[PATCH v2] DSPBRIDGE: DSP recovery feature

[PATCH v2] DSPBRIDGE: DSP recovery feature

[Guzman Lugo, Fernando]

Patch updated with new O_NONBLOCK option in bridge_open function.


>From aa1b238e4bd30eb52a96a7000998cd7d32e37050 Mon Sep 17 00:00:00 2001
From: Fernando Guzman Lugo <x0095840@ti.com>
Date: Thu, 25 Mar 2010 15:23:38 -0600
Subject: [PATCH] DSPBRIDGE: DSP recovery feature

This patch implements a workqueue in charge of reseting
DSP in case of fatal error.

Original idea taken from:
http://android.git.kernel.org/?p=kernel/omap.git;a=commit;h=8af1fbcda4b05a8777b4f3819da98340bd5d9de2

Features:
-Recovery is done reloading baseimage.
-Workqueue will load last baseimage loaded.
-In order workqueue can reload the baseimage absolute path for
baseimage must be used.
-Workqueue will wait until all the dspbridge handles are closed
-During recovery process all ioctl calls will return -EIO except
mgr_wait_for_bridge_events
-After DSP crash applications must close their dspbridge handles
and open a new dspbridge handle, bridge_open function will block until
recovery process is done or timeout unless O_NONBLOCK flag be pass to
open function in that case if dspbridge is in recovery process it
will return immediately with -EBUSY error.

Signed-off-by: Fernando Guzman Lugo <x0095840@ti.com>
---
 arch/arm/plat-omap/include/dspbridge/drv.h |    4 ++
 drivers/dsp/bridge/Kconfig                 |    7 +++
 drivers/dsp/bridge/pmgr/wcd.c              |   14 +++----
 drivers/dsp/bridge/rmgr/drv_interface.c    |   64 +++++++++++++++++++++++++++-
 drivers/dsp/bridge/rmgr/proc.c             |    9 +---
 drivers/dsp/bridge/wmd/ue_deh.c            |    6 ++-
 6 files changed, 88 insertions(+), 16 deletions(-)

diff --git a/arch/arm/plat-omap/include/dspbridge/drv.h b/arch/arm/plat-omap/include/dspbridge/drv.h
index 41deb86..947cbdd 100644
--- a/arch/arm/plat-omap/include/dspbridge/drv.h
+++ b/arch/arm/plat-omap/include/dspbridge/drv.h
@@ -386,4 +386,8 @@ extern dsp_status drv_request_resources(IN u32 dw_context,
 extern dsp_status drv_release_resources(IN u32 dw_context,
 					struct drv_object *hdrv_obj);
 
+#ifdef CONFIG_BRIDGE_RECOVERY
+void bridge_recover_schedule(void);
+#endif
+
 #endif /* DRV_ */
diff --git a/drivers/dsp/bridge/Kconfig b/drivers/dsp/bridge/Kconfig
index a3251c3..23b2afc 100644
--- a/drivers/dsp/bridge/Kconfig
+++ b/drivers/dsp/bridge/Kconfig
@@ -38,6 +38,13 @@ config BRIDGE_DEBUG
 	help
 	  Say Y to enable Bridge debugging capabilities
 
+config BRIDGE_RECOVERY
+	bool "DSP Recovery Support"
+	depends on MPU_BRIDGE
+	help
+	  In case of DSP fatal error, BRIDGE driver will try to
+	  recover itself.
+
 comment "Bridge Notifications"
 	depends on MPU_BRIDGE
 
diff --git a/drivers/dsp/bridge/pmgr/wcd.c b/drivers/dsp/bridge/pmgr/wcd.c
index f30e023..15a05a6 100644
--- a/drivers/dsp/bridge/pmgr/wcd.c
+++ b/drivers/dsp/bridge/pmgr/wcd.c
@@ -381,7 +381,7 @@ dsp_status wcd_init_complete2(void)
 	dsp_status status = DSP_SOK;
 	struct cfg_devnode *dev_node;
 	struct dev_object *hdev_obj;
-	u32 dev_type;
+	u32 dev_type, tmp;
 
 	DBC_REQUIRE(wcd_c_refs > 0);
 
@@ -396,13 +396,11 @@ dsp_status wcd_init_complete2(void)
 		if (DSP_FAILED(dev_get_dev_type(hdev_obj, &dev_type)))
 			continue;
 
