Added the fix for FW and OS hang by using MSM. Also add the fix for pending cmd in fw, after deleting the lds.
Signed-off-by Bo Yang<[email protected]>
---
drivers/scsi/megaraid/megaraid_sas.c | 75 +++++++++++++++++++++++++++++------
drivers/scsi/megaraid/megaraid_sas.h | 25 ++++++-----
2 files changed, 77 insertions(+), 23 deletions(-)
diff -rupN linux-2.6.28_orig/drivers/scsi/megaraid/megaraid_sas.c linux-2.6.28_new/drivers/scsi/megaraid/megaraid_sas.c
--- linux-2.6.28_orig/drivers/scsi/megaraid/megaraid_sas.c 2009-09-14 07:59:41.000000000 -0400
+++ linux-2.6.28_new/drivers/scsi/megaraid/megaraid_sas.c 2009-09-14 09:06:54.000000000 -0400
@@ -226,7 +226,10 @@ megasas_clear_intr_xscale(struct megasas
* @regs : MFI register set
*/
static inline void
-megasas_fire_cmd_xscale(dma_addr_t frame_phys_addr,u32 frame_count, struct megasas_register_set __iomem *regs)
+megasas_fire_cmd_xscale(struct megasas_instance *instance,
+ dma_addr_t frame_phys_addr,
+ u32 frame_count,
+ struct megasas_register_set __iomem *regs)
{
writel((frame_phys_addr >> 3)|(frame_count),
&(regs)->inbound_queue_port);
@@ -323,7 +326,10 @@ megasas_clear_intr_ppc(struct megasas_re
* @regs : MFI register set
*/
static inline void
-megasas_fire_cmd_ppc(dma_addr_t frame_phys_addr, u32 frame_count, struct megasas_register_set __iomem *regs)
+megasas_fire_cmd_ppc(struct megasas_instance *instance,
+ dma_addr_t frame_phys_addr,
+ u32 frame_count,
+ struct megasas_register_set __iomem *regs)
{
writel((frame_phys_addr | (frame_count<<1))|1,
&(regs)->inbound_queue_port);
@@ -413,12 +419,17 @@ megasas_clear_intr_skinny(struct megasas
* @regs : MFI register set
*/
static inline void
-megasas_fire_cmd_skinny(dma_addr_t frame_phys_addr, u32 frame_count,
+megasas_fire_cmd_skinny(struct megasas_instance *instance,
+ dma_addr_t frame_phys_addr,
+ u32 frame_count,
struct megasas_register_set __iomem *regs)
{
+ unsigned long flags;
+ spin_lock_irqsave(&instance->fire_lock, flags);
writel(0, &(regs)->inbound_high_queue_port);
writel((frame_phys_addr | (frame_count<<1))|1,
&(regs)->inbound_low_queue_port);
+ spin_unlock_irqrestore(&instance->fire_lock, flags);
}
static struct megasas_instance_template megasas_instance_template_skinny = {
@@ -508,7 +519,9 @@ megasas_clear_intr_gen2(struct megasas_r
* @regs : MFI register set
*/
static inline void
-megasas_fire_cmd_gen2(dma_addr_t frame_phys_addr, u32 frame_count,
+megasas_fire_cmd_gen2(struct megasas_instance *instance,
+ dma_addr_t frame_phys_addr,
+ u32 frame_count,
struct megasas_register_set __iomem *regs)
{
writel((frame_phys_addr | (frame_count<<1))|1,
@@ -550,7 +563,8 @@ megasas_issue_polled(struct megasas_inst
/*
* Issue the frame using inbound queue port
*/
- instance->instancet->fire_cmd(cmd->frame_phys_addr ,0,instance->reg_set);
+ instance->instancet->fire_cmd(instance,
+ cmd->frame_phys_addr, 0, instance->reg_set);
/*
* Wait for cmd_status to change
@@ -581,7 +595,8 @@ megasas_issue_blocked_cmd(struct megasas
