From: NeilBrown <neilb@suse.de>
To: "Williams, Dan J" <dan.j.williams@intel.com>
Cc: linux-raid@vger.kernel.org
Subject: Re: [md PATCH 08/34] md/raid5: replace sh->lock with an 'active' flag.
Date: Thu, 4 Aug 2011 10:18:10 +1000 [thread overview]
Message-ID: <20110804101810.57505cb7@notabene.brown> (raw)
In-Reply-To: <CABE8wwu9xP4yw_SRH800R9+7E9Q2kAR-hPxaycxKO4Nwc1bsHg@mail.gmail.com>
On Wed, 3 Aug 2011 16:45:35 -0700 "Williams, Dan J"
<dan.j.williams@intel.com> wrote:
> On Wed, Aug 3, 2011 at 4:35 PM, NeilBrown <neilb@suse.de> wrote:
> > On Wed, 3 Aug 2011 15:47:47 -0700 Dan Williams <dan.j.williams@intel.com>
> > wrote:
> >
> >> On Wed, Jul 20, 2011 at 7:32 PM, NeilBrown <neilb@suse.de> wrote:
> >> > sh->lock is now mainly used to ensure that two threads aren't running
> >> > in the locked part of handle_stripe[56] at the same time.
> >> >
> >> > That can more neatly be achieved with an 'active' flag which we set
> >> > while running handle_stripe. If we find the flag is set, we simply
> >> > requeue the stripe for later by setting STRIPE_HANDLE.
> >> >
> >> > For safety we take ->device_lock while examining the state of the
> >> > stripe and creating a summary in 'stripe_head_state / r6_state'.
> >> > This possibly isn't needed but as shared fields like ->toread,
> >> > ->towrite are checked it is safer for now at least.
> >> >
> >> > We leave the label after the old 'unlock' called "unlock" because it
> >> > will disappear in a few patches, so renaming seems pointless.
> >> >
> >> > This leaves the stripe 'locked' for longer as we clear STRIPE_ACTIVE
> >> > later, but that is not a problem.
> >> >
> >>
> >> This removal reminds me of one thing I have wondered about, but to
> >> date have not found the time to measure (maybe someone might beat me
> >> to it if the idea is out there), is what is the overhead of all the
> >> atomic operations that raid5.c generates? If we can guarantee that
> >> certain updates only happen under sh->lock (now STRIPE_ACTIVE) can we
> >> downgrade set_bit and clear_bit to their non-atomic __set_bit and
> >> __clear_bit versions and recover some cpu cycles?
> >>
> >
> > You can only used the unlocked version if you know that no other CPU will
> > change any of the bits in the 'unsigned long'. As STRIPE_HANDLE can be set
> > at any time, I think all accesses to sh->state must be atomic.
> >
> > However 'struct stripe_head_state' is thread-local so setting/clearing flags
> > in ops_request can probably benefit from non-atomic ops.
>
> I was particularly eyeing 'flags' in struct r5dev...
good point.... yes, they would probably be safe.
Async updates only happen when the refcount on the stripe head is != 0,
and handle_stripe only updates things when it own the sole reference.
So the updates in handle_stripe and functions directly call from there should
be safe to use unlocked versions...
NeilBrown
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
next prev parent reply other threads:[~2011-08-04 0:18 UTC|newest]
Thread overview: 71+ messages / expand[flat|nested] mbox.gz Atom feed top
2011-07-21 2:32 [md PATCH 00/34] md patches for 3.1 - part 1 NeilBrown
2011-07-21 2:32 ` [md PATCH 03/34] md/raid10: share pages between read and write bio's during recovery NeilBrown
2011-07-21 2:32 ` [md PATCH 02/34] md/raid10: factor out common bio handling code NeilBrown
2011-07-21 2:32 ` [md PATCH 01/34] md/raid10: get rid of duplicated conditional expression NeilBrown
2011-07-21 2:32 ` [md PATCH 10/34] md/raid5: unify stripe_head_state and r6_state NeilBrown
2011-07-22 4:49 ` Namhyung Kim
2011-07-22 5:15 ` NeilBrown
2011-07-22 5:37 ` NeilBrown
2011-07-22 5:53 ` Namhyung Kim
2011-07-26 6:44 ` Namhyung Kim
2011-07-21 2:32 ` [md PATCH 09/34] md/raid5: move common code into handle_stripe NeilBrown
2011-07-22 4:30 ` Namhyung Kim
2011-07-21 2:32 ` [md PATCH 05/34] md/raid5: get rid of duplicated call to bio_data_dir() NeilBrown
