf_lib.c | 9 +--------
1 file changed, 1 insertion(+), 8 deletions(-)
diff --git a/drivers/net/ethernet/intel/idpf/idpf_lib.c b/drivers/net/ethernet/intel/idpf/idpf_lib.c
index 1aae6963628b..1e30ef98cd68 100644
--- a/drivers/net/ethernet/intel/idpf/idpf_lib.c
+++ b/drivers/net/ethernet/intel/idpf/idpf_lib.c
@@ -78,19 +78,12 @@ static void idpf_mb_intr_rel_irq(struct idpf_adapter *adapter)
*/
void idpf_intr_rel(struct idpf_adapter *adapter)
{
- int err;
-
if (!adapter->msix_entries)
return;
idpf_mb_intr_rel_irq(adapter);
pci_free_irq_vectors(adapter->pdev);
-
- err = idpf_send_dealloc_vectors_msg(adapter);
- if (err)
- dev_err(&adapter->pdev->dev,
- "Failed to deallocate vectors: %d\n", err);
-
+ idpf_send_dealloc_vectors_msg(adapter);
idpf_deinit_vector_stack(adapter);
kfree(adapter->msix_entries);
adapter->msix_entries = NULL;
--
2.40.1
^ permalink raw reply related [flat|nested] 44+ messages in thread
* Re: [PATCH v4 00/10 iwl-next] idpf: refactor virtchnl messages
2024-02-06 3:37 ` [Intel-wired-lan] " Alan Brady
@ 2024-02-06 17:02 ` Alexander Lobakin
-1 siblings, 0 replies; 44+ messages in thread
From: Alexander Lobakin @ 2024-02-06 17:02 UTC (permalink / raw)
To: Alan Brady
Cc: intel-wired-lan, netdev, willemdebruijn.kernel,
przemyslaw.kitszel, igor.bagnucki
From: Alan Brady <alan.brady@intel.com>
Date: Mon, 5 Feb 2024 19:37:54 -0800
> The motivation for this series has two primary goals. We want to enable
> support of multiple simultaneous messages and make the channel more
> robust. The way it works right now, the driver can only send and receive
> a single message at a time and if something goes really wrong, it can
> lead to data corruption and strange bugs.
This works better than v3.
For the basic scenarios:
Tested-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Thanks,
Olek
^ permalink raw reply [flat|nested] 44+ messages in thread
* Re: [Intel-wired-lan] [PATCH v4 00/10 iwl-next] idpf: refactor virtchnl messages
@ 2024-02-06 17:02 ` Alexander Lobakin
0 siblings, 0 replies; 44+ messages in thread
From: Alexander Lobakin @ 2024-02-06 17:02 UTC (permalink / raw)
To: Alan Brady
Cc: netdev, willemdebruijn.kernel, intel-wired-lan, igor.bagnucki,
przemyslaw.kitszel
From: Alan Brady <alan.brady@intel.com>
Date: Mon, 5 Feb 2024 19:37:54 -0800
> The motivation for this series has two primary goals. We want to enable
> support of multiple simultaneous messages and make the channel more
> robust. The way it works right now, the driver can only send and receive
> a single message at a time and if something goes really wrong, it can
> lead to data corruption and strange bugs.
This works better than v3.
For the basic scenarios:
Tested-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Thanks,
Olek
^ permalink raw reply [flat|nested] 44+ messages in thread
* Re: [PATCH v4 00/10 iwl-next] idpf: refactor virtchnl messages
2024-02-06 3:37 ` [Intel-wired-lan] " Alan Brady
@ 2024-02-06 18:57 ` Jakub Kicinski
-1 siblings, 0 replies; 44+ messages in thread
From: Jakub Kicinski @ 2024-02-06 18:57 UTC (permalink / raw)
To: Alan Brady
Cc: intel-wired-lan, netdev, willemdebruijn.kernel,
przemyslaw.kitszel, igor.bagnucki, aleksander.lobakin
On Mon, 5 Feb 2024 19:37:54 -0800 Alan Brady wrote:
> The motivation for this series has two primary goals. We want to enable
> support of multiple simultaneous messages and make the channel more
> robust. The way it works right now, the driver can only send and receive
> a single message at a time and if something goes really wrong, it can
> lead to data corruption and strange bugs.
Coccinelle points out some potential places to use min()
testing/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c:1956:17-18: WARNING opportunity for min()
testing/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c:1271:17-18: WARNING opportunity for min()
testing/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c:1319:17-18: WARNING opportunity for min()
testing/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c:2640:17-18: WARNING opportunity for min()
testing/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c:1295:17-18: WARNING opportunity for min()
testing/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c:2157:17-18: WARNING opportunity for min()
testing/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c:3582:17-18: WARNING opportunity for min()
--
pw-bot: au
^ permalink raw reply [flat|nested] 44+ messages in thread
* Re: [Intel-wired-lan] [PATCH v4 00/10 iwl-next] idpf: refactor virtchnl messages
@ 2024-02-06 18:57 ` Jakub Kicinski
0 siblings, 0 replies; 44+ messages in thread
From: Jakub Kicinski @ 2024-02-06 18:57 UTC (permalink / raw)
To: Alan Brady
Cc: willemdebruijn.kernel, netdev, aleksander.lobakin,
intel-wired-lan, przemyslaw.kitszel, igor.bagnucki
On Mon, 5 Feb 2024 19:37:54 -0800 Alan Brady wrote:
> The motivation for this series has two primary goals. We want to enable
> support of multiple simultaneous messages and make the channel more
> robust. The way it works right now, the driver can only send and receive
> a single message at a time and if something goes really wrong, it can
> lead to data corruption and strange bugs.
Coccinelle points out some potential places to use min()
testing/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c:1956:17-18: WARNING opportunity for min()
testing/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c:1271:17-18: WARNING opportunity for min()
testing/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c:1319:17-18: WARNING opportunity for min()
testing/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c:2640:17-18: WARNING opportunity for min()
testing/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c:1295:17-18: WARNING opportunity for min()
testing/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c:2157:17-18: WARNING opportunity for min()
testing/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c:3582:17-18: WARNING opportunity for min()
--
pw-bot: au
^ permalink raw reply [flat|nested] 44+ messages in thread
* Re: [PATCH v4 00/10 iwl-next] idpf: refactor virtchnl messages
2024-02-06 18:57 ` [Intel-wired-lan] " Jakub Kicinski
@ 2024-02-06 19:18 ` Alan Brady
-1 siblings, 0 replies; 44+ messages in thread
From: Alan Brady @ 2024-02-06 19:18 UTC (permalink / raw)
To: Jakub Kicinski
Cc: intel-wired-lan, netdev, willemdebruijn.kernel,
przemyslaw.kitszel, igor.bagnucki, aleksander.lobakin
On 2/6/2024 10:57 AM, Jakub Kicinski wrote:
> On Mon, 5 Feb 2024 19:37:54 -0800 Alan Brady wrote:
>> The motivation for this series has two primary goals. We want to enable
>> support of multiple simultaneous messages and make the channel more
>> robust. The way it works right now, the driver can only send and receive
>> a single message at a time and if something goes really wrong, it can
>> lead to data corruption and strange bugs.
>
> Coccinelle points out some potential places to use min()
>
> testing/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c:1956:17-18: WARNING opportunity for min()
> testing/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c:1271:17-18: WARNING opportunity for min()
> testing/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c:1319:17-18: WARNING opportunity for min()
> testing/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c:2640:17-18: WARNING opportunity for min()
> testing/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c:1295:17-18: WARNING opportunity for min()
> testing/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c:2157:17-18: WARNING opportunity for min()
> testing/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c:3582:17-18: WARNING opportunity for min()
We did run coccinelle check and see the min suggestions. It's triggering
on these statements I added:
return reply_sz < 0 ? reply_sz : 0;
A min here would change it to:
return min(reply_sz, 0);
I didn't really like that because it's misleading as though we're
returning the size of the reply and might accidentally encourage someone
to change it to a max. Here reply_sz will be negative if an error was
returned from message sending. But this function we only want to return
0 or negative. By being explicit in what we want to do, it seems clearer
to me what the intention is but I could be wrong.