{
cmd->cmd_status = ENODATA;
- instance->instancet->fire_cmd(cmd->frame_phys_addr ,0,instance->reg_set);
+ instance->instancet->fire_cmd(instance,
+ cmd->frame_phys_addr, 0, instance->reg_set);
wait_event_timeout(instance->int_cmd_wait_q, (cmd->cmd_status != ENODATA),
MEGASAS_INTERNAL_CMD_WAIT_TIME*HZ);
@@ -626,7 +641,8 @@ megasas_issue_blocked_abort_cmd(struct m
cmd->sync_cmd = 1;
cmd->cmd_status = 0xFF;
- instance->instancet->fire_cmd(cmd->frame_phys_addr ,0,instance->reg_set);
+ instance->instancet->fire_cmd(instance,
+ cmd->frame_phys_addr, 0, instance->reg_set);
/*
* Wait for this cmd to complete
@@ -1153,7 +1169,8 @@ megasas_queue_command(struct scsi_cmnd *
*/
atomic_inc(&instance->fw_outstanding);
- instance->instancet->fire_cmd(cmd->frame_phys_addr ,cmd->frame_count-1,instance->reg_set);
+ instance->instancet->fire_cmd(instance, cmd->frame_phys_addr,
+ cmd->frame_count-1, instance->reg_set);
/*
* Check if we have pend cmds to be completed
*/
@@ -1346,8 +1363,16 @@ static int megasas_wait_for_outstanding(
* Send signal to FW to stop processing any pending cmds.
* The controller will be taken offline by the OS now.
*/
- writel(MFI_STOP_ADP,
+ if ((instance->pdev->device ==
+ PCI_DEVICE_ID_LSI_SAS0073SKINNY) ||
+ (instance->pdev->device ==
+ PCI_DEVICE_ID_LSI_SAS0071SKINNY)) {
+ writel(MFI_STOP_ADP,
+ &instance->reg_set->reserved_0[0]);
+ } else {
+ writel(MFI_STOP_ADP,
&instance->reg_set->inbound_doorbell);
+ }
megasas_dump_pending_frames(instance);
instance->hw_crit_error = 1;
return FAILED;
@@ -1799,7 +1824,7 @@ megasas_transition_to_ready(struct megas
/*
* Set the CLR bit in inbound doorbell
*/
- if ((instance->pdev->device == \
+ if ((instance->pdev->device ==
PCI_DEVICE_ID_LSI_SAS0073SKINNY) ||
(instance->pdev->device ==
PCI_DEVICE_ID_LSI_SAS0071SKINNY)) {
@@ -2799,7 +2824,8 @@ megasas_register_aen(struct megasas_inst
/*
* Issue the aen registration frame
*/
- instance->instancet->fire_cmd(cmd->frame_phys_addr ,0,instance->reg_set);
+ instance->instancet->fire_cmd(instance,
+ cmd->frame_phys_addr, 0, instance->reg_set);
return 0;
}
@@ -2983,6 +3009,7 @@ megasas_probe_one(struct pci_dev *pdev,
init_waitqueue_head(&instance->abort_cmd_wait_q);
spin_lock_init(&instance->cmd_pool_lock);
+ spin_lock_init(&instance->fire_lock);
spin_lock_init(&instance->completion_lock);
spin_lock_init(&poll_aen_lock);
@@ -3005,7 +3032,7 @@ megasas_probe_one(struct pci_dev *pdev,
megasas_dbg_lvl = 0;
instance->flag = 0;
- instance->unload = 0;
+ instance->unload = 1;
instance->last_time = 0;
/*
@@ -3051,6 +3078,7 @@ megasas_probe_one(struct pci_dev *pdev,
if (megasas_io_attach(instance))
goto fail_io_attach;
+ instance->unload = 0;
return 0;
fail_start_aen:
@@ -3174,6 +3202,7 @@ megasas_suspend(struct pci_dev *pdev, pm
instance = pci_get_drvdata(pdev);
host = instance->host;
+ instance->unload = 1;
if (poll_mode_io)
del_timer_sync(&instance->io_completion_timer);
@@ -3269,6 +3298,8 @@ megasas_resume(struct pci_dev *pdev)