2011-07-21 2:32 ` [md PATCH 08/34] md/raid5: replace sh->lock with an 'active' flag NeilBrown
2011-07-22 4:27 ` Namhyung Kim
2011-07-22 4:49 ` NeilBrown
2011-07-22 5:03 ` Namhyung Kim
2011-08-03 22:47 ` Dan Williams
2011-08-03 23:35 ` NeilBrown
2011-08-03 23:45 ` Williams, Dan J
2011-08-04 0:18 ` NeilBrown [this message]
2011-07-21 2:32 ` [md PATCH 07/34] md/raid5: Protect some more code with ->device_lock NeilBrown
2011-07-22 3:54 ` Namhyung Kim
2011-07-21 2:32 ` [md PATCH 04/34] md/raid5: use kmem_cache_zalloc() NeilBrown
2011-07-21 2:32 ` [md PATCH 11/34] md/raid5: add some more fields to stripe_head_state NeilBrown
2011-07-22 5:31 ` Namhyung Kim
2011-07-26 1:35 ` NeilBrown
2011-07-21 2:32 ` [md PATCH 06/34] md/raid5: Remove use of sh->lock in sync_request NeilBrown
2011-07-22 3:39 ` Namhyung Kim
2011-07-21 2:32 ` [md PATCH 19/34] md/raid5: move some more common code into handle_stripe NeilBrown
2011-07-22 9:29 ` Namhyung Kim
2011-07-26 1:59 ` NeilBrown
2011-07-21 2:32 ` [md PATCH 13/34] md/raid5: Move code for finishing a reconstruction " NeilBrown
2011-07-22 7:09 ` Namhyung Kim
2011-07-26 1:44 ` NeilBrown
2011-07-21 2:32 ` [md PATCH 12/34] md/raid5: move stripe_head_state and more code " NeilBrown
2011-07-22 5:41 ` Namhyung Kim
2011-07-21 2:32 ` [md PATCH 17/34] md/raid5: unite handle_stripe_dirtying5 and handle_stripe_dirtying6 NeilBrown
2011-07-22 9:10 ` Namhyung Kim
2011-07-26 1:52 ` NeilBrown
2011-07-26 2:41 ` H. Peter Anvin
2011-07-26 9:40 ` David Brown
2011-07-26 13:23 ` Namhyung Kim
2011-07-26 15:01 ` David Brown
2011-07-21 2:32 ` [md PATCH 18/34] md/raid5: move more common code into handle_stripe NeilBrown
2011-07-22 9:20 ` Namhyung Kim
2011-07-21 2:32 ` [md PATCH 14/34] md/raid5: move more code into common handle_stripe NeilBrown
2011-07-22 7:32 ` Namhyung Kim
2011-07-26 1:48 ` NeilBrown
2011-07-21 2:32 ` [md PATCH 15/34] md/raid5: rearrange a test in fetch_block6 NeilBrown
2011-07-22 7:37 ` Namhyung Kim
2011-07-21 2:32 ` [md PATCH 16/34] md/raid5: unite fetch_block5 and fetch_block6 NeilBrown
2011-07-22 8:24 ` Namhyung Kim
2011-07-21 2:32 ` [md PATCH 24/34] md: remove ro check in md_check_recovery() NeilBrown
2011-07-21 2:32 ` [md PATCH 22/34] md/raid: use printk_ratelimited instead of printk_ratelimit NeilBrown
2011-07-21 2:32 ` [md PATCH 20/34] md/raid5: finalise new merged handle_stripe NeilBrown
2011-07-22 9:36 ` Namhyung Kim
2011-07-26 2:02 ` NeilBrown
2011-07-26 4:50 ` Namhyung Kim
2011-07-21 2:32 ` [md PATCH 25/34] md: change managed of recovery_disabled NeilBrown
2011-07-21 2:32 ` [md PATCH 26/34] md/raid10: Make use of new recovery_disabled handling NeilBrown
2011-07-21 2:32 ` [md PATCH 21/34] md: use proper little-endian bitops NeilBrown
2011-07-21 2:32 ` [md PATCH 27/34] md/raid10: Improve decision on whether to fail a device with a read error NeilBrown
2011-07-21 2:32 ` [md PATCH 23/34] md: introduce link/unlink_rdev() helpers NeilBrown
2011-07-21 2:32 ` [md PATCH 31/34] md/raid10: move rdev->corrected_errors counting NeilBrown
2011-07-21 2:32 ` [md PATCH 34/34] MD bitmap: Revert DM dirty log hooks NeilBrown
2011-07-21 2:32 ` [md PATCH 33/34] MD: raid1 s/sysfs_notify_dirent/sysfs_notify_dirent_safe NeilBrown
2011-07-21 2:32 ` [md PATCH 32/34] md/raid5: Avoid BUG caused by multiple failures NeilBrown
2011-07-21 2:32 ` [md PATCH 28/34] md: get rid of unnecessary casts on page_address() NeilBrown
2011-07-21 2:32 ` [md PATCH 29/34] md/raid1: move rdev->corrected_errors counting NeilBrown
2011-07-21 2:32 ` [md PATCH 30/34] md/raid5: " NeilBrown
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=20110804101810.57505cb7@notabene.brown \
--to=neilb@suse.de \
--cc=dan.j.williams@intel.com \
--cc=linux-raid@vger.kernel.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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).