We can definitely change it however if that's preferred here.
^ permalink raw reply [flat|nested] 44+ messages in thread
* Re: [Intel-wired-lan] [PATCH v4 00/10 iwl-next] idpf: refactor virtchnl messages
@ 2024-02-06 19:18 ` Alan Brady
0 siblings, 0 replies; 44+ messages in thread
From: Alan Brady @ 2024-02-06 19:18 UTC (permalink / raw)
To: Jakub Kicinski
Cc: willemdebruijn.kernel, netdev, aleksander.lobakin,
intel-wired-lan, przemyslaw.kitszel, igor.bagnucki
On 2/6/2024 10:57 AM, Jakub Kicinski wrote:
> On Mon, 5 Feb 2024 19:37:54 -0800 Alan Brady wrote:
>> The motivation for this series has two primary goals. We want to enable
>> support of multiple simultaneous messages and make the channel more
>> robust. The way it works right now, the driver can only send and receive
>> a single message at a time and if something goes really wrong, it can
>> lead to data corruption and strange bugs.
>
> Coccinelle points out some potential places to use min()
>
> testing/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c:1956:17-18: WARNING opportunity for min()
> testing/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c:1271:17-18: WARNING opportunity for min()
> testing/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c:1319:17-18: WARNING opportunity for min()
> testing/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c:2640:17-18: WARNING opportunity for min()
> testing/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c:1295:17-18: WARNING opportunity for min()
> testing/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c:2157:17-18: WARNING opportunity for min()
> testing/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c:3582:17-18: WARNING opportunity for min()
We did run coccinelle check and see the min suggestions. It's triggering
on these statements I added:
return reply_sz < 0 ? reply_sz : 0;
A min here would change it to:
return min(reply_sz, 0);
I didn't really like that because it's misleading as though we're
returning the size of the reply and might accidentally encourage someone
to change it to a max. Here reply_sz will be negative if an error was
returned from message sending. But this function we only want to return
0 or negative. By being explicit in what we want to do, it seems clearer
to me what the intention is but I could be wrong.
We can definitely change it however if that's preferred here.
^ permalink raw reply [flat|nested] 44+ messages in thread
* Re: [Intel-wired-lan] [PATCH v4 00/10 iwl-next] idpf: refactor virtchnl messages
2024-02-06 19:18 ` [Intel-wired-lan] " Alan Brady
@ 2024-02-06 20:03 ` Jakub Kicinski
-1 siblings, 0 replies; 44+ messages in thread
From: Jakub Kicinski @ 2024-02-06 20:03 UTC (permalink / raw)
To: Alan Brady
Cc: willemdebruijn.kernel, netdev, aleksander.lobakin,
intel-wired-lan, przemyslaw.kitszel, igor.bagnucki
On Tue, 6 Feb 2024 11:18:48 -0800 Alan Brady wrote:
> We did run coccinelle check and see the min suggestions. It's triggering
> on these statements I added:
>
> return reply_sz < 0 ? reply_sz : 0;
>
> A min here would change it to:
>
> return min(reply_sz, 0);
>
> I didn't really like that because it's misleading as though we're
> returning the size of the reply and might accidentally encourage someone
> to change it to a max. Here reply_sz will be negative if an error was
> returned from message sending. But this function we only want to return
> 0 or negative. By being explicit in what we want to do, it seems clearer
> to me what the intention is but I could be wrong.
>
> We can definitely change it however if that's preferred here.
Hm, okay, that does sound like making it worse.
I'll disable the minmax coccicheck for now, it seems noisy.
^ permalink raw reply [flat|nested] 44+ messages in thread
* Re: [PATCH v4 00/10 iwl-next] idpf: refactor virtchnl messages
@ 2024-02-06 20:03 ` Jakub Kicinski
0 siblings, 0 replies; 44+ messages in thread
From: Jakub Kicinski @ 2024-02-06 20:03 UTC (permalink / raw)
To: Alan Brady
Cc: intel-wired-lan, netdev, willemdebruijn.kernel,
przemyslaw.kitszel, igor.bagnucki, aleksander.lobakin
On Tue, 6 Feb 2024 11:18:48 -0800 Alan Brady wrote:
> We did run coccinelle check and see the min suggestions. It's triggering
> on these statements I added:
>
> return reply_sz < 0 ? reply_sz : 0;
>
> A min here would change it to:
>
> return min(reply_sz, 0);
>
> I didn't really like that because it's misleading as though we're
> returning the size of the reply and might accidentally encourage someone
> to change it to a max. Here reply_sz will be negative if an error was
> returned from message sending. But this function we only want to return
> 0 or negative. By being explicit in what we want to do, it seems clearer
> to me what the intention is but I could be wrong.
>
> We can definitely change it however if that's preferred here.
Hm, okay, that does sound like making it worse.
I'll disable the minmax coccicheck for now, it seems noisy.
^ permalink raw reply [flat|nested] 44+ messages in thread
* Re: [Intel-wired-lan] [PATCH v4 00/10 iwl-next] idpf: refactor virtchnl messages
2024-02-06 20:03 ` Jakub Kicinski
@ 2024-02-06 22:50 ` Keller, Jacob E
-1 siblings, 0 replies; 44+ messages in thread
From: Keller, Jacob E @ 2024-02-06 22:50 UTC (permalink / raw)
To: Jakub Kicinski, Brady, Alan
Cc: willemdebruijn.kernel, netdev, Lobakin, Aleksander,
intel-wired-lan, Kitszel, Przemyslaw, Bagnucki, Igor
> -----Original Message-----
> From: Jakub Kicinski <kuba@kernel.org>
> Sent: Tuesday, February 6, 2024 12:03 PM
> To: Brady, Alan <alan.brady@intel.com>
> Cc: intel-wired-lan@lists.osuosl.org; netdev@vger.kernel.org;
> willemdebruijn.kernel@gmail.com; Kitszel, Przemyslaw
> <przemyslaw.kitszel@intel.com>; Bagnucki, Igor <igor.bagnucki@intel.com>;
> Lobakin, Aleksander <aleksander.lobakin@intel.com>
> Subject: Re: [PATCH v4 00/10 iwl-next] idpf: refactor virtchnl messages
>
> On Tue, 6 Feb 2024 11:18:48 -0800 Alan Brady wrote:
> > We did run coccinelle check and see the min suggestions. It's triggering
> > on these statements I added:
> >
> > return reply_sz < 0 ? reply_sz : 0;
> >
> > A min here would change it to:
> >
> > return min(reply_sz, 0);
> >
> > I didn't really like that because it's misleading as though we're
> > returning the size of the reply and might accidentally encourage someone
> > to change it to a max. Here reply_sz will be negative if an error was
> > returned from message sending. But this function we only want to return
> > 0 or negative. By being explicit in what we want to do, it seems clearer
> > to me what the intention is but I could be wrong.
> >
> > We can definitely change it however if that's preferred here.
>
> Hm, okay, that does sound like making it worse.
> I'll disable the minmax coccicheck for now, it seems noisy.