megasas_start_timer(instance, &instance->io_completion_timer,
megasas_io_completion_timer,
MEGASAS_COMPLETION_TIMER_INTERVAL);
+ instance->unload = 0;
+
return 0;
fail_irq:
@@ -3366,6 +3397,7 @@ static void __devexit megasas_detach_one
static void megasas_shutdown(struct pci_dev *pdev)
{
struct megasas_instance *instance = pci_get_drvdata(pdev);
+ instance->unload = 1;
megasas_flush_cache(instance);
megasas_shutdown_controller(instance, MR_DCMD_CTRL_SHUTDOWN);
}
@@ -3615,6 +3647,17 @@ static int megasas_mgmt_ioctl_fw(struct
goto out_kfree_ioc;
}
+ if (instance->hw_crit_error == 1) {
+ printk(KERN_DEBUG "Controller in Crit ERROR\n");
+ error = -ENODEV;
+ goto out_kfree_ioc;
+ }
+
+ if (instance->unload == 1) {
+ error = -ENODEV;
+ goto out_kfree_ioc;
+ }
+
/*
* We will allow only MEGASAS_INT_CMDS number of parallel ioctl cmds
*/
@@ -3650,6 +3693,14 @@ static int megasas_mgmt_ioctl_aen(struct
if (!instance)
return -ENODEV;
+ if (instance->hw_crit_error == 1) {
+ error = -ENODEV;
+ }
+
+ if (instance->unload == 1) {
+ return -ENODEV;
+ }
+
mutex_lock(&instance->aen_mutex);
error = megasas_register_aen(instance, aen.seq_num,
aen.class_locale_word);
diff -rupN linux-2.6.28_orig/drivers/scsi/megaraid/megaraid_sas.h linux-2.6.28_new/drivers/scsi/megaraid/megaraid_sas.h
--- linux-2.6.28_orig/drivers/scsi/megaraid/megaraid_sas.h 2009-09-14 07:59:41.000000000 -0400
+++ linux-2.6.28_new/drivers/scsi/megaraid/megaraid_sas.h 2009-09-14 09:02:28.000000000 -0400
@@ -1157,17 +1157,6 @@ struct megasas_evt_detail {
} __attribute__ ((packed));
- struct megasas_instance_template {
- void (*fire_cmd)(dma_addr_t ,u32 ,struct megasas_register_set __iomem *);
-
- void (*enable_intr)(struct megasas_register_set __iomem *) ;
- void (*disable_intr)(struct megasas_register_set __iomem *);
-
- int (*clear_intr)(struct megasas_register_set __iomem *);
-
- u32 (*read_fw_status_reg)(struct megasas_register_set __iomem *);
- };
-
struct megasas_instance {
u32 *producer;
@@ -1193,6 +1182,8 @@ struct megasas_instance {
spinlock_t cmd_pool_lock;
/* used to synch producer, consumer ptrs in dpc */
spinlock_t completion_lock;
+ /* used to sync fire the cmd to fw */
+ spinlock_t fire_lock;
struct dma_pool *frame_dma_pool;
struct dma_pool *sense_dma_pool;
@@ -1224,6 +1215,18 @@ struct megasas_instance {
struct timer_list io_completion_timer;
};
+struct megasas_instance_template {
+ void (*fire_cmd)(struct megasas_instance *, dma_addr_t, \
+ u32, struct megasas_register_set __iomem *);
+
+ void (*enable_intr)(struct megasas_register_set __iomem *) ;
+ void (*disable_intr)(struct megasas_register_set __iomem *);
+
+ int (*clear_intr)(struct megasas_register_set __iomem *);
+
+ u32 (*read_fw_status_reg)(struct megasas_register_set __iomem *);
+};
+
#define MEGASAS_IS_LOGICAL(scp) \
(scp->device->channel < MEGASAS_MAX_PD_CHANNELS) ? 0 : 1
RESUBMIT:
Added the fix for FW and OS hang by using MSM. Also add the fix for pending cmd in fw, after deleting the lds.
