OSDN Git Service

scsi: aacraid: fix PCI error recovery path
authorGuilherme G. Piccoli <gpiccoli@linux.vnet.ibm.com>
Thu, 6 Apr 2017 21:12:09 +0000 (18:12 -0300)
committerMartin K. Petersen <martin.petersen@oracle.com>
Wed, 12 Apr 2017 00:45:59 +0000 (20:45 -0400)
During a PCI error recovery, if aac_check_health() is not aware that a
PCI error happened and we have an offline PCI channel, it might trigger
some errors (like NULL pointer dereference) and inhibit the error
recovery process to complete.

This patch makes the health check procedure aware of PCI channel issues,
and in case of error recovery process, the function
aac_adapter_check_health() returns -1 and let the recovery process to
complete successfully. This patch was tested on upstream kernel
v4.11-rc5 in PowerPC ppc64le architecture with adapter 9005:028d
(VID:DID) - the error recovery procedure was able to recover fine.

Fixes: 5c63f7f710bd ("aacraid: Added EEH support")
Cc: stable@vger.kernel.org # v4.6+
Signed-off-by: Guilherme G. Piccoli <gpiccoli@linux.vnet.ibm.com>
Reviewed-by: Dave Carroll <david.carroll@microsemi.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
drivers/scsi/aacraid/aacraid.h
drivers/scsi/aacraid/commsup.c

index d036a80..d281492 100644 (file)
@@ -1690,9 +1690,6 @@ struct aac_dev
 #define aac_adapter_sync_cmd(dev, command, p1, p2, p3, p4, p5, p6, status, r1, r2, r3, r4) \
        (dev)->a_ops.adapter_sync_cmd(dev, command, p1, p2, p3, p4, p5, p6, status, r1, r2, r3, r4)
 
-#define aac_adapter_check_health(dev) \
-       (dev)->a_ops.adapter_check_health(dev)
-
 #define aac_adapter_restart(dev, bled, reset_type) \
        ((dev)->a_ops.adapter_restart(dev, bled, reset_type))
 
@@ -2615,6 +2612,14 @@ static inline unsigned int cap_to_cyls(sector_t capacity, unsigned divisor)
        return capacity;
 }
 
+static inline int aac_adapter_check_health(struct aac_dev *dev)
+{
+       if (unlikely(pci_channel_offline(dev->pdev)))
+               return -1;
+
+       return (dev)->a_ops.adapter_check_health(dev);
+}
+
 /* SCp.phase values */
 #define AAC_OWNER_MIDLEVEL     0x101
 #define AAC_OWNER_LOWLEVEL     0x102
index c8172f1..1f49183 100644 (file)
@@ -1873,7 +1873,8 @@ int aac_check_health(struct aac_dev * aac)
        spin_unlock_irqrestore(&aac->fib_lock, flagv);
 
        if (BlinkLED < 0) {
-               printk(KERN_ERR "%s: Host adapter dead %d\n", aac->name, BlinkLED);
+               printk(KERN_ERR "%s: Host adapter is dead (or got a PCI error) %d\n",
+                               aac->name, BlinkLED);
                goto out;
        }