All of lore.kernel.org
 help / color / mirror / Atom feed
From: NeilBrown <neilb@suse.com>
To: lustre-devel@lists.lustre.org
Subject: [lustre-devel] [PATCH 02/21] lustre: obd_class: remove csi_barrier from struct cl_sync_io
Date: Mon, 11 Feb 2019 11:24:26 +1100	[thread overview]
Message-ID: <87va1rgojp.fsf@notabene.neil.brown.name> (raw)
In-Reply-To: <28122F36-DB60-49B8-82E5-1C2F9546C83E@whamcloud.com>

On Fri, Feb 08 2019, Andreas Dilger wrote:

> On Feb 6, 2019, at 17:03, NeilBrown <neilb@suse.com> wrote:
>> 
>> This flag is used to ensure that structure isn't freed before
>> the wakeup completes.  The same can be achieved using the
>> csi_waitq.lock and calling wake_up_all_locked().
>> 
>> Signed-off-by: NeilBrown <neilb@suse.com>
>> ---
>> drivers/staging/lustre/lustre/include/cl_object.h |    2 --
>> drivers/staging/lustre/lustre/obdclass/cl_io.c    |   16 +++++++---------
>> 2 files changed, 7 insertions(+), 11 deletions(-)
>> 
>> @@ -1080,11 +1079,10 @@ int cl_sync_io_wait(const struct lu_env *env, struct cl_sync_io *anchor,
>> 	} else {
>> 		rc = anchor->csi_sync_rc;
>> 	}
>> +	/* We take the lock to ensure that cl_sync_io_note() has finished */
>> +	spin_lock(&anchor->csi_waitq.lock);
>> 	LASSERT(atomic_read(&anchor->csi_sync_nr) == 0);
>> -
>> -	/* wait until cl_sync_io_note() has done wakeup */
>> -	while (unlikely(atomic_read(&anchor->csi_barrier) != 0))
>> -		cpu_relax();
>> +	spin_unlock(&anchor->csi_waitq.lock);
>> 
>> 	return rc;
>> }
>> @@ -1104,11 +1102,11 @@ void cl_sync_io_note(const struct lu_env *env, struct cl_sync_io *anchor,
>> 	 * IO.
>> 	 */
>> 	LASSERT(atomic_read(&anchor->csi_sync_nr) > 0);
>> -	if (atomic_dec_and_test(&anchor->csi_sync_nr)) {
>> +	if (atomic_dec_and_lock(&anchor->csi_sync_nr,
>> +				&anchor->csi_waitq.lock)) {
>> 
>> -		wake_up_all(&anchor->csi_waitq);
>> -		/* it's safe to nuke or reuse anchor now */
>> -		atomic_set(&anchor->csi_barrier, 0);
>> +		wake_up_all_locked(&anchor->csi_waitq);
>> +		spin_unlock(&anchor->csi_waitq.lock);
>
> Maybe a matching comment here that the lock is to synchronize with cl_sync_io_wait() cleanup?
> Either way,
>
Good idea.  I added

		/*
		 * Holding the lock across both the decrement and
		 * the wakeup ensures cl_sync_io_wait() doesn't complete
		 * before the wakeup completes.
		 */

> Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

Thanks,
NeilBrown

>
> Cheers, Andreas
> ---
> Andreas Dilger
> Principal Lustre Architect
> Whamcloud
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 832 bytes
Desc: not available
URL: <http://lists.lustre.org/pipermail/lustre-devel-lustre.org/attachments/20190211/56192e20/attachment.sig>

  reply	other threads:[~2019-02-11  0:24 UTC|newest]

