From: Hannes Reinecke <hare@suse.de>
To: James Bottomley <jbottomley@parallels.com>
Cc: Christoph Hellwig <hch@infradead.org>,
"linux-scsi@vger.kernel.org" <linux-scsi@vger.kernel.org>,
Ren Mingxin <renmx@cn.fujitsu.com>, Joern Engel <joern@logfs.org>,
James Smart <james.smart@emulex.com>
Subject: Re: [PATCH 2/3] scsi: improved eh timeout handler
Date: Mon, 04 Nov 2013 15:46:30 +0100 [thread overview]
Message-ID: <5277B346.201@suse.de> (raw)
In-Reply-To: <1383575122.2485.4.camel@dabdike>
[-- Attachment #1: Type: text/plain, Size: 1517 bytes --]
On 11/04/2013 03:25 PM, James Bottomley wrote:
> On Mon, 2013-11-04 at 14:36 +0100, Hannes Reinecke wrote:
>> On 10/31/2013 04:49 PM, Christoph Hellwig wrote:
>>> Looks reasonable to me, but a few minor nitpicks:
>>>
>>>> + spin_lock_irqsave(sdev->host->host_lock, flags);
>>>> + if (scsi_host_eh_past_deadline(sdev->host)) {
>>>
>>> I don't have the implementation of scsi_host_eh_past_deadline in my
>>> local tree, but do we really need the host lock for it?
>>>
>> Yes. The eh_deadline variable might be set from an interrupt context
>> or from userland, so we need to protect access to it.
>
> That's not really true. on all our supported architectures 32 bit
> reads/writes are atomic, which means that if one CPU writes a word at
> the same time another reads one, the reader is guaranteed to see either
> the old or the new data. Given the expense of lock cache line bouncing
> on the newer architectures, we really want to avoid a spinlock where
> possible.
>
> In this case, the problem with the implementation is that the writer
> might set eh_deadline to zero, but this is fixable in
> scsi_host_eh_past_deadline() by checking for zero before and after the
> time_before (for the zero to non-zero and non-zero to zero cases).
>
IE you mean something like that attached patch?
Cheers,
Hannes
--
Dr. Hannes Reinecke zSeries & Storage
hare@suse.de +49 911 74053 688
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg
GF: J. Hawn, J. Guild, F. Imendörffer, HRB 16746 (AG Nürnberg)
[-- Attachment #2: tmp-eh-unlock.patch --]
[-- Type: text/x-patch, Size: 5597 bytes --]
diff --git a/drivers/scsi/scsi_error.c b/drivers/scsi/scsi_error.c
index 6a137fa..8abf7ba 100644
--- a/drivers/scsi/scsi_error.c
+++ b/drivers/scsi/scsi_error.c
@@ -94,8 +94,10 @@ static int scsi_host_eh_past_deadline(struct Scsi_Host *shost)
if (!shost->last_reset || !shost->eh_deadline)
return 0;
+ /* Double check eh_deadline to catch atomic updates */
if (time_before(jiffies,
- shost->last_reset + shost->eh_deadline))
+ shost->last_reset + shost->eh_deadline) &&
+ shost->eh_deadline)
return 0;
return 1;
@@ -114,15 +116,12 @@ scmd_eh_abort_handler(struct work_struct *work)
unsigned long flags;
int rtn;
- spin_lock_irqsave(sdev->host->host_lock, flags);
if (scsi_host_eh_past_deadline(sdev->host)) {
- spin_unlock_irqrestore(sdev->host->host_lock, flags);
SCSI_LOG_ERROR_RECOVERY(3,
scmd_printk(KERN_INFO, scmd,
"scmd %p eh timeout, not aborting\n",
scmd));
} else {
- spin_unlock_irqrestore(sdev->host->host_lock, flags);
SCSI_LOG_ERROR_RECOVERY(3,
scmd_printk(KERN_INFO, scmd,
"aborting command %p\n", scmd));
@@ -1140,16 +1139,13 @@ int scsi_eh_get_sense(struct list_head *work_q,
continue;
shost = scmd->device->host;
- spin_lock_irqsave(shost->host_lock, flags);
if (scsi_host_eh_past_deadline(shost)) {
- spin_unlock_irqrestore(shost->host_lock, flags);
SCSI_LOG_ERROR_RECOVERY(3,
shost_printk(KERN_INFO, shost,
"skip %s, past eh deadline\n",
__func__));
break;
}
- spin_unlock_irqrestore(shost->host_lock, flags);
SCSI_LOG_ERROR_RECOVERY(2, scmd_printk(KERN_INFO, scmd,
"%s: requesting sense\n",
current->comm));
@@ -1242,19 +1238,15 @@ static int scsi_eh_test_devices(struct list_head *cmd_list,
sdev = scmd->device;
if (!try_stu) {
- spin_lock_irqsave(sdev->host->host_lock, flags);
if (scsi_host_eh_past_deadline(sdev->host)) {
/* Push items back onto work_q */
list_splice_init(cmd_list, work_q);
- spin_unlock_irqrestore(sdev->host->host_lock,
- flags);
SCSI_LOG_ERROR_RECOVERY(3,
shost_printk(KERN_INFO, sdev->host,
"skip %s, past eh deadline",
__func__));
break;
}
- spin_unlock_irqrestore(sdev->host->host_lock, flags);
}
finish_cmds = !scsi_device_online(scmd->device) ||
@@ -1301,9 +1293,7 @@ static int scsi_eh_abort_cmds(struct list_head *work_q,