-		if ((dev_type == DSP_UNIT) || (dev_type == IVA_UNIT)) {
-			if (DSP_FAILED(proc_auto_start(dev_node, hdev_obj))) {
-				status = DSP_EFAIL;
-				/* break; */
-			}
-		}
-	}			/* End For Loop */
+		if ((dev_type == DSP_UNIT) || (dev_type == IVA_UNIT))
+			if (cfg_get_auto_start(dev_node, &tmp) == DSP_SOK
+									&& tmp)
+				proc_auto_start(dev_node, hdev_obj);
+	}
 
 	return status;
 }
diff --git a/drivers/dsp/bridge/rmgr/drv_interface.c b/drivers/dsp/bridge/rmgr/drv_interface.c
index d592181..5f01ab9 100644
--- a/drivers/dsp/bridge/rmgr/drv_interface.c
+++ b/drivers/dsp/bridge/rmgr/drv_interface.c
@@ -91,6 +91,15 @@ static s32 shm_size = 0x500000;	/* 5 MB */
 static u32 phys_mempool_base;
 static u32 phys_mempool_size;
 static int tc_wordswapon;	/* Default value is always false */
+#ifdef CONFIG_BRIDGE_RECOVERY
+#define REC_TIMEOUT 5000	/*recovery timeout in msecs */
+static atomic_t bridge_cref;	/* number of bridge open handles */
+static struct workqueue_struct *bridge_rec_queue;
+static struct work_struct bridge_recovery_work;
+static DECLARE_COMPLETION_ONSTACK(bridge_comp);
+static DECLARE_COMPLETION_ONSTACK(bridge_open_comp);
+static bool recover;
+#endif
 
 #ifdef CONFIG_PM
 struct omap34_xx_bridge_suspend_data {
@@ -189,6 +198,33 @@ u32 vdd1_dsp_freq[6][4] = {
 	{0, 430000, 355000, 430000},
 };
 
+#ifdef CONFIG_BRIDGE_RECOVERY
+static void bridge_recover(struct work_struct *work)
+{
+	struct dev_object *dev;
+	struct cfg_devnode *dev_node;
+	if (atomic_read(&bridge_cref)) {
+		INIT_COMPLETION(bridge_comp);
+		while (!wait_for_completion_timeout(&bridge_comp,
+						msecs_to_jiffies(REC_TIMEOUT)))
+			pr_info("%s:%d handle(s) still opened\n",
+					__func__, atomic_read(&bridge_cref));
+	}
+	dev = dev_get_first();
+	dev_get_dev_node(dev, &dev_node);
+	if (!dev_node || DSP_FAILED(proc_auto_start(dev_node, dev)))
+		pr_err("DSP could not be restarted\n");
+	recover = false;
+	complete_all(&bridge_open_comp);
+}
+
+void bridge_recover_schedule(void)
+{
+	INIT_COMPLETION(bridge_open_comp);
+	recover = true;
+	queue_work(bridge_rec_queue, &bridge_recovery_work);
+}
+#endif
 #ifdef CONFIG_BRIDGE_DVFS
 static int dspbridge_post_scale(struct notifier_block *op, unsigned long level,
 				void *ptr)
@@ -323,6 +359,11 @@ static int __devinit omap34_xx_bridge_probe(struct platform_device *pdev)
 		}
 	}
 
+#ifdef CONFIG_BRIDGE_RECOVERY
+	bridge_rec_queue = create_workqueue("bridge_rec_queue");
+	INIT_WORK(&bridge_recovery_work, bridge_recover);
+	INIT_COMPLETION(bridge_comp);
+#endif
 	DBC_ASSERT(status == 0);
 	DBC_ASSERT(DSP_SUCCEEDED(init_status));
 
@@ -443,6 +484,14 @@ static int bridge_open(struct inode *ip, struct file *filp)
 	 * Allocate a new process context and insert it into global
 	 * process context list.
 	 */
+
+#ifdef CONFIG_BRIDGE_RECOVERY
+	if (recover) {
+		if (filp->f_flags & O_NONBLOCK ||
+			wait_for_completion_interruptible(&bridge_open_comp))
+			return -EBUSY;
+	}
+#endif
 	pr_ctxt = mem_calloc(sizeof(struct process_context), MEM_PAGED);
 	if (pr_ctxt) {
 		pr_ctxt->res_state = PROC_RES_ALLOCATED;
@@ -457,7 +506,10 @@ static int bridge_open(struct inode *ip, struct file *filp)
 	}
 
 	filp->private_data = pr_ctxt;
-
+#ifdef CONFIG_BRIDGE_RECOVERY
+	if (!status)
+		atomic_inc(&bridge_cref);
+#endif
 	return status;
 }
 
@@ -484,6 +536,10 @@ static int bridge_release(struct inode *ip, struct file *filp)
 	filp->private_data = NULL;
 
 err:
+#ifdef CONFIG_BRIDGE_RECOVERY
+	if (!atomic_dec_return(&bridge_cref))
+		complete(&bridge_comp);
+#endif
 	return status;
 }
 
@@ -496,6 +552,12 @@ static long bridge_ioctl(struct file *filp, unsigned int code,
 	union Trapped_Args buf_in;
 