Maybe you could make the coccicheck only complain if the value is non-zero or not const? Maybe that's a bit too complicated... Hm
^ permalink raw reply [flat|nested] 44+ messages in thread
* RE: [PATCH v4 00/10 iwl-next] idpf: refactor virtchnl messages
@ 2024-02-06 22:50 ` Keller, Jacob E
0 siblings, 0 replies; 44+ messages in thread
From: Keller, Jacob E @ 2024-02-06 22:50 UTC (permalink / raw)
To: Jakub Kicinski, Brady, Alan
Cc: intel-wired-lan, netdev, willemdebruijn.kernel, Kitszel,
Przemyslaw, Bagnucki, Igor, Lobakin, Aleksander
> -----Original Message-----
> From: Jakub Kicinski <kuba@kernel.org>
> Sent: Tuesday, February 6, 2024 12:03 PM
> To: Brady, Alan <alan.brady@intel.com>
> Cc: intel-wired-lan@lists.osuosl.org; netdev@vger.kernel.org;
> willemdebruijn.kernel@gmail.com; Kitszel, Przemyslaw
> <przemyslaw.kitszel@intel.com>; Bagnucki, Igor <igor.bagnucki@intel.com>;
> Lobakin, Aleksander <aleksander.lobakin@intel.com>
> Subject: Re: [PATCH v4 00/10 iwl-next] idpf: refactor virtchnl messages
>
> On Tue, 6 Feb 2024 11:18:48 -0800 Alan Brady wrote:
> > We did run coccinelle check and see the min suggestions. It's triggering
> > on these statements I added:
> >
> > return reply_sz < 0 ? reply_sz : 0;
> >
> > A min here would change it to:
> >
> > return min(reply_sz, 0);
> >
> > I didn't really like that because it's misleading as though we're
> > returning the size of the reply and might accidentally encourage someone
> > to change it to a max. Here reply_sz will be negative if an error was
> > returned from message sending. But this function we only want to return
> > 0 or negative. By being explicit in what we want to do, it seems clearer
> > to me what the intention is but I could be wrong.
> >
> > We can definitely change it however if that's preferred here.
>
> Hm, okay, that does sound like making it worse.
> I'll disable the minmax coccicheck for now, it seems noisy.
Maybe you could make the coccicheck only complain if the value is non-zero or not const? Maybe that's a bit too complicated... Hm
^ permalink raw reply [flat|nested] 44+ messages in thread
* Re: [Intel-wired-lan] [PATCH v4 00/10 iwl-next] idpf: refactor virtchnl messages
2024-02-06 22:50 ` Keller, Jacob E
@ 2024-02-06 23:17 ` Jakub Kicinski
-1 siblings, 0 replies; 44+ messages in thread
From: Jakub Kicinski @ 2024-02-06 23:17 UTC (permalink / raw)
To: Keller, Jacob E
Cc: willemdebruijn.kernel, netdev, Lobakin, Aleksander,
intel-wired-lan, Brady, Alan, Kitszel, Przemyslaw, Bagnucki,
Igor
On Tue, 6 Feb 2024 22:50:25 +0000 Keller, Jacob E wrote:
> > Hm, okay, that does sound like making it worse.
> > I'll disable the minmax coccicheck for now, it seems noisy.
>
> Maybe you could make the coccicheck only complain if the value is
> non-zero or not const? Maybe that's a bit too complicated... Hm
Non-zero could work. It may be worthwhile to look at the warnings cocci
currently generates and figure out what's common among the cases where
warning doesn't make sense.
^ permalink raw reply [flat|nested] 44+ messages in thread
* Re: [PATCH v4 00/10 iwl-next] idpf: refactor virtchnl messages
@ 2024-02-06 23:17 ` Jakub Kicinski
0 siblings, 0 replies; 44+ messages in thread
From: Jakub Kicinski @ 2024-02-06 23:17 UTC (permalink / raw)
To: Keller, Jacob E
Cc: Brady, Alan, intel-wired-lan, netdev, willemdebruijn.kernel,
Kitszel, Przemyslaw, Bagnucki, Igor, Lobakin, Aleksander
On Tue, 6 Feb 2024 22:50:25 +0000 Keller, Jacob E wrote:
> > Hm, okay, that does sound like making it worse.
> > I'll disable the minmax coccicheck for now, it seems noisy.
>
> Maybe you could make the coccicheck only complain if the value is
> non-zero or not const? Maybe that's a bit too complicated... Hm
Non-zero could work. It may be worthwhile to look at the warnings cocci
currently generates and figure out what's common among the cases where
warning doesn't make sense.
^ permalink raw reply [flat|nested] 44+ messages in thread
* Re: [Intel-wired-lan] [PATCH v4 00/10 iwl-next] idpf: refactor virtchnl messages
2024-02-06 17:02 ` [Intel-wired-lan] " Alexander Lobakin
@ 2024-02-14 14:49 ` Alexander Lobakin
-1 siblings, 0 replies; 44+ messages in thread
From: Alexander Lobakin @ 2024-02-14 14:49 UTC (permalink / raw)
To: Alan Brady
Cc: netdev, willemdebruijn.kernel, intel-wired-lan, igor.bagnucki,
przemyslaw.kitszel
From: Alexander Lobakin <aleksander.lobakin@intel.com>
Date: Tue, 6 Feb 2024 18:02:33 +0100
> From: Alan Brady <alan.brady@intel.com>
> Date: Mon, 5 Feb 2024 19:37:54 -0800
>
>> The motivation for this series has two primary goals. We want to enable
>> support of multiple simultaneous messages and make the channel more
>> robust. The way it works right now, the driver can only send and receive
>> a single message at a time and if something goes really wrong, it can
>> lead to data corruption and strange bugs.
>
> This works better than v3.
> For the basic scenarios:
>
> Tested-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Sorry I'm reverting my tag.
After the series, the CP segfaults on each rmmod idpf:
root@mev-imc:/usr/bin/cplane# cp_pim_mdd_handler: MDD interrupt detected
cp_pim_mdd_event_handler: reg = 40
Jan 1 00:27:57 mev-imc local0.warn LanCP[190]: cp_pim_mdd_handler: MDD
interrupt detected
cp_pim_mdd_event_handler: reg = 1
Jan 1 00:28:54 mev-imc local0.warn LanCP[190]: [hma_create_vport/4257]
WARNING: RSS is configured on 1st contiguous num of queuJan 1 00:28:54
mev-imc local0.warn LanCP[190]: [hma_create_vport/4257] WARNING: RSS is
configured on 1st contiguous num of queuJan 1 00:28:55 mev-imc
local0.warn LanCP[190]: [hma_create_vport/4257] WARNING: RSS is
configured on 1st contiguous num of queues= 16 start Qid= 34
Jan 1 00:28:55 mev-imc local0.warn LanCP[190]: [hma_create_vport/4257]
WARNING: RSS is configured on 1st contiguous num of queu16 start Qid= 640
Jan 1 00:28:55 mev-imc local0.err LanCP[190]:
[cp_del_node_rxbuff_lst/4179] ERR: Resource list is empty, so nothing to
delete here
Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
[cp_uninit_vsi_tc_q_region/222] ERR: Failed to init vsi LUT on vsi 1.
Jan 1 00::08 mev-imc local0.err LanCP[190]:
[cp_uninit_vsi_fxp_config/1101] ERR: cp_uninit_vsi_rss_config() failed
on vsi (1).
Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
[cp_uninit_vsi_tc_q_region/222] ERR: Failed to init vsi LUT on vsi 6.
Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
[cp_uninit_vsi_rss_config/340] ERR: faininit_vsi_rss_config() failed on
vsi (6).
Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
[cp_uninit_vsi_tc_q_region/222] ERR: Failed to init vsi LUT on vsi 7.
Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
[cp_uninit_vsi_rss_config/340] ERR: failed to remove vsi (7)'s queue
regions.
Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
[cp_uninit_vsi_fxp_config/1101] ERR: cp_uninit_vo init vsi LUT on vsi 8.
Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
[cp_uninit_vsi_rss_config/340] ERR: failed to remove vsi (8)'s queue
regions.
Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
[cp_uninit_vsi_fxp_config/1101] ERR: cp_uninit_vsi_rss_config() failed
on vsi (8).
Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
[cp_uninit_vsi_tc_q_region/222] ERR: Failed to init vsi LUT on vsi 1.
Jan 1 00::08 mev-imc local0.err LanCP[190]:
[cp_uninit_vsi_fxp_config/1101] ERR: cp_uninit_vsi_rss_config() failed
on vsi (1).
[1]+ Segmentation fault ./imccp 0000:00:01.6 0 cp_init.cfg
Only restarting the CP helps -- restarting the imccp daemon makes it
immediately crash again.
This should be dropped from the next-queue until it's fixed.