if (!(scmd->eh_eflags & SCSI_EH_CANCEL_CMD))
continue;
shost = scmd->device->host;
- spin_lock_irqsave(shost->host_lock, flags);
if (scsi_host_eh_past_deadline(shost)) {
- spin_unlock_irqrestore(shost->host_lock, flags);
list_splice_init(&check_list, work_q);
SCSI_LOG_ERROR_RECOVERY(3,
shost_printk(KERN_INFO, shost,
@@ -1311,7 +1301,6 @@ static int scsi_eh_abort_cmds(struct list_head *work_q,
__func__));
return list_empty(work_q);
}
- spin_unlock_irqrestore(shost->host_lock, flags);
SCSI_LOG_ERROR_RECOVERY(3, printk("%s: aborting cmd:"
"0x%p\n", current->comm,
scmd));
@@ -1378,16 +1367,13 @@ static int scsi_eh_stu(struct Scsi_Host *shost,
unsigned long flags;
shost_for_each_device(sdev, shost) {
- spin_lock_irqsave(shost->host_lock, flags);
if (scsi_host_eh_past_deadline(shost)) {
- spin_unlock_irqrestore(shost->host_lock, flags);
SCSI_LOG_ERROR_RECOVERY(3,
shost_printk(KERN_INFO, shost,
"skip %s, past eh deadline\n",
__func__));
break;
}
- spin_unlock_irqrestore(shost->host_lock, flags);
stu_scmd = NULL;
list_for_each_entry(scmd, work_q, eh_entry)
if (scmd->device == sdev && SCSI_SENSE_VALID(scmd) &&
@@ -1445,16 +1431,13 @@ static int scsi_eh_bus_device_reset(struct Scsi_Host *shost,
int rtn;
shost_for_each_device(sdev, shost) {
- spin_lock_irqsave(shost->host_lock, flags);
if (scsi_host_eh_past_deadline(shost)) {
- spin_unlock_irqrestore(shost->host_lock, flags);
SCSI_LOG_ERROR_RECOVERY(3,
shost_printk(KERN_INFO, shost,
"skip %s, past eh deadline\n",
__func__));
break;
}
- spin_unlock_irqrestore(shost->host_lock, flags);
bdr_scmd = NULL;
list_for_each_entry(scmd, work_q, eh_entry)
if (scmd->device == sdev) {
@@ -1517,9 +1500,7 @@ static int scsi_eh_target_reset(struct Scsi_Host *shost,
unsigned int id;
unsigned long flags;
- spin_lock_irqsave(shost->host_lock, flags);
if (scsi_host_eh_past_deadline(shost)) {
- spin_unlock_irqrestore(shost->host_lock, flags);
/* push back on work queue for further processing */
list_splice_init(&check_list, work_q);
list_splice_init(&tmp_list, work_q);
@@ -1529,7 +1510,6 @@ static int scsi_eh_target_reset(struct Scsi_Host *shost,
__func__));
return list_empty(work_q);
}
- spin_unlock_irqrestore(shost->host_lock, flags);
scmd = list_entry(tmp_list.next, struct scsi_cmnd, eh_entry);
id = scmd_id(scmd);
@@ -1584,9 +1564,7 @@ static int scsi_eh_bus_reset(struct Scsi_Host *shost,
*/
for (channel = 0; channel <= shost->max_channel; channel++) {
- spin_lock_irqsave(shost->host_lock, flags);
if (scsi_host_eh_past_deadline(shost)) {
- spin_unlock_irqrestore(shost->host_lock, flags);
list_splice_init(&check_list, work_q);
SCSI_LOG_ERROR_RECOVERY(3,
shost_printk(KERN_INFO, shost,
@@ -1594,7 +1572,6 @@ static int scsi_eh_bus_reset(struct Scsi_Host *shost,
__func__));
return list_empty(work_q);
}
- spin_unlock_irqrestore(shost->host_lock, flags);
chan_scmd = NULL;
list_for_each_entry(scmd, work_q, eh_entry) {
next prev parent reply other threads:[~2013-11-04 14:46 UTC|newest]
Thread overview: 17+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-10-31 13:02 [PATCHv8 0/3] New EH command timeout handler Hannes Reinecke
2013-10-31 13:02 ` [PATCH 1/3] scsi: Fix erratic device offline during EH Hannes Reinecke
2013-10-31 13:02 ` [PATCH 2/3] scsi: improved eh timeout handler Hannes Reinecke
2013-10-31 15:49 ` Christoph Hellwig
2013-11-04 13:36 ` Hannes Reinecke
2013-11-04 14:25 ` James Bottomley
2013-11-04 14:46 ` Hannes Reinecke [this message]
2013-11-04 14:50 ` James Bottomley
2013-11-04 15:43 ` Hannes Reinecke
2013-11-05 1:07 ` James Bottomley
2013-11-01 6:10 ` Ren Mingxin
2013-10-31 13:02 ` [PATCH 3/3] scsi: Update documentation Hannes Reinecke
-- strict thread matches above, loose matches on Subject: below --
2013-09-02 11:58 [PATCHv6 0/3] New EH command timeout handler Hannes Reinecke
2013-09-02 11:58 ` [PATCH 2/3] scsi: improved eh " Hannes Reinecke
2013-09-11 9:16 ` Ren Mingxin
2013-09-12 20:49 ` Hannes Reinecke
2013-09-20 7:59 ` Ren Mingxin
2013-10-02 16:24 ` Hannes Reinecke
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=5277B346.201@suse.de \
--to=hare@suse.de \
--cc=hch@infradead.org \
--cc=james.smart@emulex.com \
--cc=jbottomley@parallels.com \
--cc=joern@logfs.org \
--cc=linux-scsi@vger.kernel.org \
--cc=renmx@cn.fujitsu.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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.