Signed-off-by Bo Yang<[email protected]>
---
drivers/scsi/megaraid/megaraid_sas.c | 75 +++++++++++++++++++++++++++++------
drivers/scsi/megaraid/megaraid_sas.h | 25 ++++++-----
2 files changed, 77 insertions(+), 23 deletions(-)
diff -rupN linux-2.6.28_orig/drivers/scsi/megaraid/megaraid_sas.c linux-2.6.28_new/drivers/scsi/megaraid/megaraid_sas.c
--- linux-2.6.28_orig/drivers/scsi/megaraid/megaraid_sas.c 2009-09-14 07:59:41.000000000 -0400
+++ linux-2.6.28_new/drivers/scsi/megaraid/megaraid_sas.c 2009-09-14 09:06:54.000000000 -0400
@@ -226,7 +226,10 @@ megasas_clear_intr_xscale(struct megasas
* @regs : MFI register set
*/
static inline void
-megasas_fire_cmd_xscale(dma_addr_t frame_phys_addr,u32 frame_count, struct megasas_register_set __iomem *regs)
+megasas_fire_cmd_xscale(struct megasas_instance *instance,
+ dma_addr_t frame_phys_addr,
+ u32 frame_count,
+ struct megasas_register_set __iomem *regs)
{
writel((frame_phys_addr >> 3)|(frame_count),
&(regs)->inbound_queue_port);
@@ -323,7 +326,10 @@ megasas_clear_intr_ppc(struct megasas_re
* @regs : MFI register set
*/
static inline void
-megasas_fire_cmd_ppc(dma_addr_t frame_phys_addr, u32 frame_count, struct megasas_register_set __iomem *regs)
+megasas_fire_cmd_ppc(struct megasas_instance *instance,
+ dma_addr_t frame_phys_addr,
+ u32 frame_count,
+ struct megasas_register_set __iomem *regs)
{
writel((frame_phys_addr | (frame_count<<1))|1,
&(regs)->inbound_queue_port);
@@ -413,12 +419,17 @@ megasas_clear_intr_skinny(struct megasas
* @regs : MFI register set
*/
static inline void
-megasas_fire_cmd_skinny(dma_addr_t frame_phys_addr, u32 frame_count,
+megasas_fire_cmd_skinny(struct megasas_instance *instance,
+ dma_addr_t frame_phys_addr,
+ u32 frame_count,
struct megasas_register_set __iomem *regs)
{
+ unsigned long flags;
+ spin_lock_irqsave(&instance->fire_lock, flags);
writel(0, &(regs)->inbound_high_queue_port);
writel((frame_phys_addr | (frame_count<<1))|1,
&(regs)->inbound_low_queue_port);
+ spin_unlock_irqrestore(&instance->fire_lock, flags);
}
static struct megasas_instance_template megasas_instance_template_skinny = {
@@ -508,7 +519,9 @@ megasas_clear_intr_gen2(struct megasas_r
* @regs : MFI register set
*/
static inline void
-megasas_fire_cmd_gen2(dma_addr_t frame_phys_addr, u32 frame_count,
+megasas_fire_cmd_gen2(struct megasas_instance *instance,
+ dma_addr_t frame_phys_addr,
+ u32 frame_count,
struct megasas_register_set __iomem *regs)
{
writel((frame_phys_addr | (frame_count<<1))|1,
@@ -550,7 +563,8 @@ megasas_issue_polled(struct megasas_inst
/*
* Issue the frame using inbound queue port
*/
- instance->instancet->fire_cmd(cmd->frame_phys_addr ,0,instance->reg_set);
+ instance->instancet->fire_cmd(instance,
+ cmd->frame_phys_addr, 0, instance->reg_set);
/*
* Wait for cmd_status to change
@@ -581,7 +595,8 @@ megasas_issue_blocked_cmd(struct megasas
{
cmd->cmd_status = ENODATA;
- instance->instancet->fire_cmd(cmd->frame_phys_addr ,0,instance->reg_set);
+ instance->instancet->fire_cmd(instance,
+ cmd->frame_phys_addr, 0, instance->reg_set);
wait_event_timeout(instance->int_cmd_wait_q, (cmd->cmd_status != ENODATA),
MEGASAS_INTERNAL_CMD_WAIT_TIME*HZ);
@@ -626,7 +641,8 @@ megasas_issue_blocked_abort_cmd(struct m
cmd->sync_cmd = 1;
cmd->cmd_status = 0xFF;
- instance->instancet->fire_cmd(cmd->frame_phys_addr ,0,instance->reg_set);
+ instance->instancet->fire_cmd(instance,
+ cmd->frame_phys_addr, 0, instance->reg_set);
/*
* Wait for this cmd to complete
@@ -1153,7 +1169,8 @@ megasas_queue_command(struct scsi_cmnd *
*/
atomic_inc(&instance->fw_outstanding);
- instance->instancet->fire_cmd(cmd->frame_phys_addr ,cmd->frame_count-1,instance->reg_set);
+ instance->instancet->fire_cmd(instance, cmd->frame_phys_addr,
+ cmd->frame_count-1, instance->reg_set);
/*
* Check if we have pend cmds to be completed
*/
@@ -1346,8 +1363,16 @@ static int megasas_wait_for_outstanding(
* Send signal to FW to stop processing any pending cmds.