Thanks,
Olek
^ permalink raw reply [flat|nested] 44+ messages in thread
* Re: [PATCH v4 00/10 iwl-next] idpf: refactor virtchnl messages
@ 2024-02-14 14:49 ` Alexander Lobakin
0 siblings, 0 replies; 44+ messages in thread
From: Alexander Lobakin @ 2024-02-14 14:49 UTC (permalink / raw)
To: Alan Brady
Cc: intel-wired-lan, netdev, willemdebruijn.kernel,
przemyslaw.kitszel, igor.bagnucki
From: Alexander Lobakin <aleksander.lobakin@intel.com>
Date: Tue, 6 Feb 2024 18:02:33 +0100
> From: Alan Brady <alan.brady@intel.com>
> Date: Mon, 5 Feb 2024 19:37:54 -0800
>
>> The motivation for this series has two primary goals. We want to enable
>> support of multiple simultaneous messages and make the channel more
>> robust. The way it works right now, the driver can only send and receive
>> a single message at a time and if something goes really wrong, it can
>> lead to data corruption and strange bugs.
>
> This works better than v3.
> For the basic scenarios:
>
> Tested-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Sorry I'm reverting my tag.
After the series, the CP segfaults on each rmmod idpf:
root@mev-imc:/usr/bin/cplane# cp_pim_mdd_handler: MDD interrupt detected
cp_pim_mdd_event_handler: reg = 40
Jan 1 00:27:57 mev-imc local0.warn LanCP[190]: cp_pim_mdd_handler: MDD
interrupt detected
cp_pim_mdd_event_handler: reg = 1
Jan 1 00:28:54 mev-imc local0.warn LanCP[190]: [hma_create_vport/4257]
WARNING: RSS is configured on 1st contiguous num of queuJan 1 00:28:54
mev-imc local0.warn LanCP[190]: [hma_create_vport/4257] WARNING: RSS is
configured on 1st contiguous num of queuJan 1 00:28:55 mev-imc
local0.warn LanCP[190]: [hma_create_vport/4257] WARNING: RSS is
configured on 1st contiguous num of queues= 16 start Qid= 34
Jan 1 00:28:55 mev-imc local0.warn LanCP[190]: [hma_create_vport/4257]
WARNING: RSS is configured on 1st contiguous num of queu16 start Qid= 640
Jan 1 00:28:55 mev-imc local0.err LanCP[190]:
[cp_del_node_rxbuff_lst/4179] ERR: Resource list is empty, so nothing to
delete here
Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
[cp_uninit_vsi_tc_q_region/222] ERR: Failed to init vsi LUT on vsi 1.
Jan 1 00::08 mev-imc local0.err LanCP[190]:
[cp_uninit_vsi_fxp_config/1101] ERR: cp_uninit_vsi_rss_config() failed
on vsi (1).
Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
[cp_uninit_vsi_tc_q_region/222] ERR: Failed to init vsi LUT on vsi 6.
Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
[cp_uninit_vsi_rss_config/340] ERR: faininit_vsi_rss_config() failed on
vsi (6).
Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
[cp_uninit_vsi_tc_q_region/222] ERR: Failed to init vsi LUT on vsi 7.
Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
[cp_uninit_vsi_rss_config/340] ERR: failed to remove vsi (7)'s queue
regions.
Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
[cp_uninit_vsi_fxp_config/1101] ERR: cp_uninit_vo init vsi LUT on vsi 8.
Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
[cp_uninit_vsi_rss_config/340] ERR: failed to remove vsi (8)'s queue
regions.
Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
[cp_uninit_vsi_fxp_config/1101] ERR: cp_uninit_vsi_rss_config() failed
on vsi (8).
Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
[cp_uninit_vsi_tc_q_region/222] ERR: Failed to init vsi LUT on vsi 1.
Jan 1 00::08 mev-imc local0.err LanCP[190]:
[cp_uninit_vsi_fxp_config/1101] ERR: cp_uninit_vsi_rss_config() failed
on vsi (1).
[1]+ Segmentation fault ./imccp 0000:00:01.6 0 cp_init.cfg
Only restarting the CP helps -- restarting the imccp daemon makes it
immediately crash again.
This should be dropped from the next-queue until it's fixed.
Thanks,
Olek
^ permalink raw reply [flat|nested] 44+ messages in thread
* Re: [PATCH v4 00/10 iwl-next] idpf: refactor virtchnl messages
2024-02-14 14:49 ` Alexander Lobakin
@ 2024-02-14 17:06 ` Alan Brady
-1 siblings, 0 replies; 44+ messages in thread
From: Alan Brady @ 2024-02-14 17:06 UTC (permalink / raw)
To: Alexander Lobakin
Cc: intel-wired-lan, netdev, willemdebruijn.kernel,
przemyslaw.kitszel, igor.bagnucki
On 2/14/2024 6:49 AM, Alexander Lobakin wrote:
> From: Alexander Lobakin <aleksander.lobakin@intel.com>
> Date: Tue, 6 Feb 2024 18:02:33 +0100
>
>> From: Alan Brady <alan.brady@intel.com>
>> Date: Mon, 5 Feb 2024 19:37:54 -0800
>>
>>> The motivation for this series has two primary goals. We want to enable
>>> support of multiple simultaneous messages and make the channel more
>>> robust. The way it works right now, the driver can only send and receive
>>> a single message at a time and if something goes really wrong, it can
>>> lead to data corruption and strange bugs.
>>
>> This works better than v3.
>> For the basic scenarios:
>>
>> Tested-by: Alexander Lobakin <aleksander.lobakin@intel.com>
>
> Sorry I'm reverting my tag.
> After the series, the CP segfaults on each rmmod idpf:
>
> root@mev-imc:/usr/bin/cplane# cp_pim_mdd_handler: MDD interrupt detected
> cp_pim_mdd_event_handler: reg = 40
> Jan 1 00:27:57 mev-imc local0.warn LanCP[190]: cp_pim_mdd_handler: MDD
> interrupt detected
> cp_pim_mdd_event_handler: reg = 1
> Jan 1 00:28:54 mev-imc local0.warn LanCP[190]: [hma_create_vport/4257]
> WARNING: RSS is configured on 1st contiguous num of queuJan 1 00:28:54
> mev-imc local0.warn LanCP[190]: [hma_create_vport/4257] WARNING: RSS is
> configured on 1st contiguous num of queuJan 1 00:28:55 mev-imc
> local0.warn LanCP[190]: [hma_create_vport/4257] WARNING: RSS is
> configured on 1st contiguous num of queues= 16 start Qid= 34
> Jan 1 00:28:55 mev-imc local0.warn LanCP[190]: [hma_create_vport/4257]
> WARNING: RSS is configured on 1st contiguous num of queu16 start Qid= 640
> Jan 1 00:28:55 mev-imc local0.err LanCP[190]:
> [cp_del_node_rxbuff_lst/4179] ERR: Resource list is empty, so nothing to
> delete here
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_tc_q_region/222] ERR: Failed to init vsi LUT on vsi 1.
> Jan 1 00::08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_fxp_config/1101] ERR: cp_uninit_vsi_rss_config() failed
> on vsi (1).
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_tc_q_region/222] ERR: Failed to init vsi LUT on vsi 6.
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_rss_config/340] ERR: faininit_vsi_rss_config() failed on
> vsi (6).
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_tc_q_region/222] ERR: Failed to init vsi LUT on vsi 7.
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_rss_config/340] ERR: failed to remove vsi (7)'s queue
> regions.
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_fxp_config/1101] ERR: cp_uninit_vo init vsi LUT on vsi 8.
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_rss_config/340] ERR: failed to remove vsi (8)'s queue
> regions.
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_fxp_config/1101] ERR: cp_uninit_vsi_rss_config() failed
> on vsi (8).
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_tc_q_region/222] ERR: Failed to init vsi LUT on vsi 1.
> Jan 1 00::08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_fxp_config/1101] ERR: cp_uninit_vsi_rss_config() failed
> on vsi (1).
>
> [1]+ Segmentation fault ./imccp 0000:00:01.6 0 cp_init.cfg
>
> Only restarting the CP helps -- restarting the imccp daemon makes it
> immediately crash again.
>
> This should be dropped from the next-queue until it's fixed.
