Hi Damien,
在 2024/4/1 星期一 15:08, Damien Le Moal 写道:
On 4/1/24 14:49, chenxiang wrote:
From: Yihang Li <liyihang9@xxxxxxxxxx>
We found that the second parameter of function ata_wait_after_reset() is
incorrectly used. We call smp_ata_check_ready_type() to poll the device
type until the 30s timeout, so the correct deadline should be (jiffies +
30000).
Fixes: 3c2673a09cf1 ("scsi: hisi_sas: Fix SATA devices missing issue during I_T nexus reset")
Signed-off-by: xiabing <xiabing12@xxxxxxxxxxxxxx>
Signed-off-by: Yihang Li <liyihang9@xxxxxxxxxx>
Signed-off-by: Xiang Chen <chenxiang66@xxxxxxxxxxxxx>
---
drivers/scsi/hisi_sas/hisi_sas_main.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/drivers/scsi/hisi_sas/hisi_sas_main.c b/drivers/scsi/hisi_sas/hisi_sas_main.c
index 097dfe4b620d..7245600aedb2 100644
--- a/drivers/scsi/hisi_sas/hisi_sas_main.c
+++ b/drivers/scsi/hisi_sas/hisi_sas_main.c
@@ -1796,8 +1796,10 @@ static int hisi_sas_debug_I_T_nexus_reset(struct domain_device *device)
if (dev_is_sata(device)) {
struct ata_link *link = &device->sata_dev.ap->link;
+ unsigned long deadline = ata_deadline(jiffies,
+ HISI_SAS_WAIT_PHYUP_TIMEOUT / HZ * 1000);
At the very least, you should use jiffies_to_msecs() here. But I do not see the
point though given that ata_deadline will do "jiffies + msecs_to_jiffies()".
We found the issue when reading the code of libata-eh which uses
ata_deadline() and ata_wait_after_reset().
So we modify it as above.
So may be calling:
rc = ata_wait_after_reset(link, jiffies + HISI_SAS_WAIT_PHYUP_TIMEOUT,
smp_ata_check_ready_type);
May be simpler and more readable.
OK, it is more simpler and readable. Will change it in next version.
- rc = ata_wait_after_reset(link, HISI_SAS_WAIT_PHYUP_TIMEOUT,
+ rc = ata_wait_after_reset(link, deadline,
smp_ata_check_ready_type);
} else {
msleep(2000);