* The controller will be taken offline by the OS now.
*/
- writel(MFI_STOP_ADP,
+ if ((instance->pdev->device ==
+ PCI_DEVICE_ID_LSI_SAS0073SKINNY) ||
+ (instance->pdev->device ==
+ PCI_DEVICE_ID_LSI_SAS0071SKINNY)) {
+ writel(MFI_STOP_ADP,
+ &instance->reg_set->reserved_0[0]);
+ } else {
+ writel(MFI_STOP_ADP,
&instance->reg_set->inbound_doorbell);
+ }
megasas_dump_pending_frames(instance);
instance->hw_crit_error = 1;
return FAILED;
@@ -1799,7 +1824,7 @@ megasas_transition_to_ready(struct megas
/*
* Set the CLR bit in inbound doorbell
*/
- if ((instance->pdev->device == \
+ if ((instance->pdev->device ==
PCI_DEVICE_ID_LSI_SAS0073SKINNY) ||
(instance->pdev->device ==
PCI_DEVICE_ID_LSI_SAS0071SKINNY)) {
@@ -2799,7 +2824,8 @@ megasas_register_aen(struct megasas_inst
/*
* Issue the aen registration frame
*/
- instance->instancet->fire_cmd(cmd->frame_phys_addr ,0,instance->reg_set);
+ instance->instancet->fire_cmd(instance,
+ cmd->frame_phys_addr, 0, instance->reg_set);
return 0;
}
@@ -2983,6 +3009,7 @@ megasas_probe_one(struct pci_dev *pdev,
init_waitqueue_head(&instance->abort_cmd_wait_q);
spin_lock_init(&instance->cmd_pool_lock);
+ spin_lock_init(&instance->fire_lock);
spin_lock_init(&instance->completion_lock);
spin_lock_init(&poll_aen_lock);
@@ -3005,7 +3032,7 @@ megasas_probe_one(struct pci_dev *pdev,
megasas_dbg_lvl = 0;
instance->flag = 0;
- instance->unload = 0;
+ instance->unload = 1;
instance->last_time = 0;
/*
@@ -3051,6 +3078,7 @@ megasas_probe_one(struct pci_dev *pdev,
if (megasas_io_attach(instance))
goto fail_io_attach;
+ instance->unload = 0;
return 0;
fail_start_aen:
@@ -3174,6 +3202,7 @@ megasas_suspend(struct pci_dev *pdev, pm
instance = pci_get_drvdata(pdev);
host = instance->host;
+ instance->unload = 1;
if (poll_mode_io)
del_timer_sync(&instance->io_completion_timer);
@@ -3269,6 +3298,8 @@ megasas_resume(struct pci_dev *pdev)
megasas_start_timer(instance, &instance->io_completion_timer,
megasas_io_completion_timer,
MEGASAS_COMPLETION_TIMER_INTERVAL);
+ instance->unload = 0;
+
return 0;
fail_irq:
@@ -3366,6 +3397,7 @@ static void __devexit megasas_detach_one
static void megasas_shutdown(struct pci_dev *pdev)
{
struct megasas_instance *instance = pci_get_drvdata(pdev);
+ instance->unload = 1;
megasas_flush_cache(instance);
megasas_shutdown_controller(instance, MR_DCMD_CTRL_SHUTDOWN);
}
@@ -3615,6 +3647,17 @@ static int megasas_mgmt_ioctl_fw(struct
goto out_kfree_ioc;
}
+ if (instance->hw_crit_error == 1) {
+ printk(KERN_DEBUG "Controller in Crit ERROR\n");
+ error = -ENODEV;
+ goto out_kfree_ioc;
+ }
+
+ if (instance->unload == 1) {
+ error = -ENODEV;
+ goto out_kfree_ioc;
+ }
+
/*
* We will allow only MEGASAS_INT_CMDS number of parallel ioctl cmds
*/
@@ -3650,6 +3693,14 @@ static int megasas_mgmt_ioctl_aen(struct