>
> Thanks,
> Olek
I definitely tested rmmod so I'm frankly not understanding how this can
be possible. If you would like I'm happy to sync on how you're able to
cause this to happen. Our internal validation verified 1000 load/unloads
passed their testing.
Is there more you're doing other than just rmmod?
Alan
^ permalink raw reply [flat|nested] 44+ messages in thread
* Re: [Intel-wired-lan] [PATCH v4 00/10 iwl-next] idpf: refactor virtchnl messages
@ 2024-02-14 17:06 ` Alan Brady
0 siblings, 0 replies; 44+ messages in thread
From: Alan Brady @ 2024-02-14 17:06 UTC (permalink / raw)
To: Alexander Lobakin
Cc: netdev, willemdebruijn.kernel, intel-wired-lan, igor.bagnucki,
przemyslaw.kitszel
On 2/14/2024 6:49 AM, Alexander Lobakin wrote:
> From: Alexander Lobakin <aleksander.lobakin@intel.com>
> Date: Tue, 6 Feb 2024 18:02:33 +0100
>
>> From: Alan Brady <alan.brady@intel.com>
>> Date: Mon, 5 Feb 2024 19:37:54 -0800
>>
>>> The motivation for this series has two primary goals. We want to enable
>>> support of multiple simultaneous messages and make the channel more
>>> robust. The way it works right now, the driver can only send and receive
>>> a single message at a time and if something goes really wrong, it can
>>> lead to data corruption and strange bugs.
>>
>> This works better than v3.
>> For the basic scenarios:
>>
>> Tested-by: Alexander Lobakin <aleksander.lobakin@intel.com>
>
> Sorry I'm reverting my tag.
> After the series, the CP segfaults on each rmmod idpf:
>
> root@mev-imc:/usr/bin/cplane# cp_pim_mdd_handler: MDD interrupt detected
> cp_pim_mdd_event_handler: reg = 40
> Jan 1 00:27:57 mev-imc local0.warn LanCP[190]: cp_pim_mdd_handler: MDD
> interrupt detected
> cp_pim_mdd_event_handler: reg = 1
> Jan 1 00:28:54 mev-imc local0.warn LanCP[190]: [hma_create_vport/4257]
> WARNING: RSS is configured on 1st contiguous num of queuJan 1 00:28:54
> mev-imc local0.warn LanCP[190]: [hma_create_vport/4257] WARNING: RSS is
> configured on 1st contiguous num of queuJan 1 00:28:55 mev-imc
> local0.warn LanCP[190]: [hma_create_vport/4257] WARNING: RSS is
> configured on 1st contiguous num of queues= 16 start Qid= 34
> Jan 1 00:28:55 mev-imc local0.warn LanCP[190]: [hma_create_vport/4257]
> WARNING: RSS is configured on 1st contiguous num of queu16 start Qid= 640
> Jan 1 00:28:55 mev-imc local0.err LanCP[190]:
> [cp_del_node_rxbuff_lst/4179] ERR: Resource list is empty, so nothing to
> delete here
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_tc_q_region/222] ERR: Failed to init vsi LUT on vsi 1.
> Jan 1 00::08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_fxp_config/1101] ERR: cp_uninit_vsi_rss_config() failed
> on vsi (1).
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_tc_q_region/222] ERR: Failed to init vsi LUT on vsi 6.
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_rss_config/340] ERR: faininit_vsi_rss_config() failed on
> vsi (6).
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_tc_q_region/222] ERR: Failed to init vsi LUT on vsi 7.
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_rss_config/340] ERR: failed to remove vsi (7)'s queue
> regions.
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_fxp_config/1101] ERR: cp_uninit_vo init vsi LUT on vsi 8.
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_rss_config/340] ERR: failed to remove vsi (8)'s queue
> regions.
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_fxp_config/1101] ERR: cp_uninit_vsi_rss_config() failed
> on vsi (8).
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_tc_q_region/222] ERR: Failed to init vsi LUT on vsi 1.
> Jan 1 00::08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_fxp_config/1101] ERR: cp_uninit_vsi_rss_config() failed
> on vsi (1).
>
> [1]+ Segmentation fault ./imccp 0000:00:01.6 0 cp_init.cfg
>
> Only restarting the CP helps -- restarting the imccp daemon makes it
> immediately crash again.
>
> This should be dropped from the next-queue until it's fixed.
>
> Thanks,
> Olek
I definitely tested rmmod so I'm frankly not understanding how this can
be possible. If you would like I'm happy to sync on how you're able to
cause this to happen. Our internal validation verified 1000 load/unloads
passed their testing.
Is there more you're doing other than just rmmod?
Alan
^ permalink raw reply [flat|nested] 44+ messages in thread
* Re: [Intel-wired-lan] [PATCH v4 00/10 iwl-next] idpf: refactor virtchnl messages
2024-02-14 14:49 ` Alexander Lobakin
@ 2024-02-20 13:47 ` Alexander Lobakin
-1 siblings, 0 replies; 44+ messages in thread
From: Alexander Lobakin @ 2024-02-20 13:47 UTC (permalink / raw)
To: Alan Brady
Cc: netdev, willemdebruijn.kernel, intel-wired-lan, igor.bagnucki,
przemyslaw.kitszel
From: Alexander Lobakin <aleksander.lobakin@intel.com>
Date: Wed, 14 Feb 2024 15:49:51 +0100
> From: Alexander Lobakin <aleksander.lobakin@intel.com>
> Date: Tue, 6 Feb 2024 18:02:33 +0100
>
>> From: Alan Brady <alan.brady@intel.com>
>> Date: Mon, 5 Feb 2024 19:37:54 -0800
>>
>>> The motivation for this series has two primary goals. We want to enable
>>> support of multiple simultaneous messages and make the channel more
>>> robust. The way it works right now, the driver can only send and receive
>>> a single message at a time and if something goes really wrong, it can
>>> lead to data corruption and strange bugs.
>>
>> This works better than v3.
>> For the basic scenarios:
>>
>> Tested-by: Alexander Lobakin <aleksander.lobakin@intel.com>
>
> Sorry I'm reverting my tag.
> After the series, the CP segfaults on each rmmod idpf:
>
> root@mev-imc:/usr/bin/cplane# cp_pim_mdd_handler: MDD interrupt detected
> cp_pim_mdd_event_handler: reg = 40
> Jan 1 00:27:57 mev-imc local0.warn LanCP[190]: cp_pim_mdd_handler: MDD
> interrupt detected
> cp_pim_mdd_event_handler: reg = 1
> Jan 1 00:28:54 mev-imc local0.warn LanCP[190]: [hma_create_vport/4257]
> WARNING: RSS is configured on 1st contiguous num of queuJan 1 00:28:54
> mev-imc local0.warn LanCP[190]: [hma_create_vport/4257] WARNING: RSS is
> configured on 1st contiguous num of queuJan 1 00:28:55 mev-imc
> local0.warn LanCP[190]: [hma_create_vport/4257] WARNING: RSS is
> configured on 1st contiguous num of queues= 16 start Qid= 34
> Jan 1 00:28:55 mev-imc local0.warn LanCP[190]: [hma_create_vport/4257]
> WARNING: RSS is configured on 1st contiguous num of queu16 start Qid= 640
> Jan 1 00:28:55 mev-imc local0.err LanCP[190]:
> [cp_del_node_rxbuff_lst/4179] ERR: Resource list is empty, so nothing to
> delete here
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_tc_q_region/222] ERR: Failed to init vsi LUT on vsi 1.
> Jan 1 00::08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_fxp_config/1101] ERR: cp_uninit_vsi_rss_config() failed
> on vsi (1).
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_tc_q_region/222] ERR: Failed to init vsi LUT on vsi 6.
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_rss_config/340] ERR: faininit_vsi_rss_config() failed on
> vsi (6).
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_tc_q_region/222] ERR: Failed to init vsi LUT on vsi 7.
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_rss_config/340] ERR: failed to remove vsi (7)'s queue
> regions.
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_fxp_config/1101] ERR: cp_uninit_vo init vsi LUT on vsi 8.
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_rss_config/340] ERR: failed to remove vsi (8)'s queue
> regions.