 	DBC_REQUIRE(filp != NULL);
+#ifdef CONFIG_BRIDGE_RECOVERY
+	if (recover) {
+		status = -EIO;
+		goto err;
+	}
+#endif
 #ifdef CONFIG_PM
 	status = omap34_xxbridge_suspend_lockout(&bridge_suspend_data, filp);
 	if (status != 0)
diff --git a/drivers/dsp/bridge/rmgr/proc.c b/drivers/dsp/bridge/rmgr/proc.c
index b3e4546..1556285 100644
--- a/drivers/dsp/bridge/rmgr/proc.c
+++ b/drivers/dsp/bridge/rmgr/proc.c
@@ -53,6 +53,7 @@
 #include <dspbridge/msg.h>
 #include <dspbridge/wmdioctl.h>
 #include <dspbridge/drv.h>
+#include <dspbridge/reg.h>
 
 /*  ----------------------------------- This */
 #include <dspbridge/proc.h>
@@ -268,7 +269,6 @@ dsp_status proc_auto_start(struct cfg_devnode *dev_node_obj,
 			   struct dev_object *hdev_obj)
 {
 	dsp_status status = DSP_EFAIL;
-	u32 dw_auto_start = 0;	/* autostart flag */
 	struct proc_object *p_proc_object;
 	char sz_exec_file[MAXCMDLINELEN];
 	char *argv[2];
@@ -304,11 +304,6 @@ dsp_status proc_auto_start(struct cfg_devnode *dev_node_obj,
 	if (DSP_FAILED(status))
 		goto func_cont;
 
-	status = cfg_get_auto_start(dev_node_obj, &dw_auto_start);
-	if (DSP_FAILED(status) || !dw_auto_start) {
-		status = DSP_EFAIL;
-		goto func_cont;
-	}
 	/* Get the default executable for this board... */
 	dev_get_dev_type(hdev_obj, (u32 *) &dev_type);
 	p_proc_object->processor_id = dev_type;
@@ -1031,6 +1026,8 @@ dsp_status proc_load(void *hprocessor, IN CONST s32 argc_index,
 		if (DSP_SUCCEEDED((*p_proc_object->intf_fxns->pfn_brd_status)
 				  (p_proc_object->hwmd_context, &brd_state))) {
 			pr_info("%s: Processor Loaded %s\n", __func__, pargv0);
+			reg_set_value(DEFEXEC, (u8 *)pargv0,
+							strlen(pargv0) + 1);
 			DBC_ASSERT(brd_state == BRD_LOADED);
 		}
 	}
diff --git a/drivers/dsp/bridge/wmd/ue_deh.c b/drivers/dsp/bridge/wmd/ue_deh.c
index 231b05e..14dd8ae 100644
--- a/drivers/dsp/bridge/wmd/ue_deh.c
+++ b/drivers/dsp/bridge/wmd/ue_deh.c
@@ -288,8 +288,12 @@ void bridge_deh_notify(struct deh_mgr *hdeh_mgr, u32 ulEventMask, u32 dwErrInfo)
 		}
 
 		/* Filter subsequent notifications when an error occurs */
-		if (dev_context->dw_brd_state != BRD_ERROR)
+		if (dev_context->dw_brd_state != BRD_ERROR) {
 			ntfy_notify(deh_mgr_obj->ntfy_obj, ulEventMask);
+#ifdef CONFIG_BRIDGE_RECOVERY
+			bridge_recover_schedule();
+#endif
+		}
 
 		/* Set the Board state as ERROR */
 		dev_context->dw_brd_state = BRD_ERROR;
-- 
1.6.0.4

Re: [PATCH v2] DSPBRIDGE: DSP recovery feature

[Felipe Contreras]

On Fri, Mar 26, 2010 at 12:17 AM, Guzman Lugo, Fernando <x0095840@ti.com> wrote:
>
> Patch updated with new O_NONBLOCK option in bridge_open function.
>
>
> From aa1b238e4bd30eb52a96a7000998cd7d32e37050 Mon Sep 17 00:00:00 2001
> From: Fernando Guzman Lugo <x0095840@ti.com>
> Date: Thu, 25 Mar 2010 15:23:38 -0600
> Subject: [PATCH] DSPBRIDGE: DSP recovery feature
>
> This patch implements a workqueue in charge of reseting
> DSP in case of fatal error.

Tested-by: Felipe Contreras <felipe.contreras@nokia.com>

-- 
Felipe Contreras

Re: [PATCH v2] DSPBRIDGE: DSP recovery feature

[Omar Ramirez Luna]

On 3/28/2010 6:29 PM, Felipe Contreras wrote:
> On Fri, Mar 26, 2010 at 12:17 AM, Guzman Lugo, Fernando<x0095840@ti.com>  wrote:
>>
>> Patch updated with new O_NONBLOCK option in bridge_open function.
>>
>>
>>  From aa1b238e4bd30eb52a96a7000998cd7d32e37050 Mon Sep 17 00:00:00 2001
>> From: Fernando Guzman Lugo<x0095840@ti.com>
>> Date: Thu, 25 Mar 2010 15:23:38 -0600
>> Subject: [PATCH] DSPBRIDGE: DSP recovery feature
>>
>> This patch implements a workqueue in charge of reseting
>> DSP in case of fatal error.
>
> Tested-by: Felipe Contreras<felipe.contreras@nokia.com>
>

Pushed to dspbridge.

- omar