if (!instance)
return -ENODEV;
+ if (instance->hw_crit_error == 1) {
+ error = -ENODEV;
+ }
+
+ if (instance->unload == 1) {
+ return -ENODEV;
+ }
+
mutex_lock(&instance->aen_mutex);
error = megasas_register_aen(instance, aen.seq_num,
aen.class_locale_word);
diff -rupN linux-2.6.28_orig/drivers/scsi/megaraid/megaraid_sas.h linux-2.6.28_new/drivers/scsi/megaraid/megaraid_sas.h
--- linux-2.6.28_orig/drivers/scsi/megaraid/megaraid_sas.h 2009-09-14 07:59:41.000000000 -0400
+++ linux-2.6.28_new/drivers/scsi/megaraid/megaraid_sas.h 2009-09-14 09:02:28.000000000 -0400
@@ -1157,17 +1157,6 @@ struct megasas_evt_detail {
} __attribute__ ((packed));
- struct megasas_instance_template {
- void (*fire_cmd)(dma_addr_t ,u32 ,struct megasas_register_set __iomem *);
-
- void (*enable_intr)(struct megasas_register_set __iomem *) ;
- void (*disable_intr)(struct megasas_register_set __iomem *);
-
- int (*clear_intr)(struct megasas_register_set __iomem *);
-
- u32 (*read_fw_status_reg)(struct megasas_register_set __iomem *);
- };
-
struct megasas_instance {
u32 *producer;
@@ -1193,6 +1182,8 @@ struct megasas_instance {
spinlock_t cmd_pool_lock;
/* used to synch producer, consumer ptrs in dpc */
spinlock_t completion_lock;
+ /* used to sync fire the cmd to fw */
+ spinlock_t fire_lock;
struct dma_pool *frame_dma_pool;
struct dma_pool *sense_dma_pool;
@@ -1224,6 +1215,18 @@ struct megasas_instance {
struct timer_list io_completion_timer;
};
+struct megasas_instance_template {
+ void (*fire_cmd)(struct megasas_instance *, dma_addr_t, \
+ u32, struct megasas_register_set __iomem *);
+
+ void (*enable_intr)(struct megasas_register_set __iomem *) ;
+ void (*disable_intr)(struct megasas_register_set __iomem *);
+
+ int (*clear_intr)(struct megasas_register_set __iomem *);
+
+ u32 (*read_fw_status_reg)(struct megasas_register_set __iomem *);
+};
+
#define MEGASAS_IS_LOGICAL(scp) \
(scp->device->channel < MEGASAS_MAX_PD_CHANNELS) ? 0 : 1
On Tue, 2009-10-06 at 14:47 -0600, Yang, Bo wrote:
> RESUBMIT:
>
> Added the fix for FW and OS hang by using MSM. Also add the fix for pending cmd in fw, after deleting the lds.
This is practically unintelligible as a changelog. These logs get used
by distibutions and other people trying to determine if they need the
patch based on bug reports they're seeing
>From the patch I think what you've done is:
1. Added a lock to the skinny firmware initialisation sequence to
prevent the two stage write being non atomic if multiple instances use
it.
2. Added a flag to the driver shutdown sequence to prevent aen ioctls
being called after shutdown begins.
For the future, could we also have separate bug fixes in separate
patches, please.
Also, does this need to go to stable?
Thanks,
James