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_fxp_config/1101] ERR: cp_uninit_vsi_rss_config() failed
> on vsi (8).
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_tc_q_region/222] ERR: Failed to init vsi LUT on vsi 1.
> Jan 1 00::08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_fxp_config/1101] ERR: cp_uninit_vsi_rss_config() failed
> on vsi (1).
>
> [1]+ Segmentation fault ./imccp 0000:00:01.6 0 cp_init.cfg
>
> Only restarting the CP helps -- restarting the imccp daemon makes it
> immediately crash again.
>
> This should be dropped from the next-queue until it's fixed.
The latest firmware works with this series -- the problem was there,
the series only revealed it.
Tested-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Can be taken back to Tony's tree.
Thanks,
Olek
^ permalink raw reply [flat|nested] 44+ messages in thread
* Re: [Intel-wired-lan] [PATCH v4 00/10 iwl-next] idpf: refactor virtchnl messages
@ 2024-02-20 13:47 ` Alexander Lobakin
0 siblings, 0 replies; 44+ messages in thread
From: Alexander Lobakin @ 2024-02-20 13:47 UTC (permalink / raw)
To: Alan Brady
Cc: netdev, willemdebruijn.kernel, intel-wired-lan, igor.bagnucki,
przemyslaw.kitszel
From: Alexander Lobakin <aleksander.lobakin@intel.com>
Date: Wed, 14 Feb 2024 15:49:51 +0100
> From: Alexander Lobakin <aleksander.lobakin@intel.com>
> Date: Tue, 6 Feb 2024 18:02:33 +0100
>
>> From: Alan Brady <alan.brady@intel.com>
>> Date: Mon, 5 Feb 2024 19:37:54 -0800
>>
>>> The motivation for this series has two primary goals. We want to enable
>>> support of multiple simultaneous messages and make the channel more
>>> robust. The way it works right now, the driver can only send and receive
>>> a single message at a time and if something goes really wrong, it can
>>> lead to data corruption and strange bugs.
>>
>> This works better than v3.
>> For the basic scenarios:
>>
>> Tested-by: Alexander Lobakin <aleksander.lobakin@intel.com>
>
> Sorry I'm reverting my tag.
> After the series, the CP segfaults on each rmmod idpf:
>
> root@mev-imc:/usr/bin/cplane# cp_pim_mdd_handler: MDD interrupt detected
> cp_pim_mdd_event_handler: reg = 40
> Jan 1 00:27:57 mev-imc local0.warn LanCP[190]: cp_pim_mdd_handler: MDD
> interrupt detected
> cp_pim_mdd_event_handler: reg = 1
> Jan 1 00:28:54 mev-imc local0.warn LanCP[190]: [hma_create_vport/4257]
> WARNING: RSS is configured on 1st contiguous num of queuJan 1 00:28:54
> mev-imc local0.warn LanCP[190]: [hma_create_vport/4257] WARNING: RSS is
> configured on 1st contiguous num of queuJan 1 00:28:55 mev-imc
> local0.warn LanCP[190]: [hma_create_vport/4257] WARNING: RSS is
> configured on 1st contiguous num of queues= 16 start Qid= 34
> Jan 1 00:28:55 mev-imc local0.warn LanCP[190]: [hma_create_vport/4257]
> WARNING: RSS is configured on 1st contiguous num of queu16 start Qid= 640
> Jan 1 00:28:55 mev-imc local0.err LanCP[190]:
> [cp_del_node_rxbuff_lst/4179] ERR: Resource list is empty, so nothing to
> delete here
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_tc_q_region/222] ERR: Failed to init vsi LUT on vsi 1.
> Jan 1 00::08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_fxp_config/1101] ERR: cp_uninit_vsi_rss_config() failed
> on vsi (1).
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_tc_q_region/222] ERR: Failed to init vsi LUT on vsi 6.
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_rss_config/340] ERR: faininit_vsi_rss_config() failed on
> vsi (6).
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_tc_q_region/222] ERR: Failed to init vsi LUT on vsi 7.
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_rss_config/340] ERR: failed to remove vsi (7)'s queue
> regions.
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_fxp_config/1101] ERR: cp_uninit_vo init vsi LUT on vsi 8.
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_rss_config/340] ERR: failed to remove vsi (8)'s queue
> regions.
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_fxp_config/1101] ERR: cp_uninit_vsi_rss_config() failed
> on vsi (8).
> Jan 1 00:29:08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_tc_q_region/222] ERR: Failed to init vsi LUT on vsi 1.
> Jan 1 00::08 mev-imc local0.err LanCP[190]:
> [cp_uninit_vsi_fxp_config/1101] ERR: cp_uninit_vsi_rss_config() failed
> on vsi (1).
>
> [1]+ Segmentation fault ./imccp 0000:00:01.6 0 cp_init.cfg
>
> Only restarting the CP helps -- restarting the imccp daemon makes it
> immediately crash again.
>
> This should be dropped from the next-queue until it's fixed.
The latest firmware works with this series -- the problem was there,
the series only revealed it.
Tested-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Can be taken back to Tony's tree.
Thanks,
Olek
^ permalink raw reply [flat|nested] 44+ messages in thread
* Re: [Intel-wired-lan] [PATCH v4 01/10 iwl-next] idpf: implement virtchnl transaction manager
2024-02-06 3:37 ` [Intel-wired-lan] " Alan Brady
@ 2024-02-20 14:30 ` Alexander Lobakin
-1 siblings, 0 replies; 44+ messages in thread
From: Alexander Lobakin @ 2024-02-20 14:30 UTC (permalink / raw)
To: Alan Brady
Cc: willemdebruijn.kernel, netdev, Joshua Hay, intel-wired-lan,
przemyslaw.kitszel, igor.bagnucki
From: Alan Brady <alan.brady@intel.com>
Date: Mon, 5 Feb 2024 19:37:55 -0800
> This starts refactoring how virtchnl messages are handled by adding a
> transaction manager (idpf_vc_xn_manager).
[...]
> +/**
> + * idpf_vc_xn_exec - Perform a send/recv virtchnl transaction
> + * @adapter: driver specific private structure with vcxn_mngr
> + * @params: parameters for this particular transaction including
> + * -vc_op: virtchannel operation to send
> + * -send_buf: kvec iov for send buf and len
> + * -recv_buf: kvec iov for recv buf and len (ignored if NULL)
> + * -timeout_ms: timeout waiting for a reply (milliseconds)
> + * -async: don't wait for message reply, will lose caller context
> + * -async_handler: callback to handle async replies
> + *
> + * @returns >= 0 for success, the size of the initial reply (may or may not be
> + * >= @recv_buf.iov_len, but we never overflow @@recv_buf_iov_base). < 0 for
> + * error.
> + */
> +static ssize_t idpf_vc_xn_exec(struct idpf_adapter *adapter,
> + struct idpf_vc_xn_params params)
Why do you pass @params by value, i.e. whole 56 bytes per each function
call instead of passing it by pointer -> 8 bytes per call?
> +{
> + struct kvec *send_buf = ¶ms.send_buf;
> + struct idpf_vc_xn *xn;
> + ssize_t retval;
> + u16 cookie;
> +
> + xn = idpf_vc_xn_pop_free(&adapter->vcxn_mngr);
> + /* no free transactions available */
> + if (!xn)
> + return -ENOSPC;
> +
> + idpf_vc_xn_lock(xn);
> + if (xn->state == IDPF_VC_XN_SHUTDOWN) {
> + retval = -ENXIO;
> + goto only_unlock;
> + } else if (xn->state != IDPF_VC_XN_IDLE) {
> + /* We're just going to clobber this transaction even though
> + * it's not IDLE. If we don't reuse it we could theoretically
> + * eventually leak all the free transactions and not be able to
> + * send any messages. At least this way we make an attempt to
> + * remain functional even though something really bad is
> + * happening that's corrupting what was supposed to be free
> + * transactions.