Thread overview: 87+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-02-07  0:03 [lustre-devel] [PATCH 00/21] lustre: Assorted cleanups for obdclass NeilBrown
2019-02-07  0:03 ` [lustre-devel] [PATCH 04/21] lustre: use list*entry macros in place of container_of() NeilBrown
2019-02-08  0:25   ` Andreas Dilger
2019-02-11  1:32   ` James Simmons
2019-02-11  3:14     ` NeilBrown
2019-02-07  0:03 ` [lustre-devel] [PATCH 05/21] lustre: use list_first_entry() in lustre subdirectory NeilBrown
2019-02-08  0:31   ` Andreas Dilger
2019-02-11  0:13     ` NeilBrown
2019-02-11  1:45   ` James Simmons
2019-02-11  3:08     ` NeilBrown
2019-02-07  0:03 ` [lustre-devel] [PATCH 03/21] lustre: obdclass: use list_sort() to sort a list NeilBrown
2019-02-08  0:13   ` Andreas Dilger
2019-02-11  0:39   ` James Simmons
2019-02-11  1:05     ` NeilBrown
2019-02-07  0:03 ` [lustre-devel] [PATCH 02/21] lustre: obd_class: remove csi_barrier from struct cl_sync_io NeilBrown
2019-02-08  0:09   ` Andreas Dilger
2019-02-11  0:24     ` NeilBrown [this message]
2019-02-11  0:34   ` James Simmons
2019-02-11  1:09     ` NeilBrown
2019-02-07  0:03 ` [lustre-devel] [PATCH 01/21] lustre: obdclass: discard csi_end_io NeilBrown
2019-02-07  0:20   ` Andreas Dilger
2019-02-11  0:19   ` James Simmons
2019-02-07  0:03 ` [lustre-devel] [PATCH 06/21] lustre: use list_first_entry() in lnet/lnet subdirectory NeilBrown
2019-02-08  0:44   ` Andreas Dilger
2019-02-11  1:46   ` James Simmons
2019-02-07  0:03 ` [lustre-devel] [PATCH 18/21] lustre: move debug.c from obdclass to obdecho NeilBrown
2019-02-08  6:02   ` Andreas Dilger
2019-02-11  4:17   ` James Simmons
2019-02-07  0:03 ` [lustre-devel] [PATCH 10/21] lustre: obdclass: use cl_object_for_each where appropriate NeilBrown
2019-02-08  1:10   ` Andreas Dilger
2019-02-11  0:42     ` NeilBrown
2019-02-11  4:19       ` James Simmons
2019-02-15 18:15       ` Andreas Dilger
2019-02-11  1:57   ` James Simmons
2019-02-11  3:19     ` NeilBrown
2019-02-07  0:03 ` [lustre-devel] [PATCH 11/21] lustre: cl_object: remove vestigial debugging NeilBrown
2019-02-08  1:31   ` Andreas Dilger
2019-02-11  0:48     ` NeilBrown
2019-02-11  2:04   ` James Simmons
2019-02-11  3:25     ` NeilBrown
2019-02-12  5:19       ` James Simmons
2019-02-12 13:56         ` Patrick Farrell
2019-02-12 22:12         ` NeilBrown
2019-02-13  0:19           ` James Simmons
2019-02-13  0:29             ` NeilBrown
2019-02-07  0:03 ` [lustre-devel] [PATCH 12/21] lustre: cl_page.c: remove PINVRNT() NeilBrown
2019-02-08  5:43   ` Andreas Dilger
2019-02-11  4:01   ` James Simmons
2019-02-07  0:03 ` [lustre-devel] [PATCH 21/21] lustre: make exp_refcount in obd_export a refcount_t NeilBrown
2019-02-08  7:07   ` Andreas Dilger
2019-02-11  4:18   ` James Simmons
2019-02-07  0:03 ` [lustre-devel] [PATCH 09/21] lustre: use list_last_entry() throughout NeilBrown
2019-02-08  1:07   ` Andreas Dilger
2019-02-11  1:48   ` James Simmons
2019-02-07  0:03 ` [lustre-devel] [PATCH 15/21] lustre: obdclass: char obd_ioctl_getdata type NeilBrown
2019-02-08  5:56   ` Andreas Dilger
2019-02-11  0:52     ` NeilBrown
2019-02-11  4:03   ` James Simmons
2019-02-07  0:03 ` [lustre-devel] [PATCH 07/21] lustre: use list_first_entry() in lnet/klnds subdirectory NeilBrown
2019-02-08  0:59   ` Andreas Dilger
2019-02-11  0:34     ` NeilBrown
2019-02-11  1:47   ` James Simmons
2019-02-07  0:03 ` [lustre-devel] [PATCH 16/21] lustre: obdclass: normalize a switch statement NeilBrown
2019-02-08  5:57   ` Andreas Dilger
2019-02-11  4:03   ` James Simmons
2019-02-07  0:03 ` [lustre-devel] [PATCH 14/21] lustre: make ccc_users in cl_client_cache a refcount_t NeilBrown
2019-02-08  5:46   ` Andreas Dilger
2019-02-11  4:01   ` James Simmons
2019-02-07  0:03 ` [lustre-devel] [PATCH 20/21] lustre: obdclass: fix module load locking NeilBrown
2019-02-13  1:53   ` James Simmons
2019-02-07  0:03 ` [lustre-devel] [PATCH 19/21] lustre: obdclass: avoid races in class_register_type() NeilBrown
2019-02-08  6:41   ` Andreas Dilger
2019-02-11  0:58     ` NeilBrown
2019-02-12  5:03   ` James Simmons
2019-02-14  3:43     ` NeilBrown
2019-02-07  0:03 ` [lustre-devel] [PATCH 17/21] lustre: obdclass: result of try_module_get() should not be ignored NeilBrown
2019-02-08  5:58   ` Andreas Dilger
2019-02-11  4:22   ` James Simmons
2019-02-11  5:01     ` NeilBrown
2019-02-11  5:09       ` [lustre-devel] [PATCH] lustre: don't manage module refs in obd_class_open/close NeilBrown
2019-02-12  4:17         ` James Simmons
2019-02-07  0:03 ` [lustre-devel] [PATCH 08/21] lustre: use list_first_entry() throughout NeilBrown
2019-02-08  1:06   ` Andreas Dilger
2019-02-11  1:48   ` James Simmons
2019-02-07  0:03 ` [lustre-devel] [PATCH 13/21] lustre: make cp_ref in cl_page a refcount_t NeilBrown
2019-02-08  5:45   ` Andreas Dilger
2019-02-11  4:00   ` James Simmons

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=87va1rgojp.fsf@notabene.neil.brown.name \
    --to=neilb@suse.com \
    --cc=lustre-devel@lists.lustre.org \
    /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.