> + */
> + WARN_ONCE(1, "There should only be idle transactions in free list (idx %d op %d)\n",
> + xn->idx, xn->vc_op);
> + }
> +
> + xn->reply = params.recv_buf;
> + xn->reply_sz = 0;
> + xn->state = params.async ? IDPF_VC_XN_ASYNC : IDPF_VC_XN_WAITING;
> + xn->vc_op = params.vc_op;
> + xn->async_handler = params.async_handler;
> + idpf_vc_xn_unlock(xn);
> +
> + if (!params.async)
> + reinit_completion(&xn->completed);
> + cookie = FIELD_PREP(IDPF_VC_XN_SALT_M, xn->salt) |
> + FIELD_PREP(IDPF_VC_XN_IDX_M, xn->idx);
> +
> + retval = idpf_send_mb_msg(adapter, params.vc_op,
> + send_buf->iov_len, send_buf->iov_base,
> + cookie);
> + if (retval) {
> + idpf_vc_xn_lock(xn);
> + goto release_and_unlock;
> + }
> +
> + if (params.async)
> + return 0;
> +
> + wait_for_completion_timeout(&xn->completed,
> + msecs_to_jiffies(params.timeout_ms));
> +
> + /* No need to check the return value; we check the final state of the
> + * transaction below. It's possible the transaction actually gets more
> + * timeout than specified if we get preempted here but after
> + * wait_for_completion_timeout returns. This should be non-issue
> + * however.
> + */
> + idpf_vc_xn_lock(xn);
> + switch (xn->state) {
> + case IDPF_VC_XN_SHUTDOWN:
> + retval = -ENXIO;
> + goto only_unlock;
> + case IDPF_VC_XN_WAITING:
> + dev_notice_ratelimited(&adapter->pdev->dev, "Transaction timed-out (op %d, %dms)\n",
> + params.vc_op, params.timeout_ms);
> + retval = -ETIME;
> + break;
> + case IDPF_VC_XN_COMPLETED_SUCCESS:
> + retval = xn->reply_sz;
> + break;
> + case IDPF_VC_XN_COMPLETED_FAILED:
> + dev_notice_ratelimited(&adapter->pdev->dev, "Transaction failed (op %d)\n",
> + params.vc_op);
> + retval = -EIO;
> + break;
> + default:
> + /* Invalid state. */
> + WARN_ON_ONCE(1);
> + retval = -EIO;
> + break;
> + }
> +
> +release_and_unlock:
> + idpf_vc_xn_push_free(&adapter->vcxn_mngr, xn);
> + /* If we receive a VC reply after here, it will be dropped. */
> +only_unlock:
> + idpf_vc_xn_unlock(xn);
> +
> + return retval;
> +}
[...]
Thanks,
Olek
^ permalink raw reply [flat|nested] 44+ messages in thread
* Re: [PATCH v4 01/10 iwl-next] idpf: implement virtchnl transaction manager
@ 2024-02-20 14:30 ` Alexander Lobakin
0 siblings, 0 replies; 44+ messages in thread
From: Alexander Lobakin @ 2024-02-20 14:30 UTC (permalink / raw)
To: Alan Brady
Cc: intel-wired-lan, netdev, willemdebruijn.kernel,
przemyslaw.kitszel, igor.bagnucki, Joshua Hay
From: Alan Brady <alan.brady@intel.com>
Date: Mon, 5 Feb 2024 19:37:55 -0800
> This starts refactoring how virtchnl messages are handled by adding a
> transaction manager (idpf_vc_xn_manager).
[...]
> +/**
> + * idpf_vc_xn_exec - Perform a send/recv virtchnl transaction
> + * @adapter: driver specific private structure with vcxn_mngr
> + * @params: parameters for this particular transaction including
> + * -vc_op: virtchannel operation to send
> + * -send_buf: kvec iov for send buf and len
> + * -recv_buf: kvec iov for recv buf and len (ignored if NULL)
> + * -timeout_ms: timeout waiting for a reply (milliseconds)
> + * -async: don't wait for message reply, will lose caller context
> + * -async_handler: callback to handle async replies
> + *
> + * @returns >= 0 for success, the size of the initial reply (may or may not be
> + * >= @recv_buf.iov_len, but we never overflow @@recv_buf_iov_base). < 0 for
> + * error.
> + */
> +static ssize_t idpf_vc_xn_exec(struct idpf_adapter *adapter,
> + struct idpf_vc_xn_params params)
Why do you pass @params by value, i.e. whole 56 bytes per each function
call instead of passing it by pointer -> 8 bytes per call?
> +{
> + struct kvec *send_buf = ¶ms.send_buf;
> + struct idpf_vc_xn *xn;
> + ssize_t retval;
> + u16 cookie;
> +
> + xn = idpf_vc_xn_pop_free(&adapter->vcxn_mngr);
> + /* no free transactions available */
> + if (!xn)
> + return -ENOSPC;
> +
> + idpf_vc_xn_lock(xn);
> + if (xn->state == IDPF_VC_XN_SHUTDOWN) {
> + retval = -ENXIO;
> + goto only_unlock;
> + } else if (xn->state != IDPF_VC_XN_IDLE) {
> + /* We're just going to clobber this transaction even though
> + * it's not IDLE. If we don't reuse it we could theoretically
> + * eventually leak all the free transactions and not be able to
> + * send any messages. At least this way we make an attempt to
> + * remain functional even though something really bad is
> + * happening that's corrupting what was supposed to be free
> + * transactions.
> + */
> + WARN_ONCE(1, "There should only be idle transactions in free list (idx %d op %d)\n",
> + xn->idx, xn->vc_op);
> + }
> +
> + xn->reply = params.recv_buf;
> + xn->reply_sz = 0;
> + xn->state = params.async ? IDPF_VC_XN_ASYNC : IDPF_VC_XN_WAITING;
> + xn->vc_op = params.vc_op;
> + xn->async_handler = params.async_handler;
> + idpf_vc_xn_unlock(xn);
> +
> + if (!params.async)
> + reinit_completion(&xn->completed);
> + cookie = FIELD_PREP(IDPF_VC_XN_SALT_M, xn->salt) |
> + FIELD_PREP(IDPF_VC_XN_IDX_M, xn->idx);
> +
> + retval = idpf_send_mb_msg(adapter, params.vc_op,
> + send_buf->iov_len, send_buf->iov_base,
> + cookie);
> + if (retval) {
> + idpf_vc_xn_lock(xn);
> + goto release_and_unlock;
> + }
> +
> + if (params.async)
> + return 0;
> +
> + wait_for_completion_timeout(&xn->completed,
> + msecs_to_jiffies(params.timeout_ms));
> +
> + /* No need to check the return value; we check the final state of the
> + * transaction below. It's possible the transaction actually gets more
> + * timeout than specified if we get preempted here but after
> + * wait_for_completion_timeout returns. This should be non-issue
> + * however.
> + */
> + idpf_vc_xn_lock(xn);
> + switch (xn->state) {
> + case IDPF_VC_XN_SHUTDOWN:
> + retval = -ENXIO;
> + goto only_unlock;
> + case IDPF_VC_XN_WAITING:
> + dev_notice_ratelimited(&adapter->pdev->dev, "Transaction timed-out (op %d, %dms)\n",
> + params.vc_op, params.timeout_ms);
> + retval = -ETIME;
> + break;
> + case IDPF_VC_XN_COMPLETED_SUCCESS:
> + retval = xn->reply_sz;
> + break;
> + case IDPF_VC_XN_COMPLETED_FAILED:
> + dev_notice_ratelimited(&adapter->pdev->dev, "Transaction failed (op %d)\n",
> + params.vc_op);
> + retval = -EIO;
> + break;
> + default:
> + /* Invalid state. */
> + WARN_ON_ONCE(1);
> + retval = -EIO;
> + break;
> + }
> +
> +release_and_unlock:
> + idpf_vc_xn_push_free(&adapter->vcxn_mngr, xn);
> + /* If we receive a VC reply after here, it will be dropped. */
> +only_unlock:
> + idpf_vc_xn_unlock(xn);
> +
> + return retval;
> +}
[...]
Thanks,
Olek
^ permalink raw reply [flat|nested] 44+ messages in thread
* Re: [Intel-wired-lan] [PATCH v4 01/10 iwl-next] idpf: implement virtchnl transaction manager
2024-02-20 14:30 ` Alexander Lobakin
@ 2024-02-20 16:23 ` Alan Brady
-1 siblings, 0 replies; 44+ messages in thread
From: Alan Brady @ 2024-02-20 16:23 UTC (permalink / raw)
To: Alexander Lobakin
Cc: willemdebruijn.kernel, netdev, Joshua Hay, intel-wired-lan,
przemyslaw.kitszel, igor.bagnucki
On 2/20/2024 6:30 AM, Alexander Lobakin wrote:
> From: Alan Brady <alan.brady@intel.com>
> Date: Mon, 5 Feb 2024 19:37:55 -0800
>
>> This starts refactoring how virtchnl messages are handled by adding a
>> transaction manager (idpf_vc_xn_manager).
>
> [...]
>
>> +/**
>> + * idpf_vc_xn_exec - Perform a send/recv virtchnl transaction
>> + * @adapter: driver specific private structure with vcxn_mngr
>> + * @params: parameters for this particular transaction including
>> + * -vc_op: virtchannel operation to send
>> + * -send_buf: kvec iov for send buf and len
>> + * -recv_buf: kvec iov for recv buf and len (ignored if NULL)
>> + * -timeout_ms: timeout waiting for a reply (milliseconds)
>> + * -async: don't wait for message reply, will lose caller context
>> + * -async_handler: callback to handle async replies
>> + *
>> + * @returns >= 0 for success, the size of the initial reply (may or may not be
>> + * >= @recv_buf.iov_len, but we never overflow @@recv_buf_iov_base). < 0 for
>> + * error.
>> + */
>> +static ssize_t idpf_vc_xn_exec(struct idpf_adapter *adapter,
>> + struct idpf_vc_xn_params params)
>
> Why do you pass @params by value, i.e. whole 56 bytes per each function
> call instead of passing it by pointer -> 8 bytes per call?
>
Not for any particularly good reason. I will fix thanks.
^ permalink raw reply [flat|nested] 44+ messages in thread
* Re: [PATCH v4 01/10 iwl-next] idpf: implement virtchnl transaction manager
@ 2024-02-20 16:23 ` Alan Brady
0 siblings, 0 replies; 44+ messages in thread
From: Alan Brady @ 2024-02-20 16:23 UTC (permalink / raw)
To: Alexander Lobakin
Cc: intel-wired-lan, netdev, willemdebruijn.kernel,
przemyslaw.kitszel, igor.bagnucki, Joshua Hay
On 2/20/2024 6:30 AM, Alexander Lobakin wrote:
> From: Alan Brady <alan.brady@intel.com>
> Date: Mon, 5 Feb 2024 19:37:55 -0800
>
>> This starts refactoring how virtchnl messages are handled by adding a
>> transaction manager (idpf_vc_xn_manager).
>
> [...]
>
>> +/**
>> + * idpf_vc_xn_exec - Perform a send/recv virtchnl transaction
>> + * @adapter: driver specific private structure with vcxn_mngr
>> + * @params: parameters for this particular transaction including
>> + * -vc_op: virtchannel operation to send
>> + * -send_buf: kvec iov for send buf and len
>> + * -recv_buf: kvec iov for recv buf and len (ignored if NULL)
>> + * -timeout_ms: timeout waiting for a reply (milliseconds)
>> + * -async: don't wait for message reply, will lose caller context
>> + * -async_handler: callback to handle async replies
>> + *
>> + * @returns >= 0 for success, the size of the initial reply (may or may not be
>> + * >= @recv_buf.iov_len, but we never overflow @@recv_buf_iov_base). < 0 for
>> + * error.
>> + */
>> +static ssize_t idpf_vc_xn_exec(struct idpf_adapter *adapter,
>> + struct idpf_vc_xn_params params)
>
> Why do you pass @params by value, i.e. whole 56 bytes per each function
> call instead of passing it by pointer -> 8 bytes per call?
>
Not for any particularly good reason. I will fix thanks.
^ permalink raw reply [flat|nested] 44+ messages in thread
end of thread, other threads:[~2024-02-20 16:23 UTC | newest]
Thread overview: 44+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2024-02-06 3:37 [PATCH v4 00/10 iwl-next] idpf: refactor virtchnl messages Alan Brady
2024-02-06 3:37 ` [Intel-wired-lan] " Alan Brady
2024-02-06 3:37 ` [PATCH v4 01/10 iwl-next] idpf: implement virtchnl transaction manager Alan Brady
2024-02-06 3:37 ` [Intel-wired-lan] " Alan Brady
2024-02-20 14:30 ` Alexander Lobakin
2024-02-20 14:30 ` Alexander Lobakin
2024-02-20 16:23 ` [Intel-wired-lan] " Alan Brady
2024-02-20 16:23 ` Alan Brady
2024-02-06 3:37 ` [PATCH v4 02/10 iwl-next] idpf: refactor vport virtchnl messages Alan Brady
2024-02-06 3:37 ` [Intel-wired-lan] " Alan Brady
2024-02-06 3:37 ` [PATCH v4 03/10 iwl-next] idpf: refactor queue related " Alan Brady
2024-02-06 3:37 ` [Intel-wired-lan] " Alan Brady
2024-02-06 3:37 ` [PATCH v4 04/10 iwl-next] idpf: refactor remaining " Alan Brady
2024-02-06 3:37 ` [Intel-wired-lan] " Alan Brady
2024-02-06 3:37 ` [PATCH v4 05/10 iwl-next] idpf: add async_handler for MAC filter messages Alan Brady
2024-02-06 3:37 ` [Intel-wired-lan] " Alan Brady
2024-02-06 3:38 ` [PATCH v4 06/10 iwl-next] idpf: refactor idpf_recv_mb_msg Alan Brady
2024-02-06 3:38 ` [Intel-wired-lan] " Alan Brady
2024-02-06 3:38 ` [PATCH v4 07/10 iwl-next] idpf: cleanup virtchnl cruft Alan Brady
2024-02-06 3:38 ` [Intel-wired-lan] " Alan Brady
2024-02-06 3:38 ` [PATCH v4 08/10 iwl-next] idpf: prevent deinit uninitialized virtchnl core Alan Brady
2024-02-06 3:38 ` [Intel-wired-lan] " Alan Brady
2024-02-06 3:38 ` [PATCH v4 09/10 iwl-next] idpf: fix minor controlq issues Alan Brady
2024-02-06 3:38 ` [Intel-wired-lan] " Alan Brady
2024-02-06 3:38 ` [PATCH v4 10/10 iwl-next] idpf: remove dealloc vector msg err in idpf_intr_rel Alan Brady
2024-02-06 3:38 ` [Intel-wired-lan] " Alan Brady
2024-02-06 17:02 ` [PATCH v4 00/10 iwl-next] idpf: refactor virtchnl messages Alexander Lobakin
2024-02-06 17:02 ` [Intel-wired-lan] " Alexander Lobakin
2024-02-14 14:49 ` Alexander Lobakin
2024-02-14 14:49 ` Alexander Lobakin
2024-02-14 17:06 ` Alan Brady
2024-02-14 17:06 ` [Intel-wired-lan] " Alan Brady
2024-02-20 13:47 ` Alexander Lobakin
2024-02-20 13:47 ` Alexander Lobakin
2024-02-06 18:57 ` Jakub Kicinski
2024-02-06 18:57 ` [Intel-wired-lan] " Jakub Kicinski
2024-02-06 19:18 ` Alan Brady
2024-02-06 19:18 ` [Intel-wired-lan] " Alan Brady
2024-02-06 20:03 ` Jakub Kicinski
2024-02-06 20:03 ` Jakub Kicinski
2024-02-06 22:50 ` [Intel-wired-lan] " Keller, Jacob E
2024-02-06 22:50 ` Keller, Jacob E
2024-02-06 23:17 ` [Intel-wired-lan] " Jakub Kicinski
2024-02-06 23:17 ` Jakub Kicinski
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.