From: Edward Shishkin <edward.shishkin@gmail.com>
To: Ivan Shapovalov <intelfx100@gmail.com>, reiserfs-devel@vger.kernel.org
Subject: Re: [PATCHv2 2/3] reiser4: block_alloc: sanitize grab_enabled modifications.
Date: Sun, 16 Nov 2014 10:53:03 +0100 [thread overview]
Message-ID: <546873FF.6040804@gmail.com> (raw)
In-Reply-To: <2228941.DWoZ5v2hiQ@intelfx-laptop>
On 11/16/2014 06:38 AM, Ivan Shapovalov wrote:
> On Thursday 23 October 2014 at 11:18:04, Ivan Shapovalov wrote:
>> - check and modify ctx->grab_enabled in reiser4_grab_space(), not reiser4_grab()
>> - do not re-enable grab before doing second attempt in BA_CAN_COMMIT sequence
>> (it is unneeded given the first change)
> ...testing has revealed that this is wrong. Please see below.
>
>> Signed-off-by: Ivan Shapovalov <intelfx100@gmail.com>
>> ---
>> fs/reiser4/block_alloc.c | 22 ++++++++++++----------
>> 1 file changed, 12 insertions(+), 10 deletions(-)
>>
>> diff --git a/fs/reiser4/block_alloc.c b/fs/reiser4/block_alloc.c
>> index 7f9f910..9339de7 100644
>> --- a/fs/reiser4/block_alloc.c
>> +++ b/fs/reiser4/block_alloc.c
>> @@ -270,12 +270,6 @@ reiser4_grab(reiser4_context * ctx, __u64 count, reiser4_ba_flags_t flags)
>>
>> assert("vs-1276", ctx == get_current_context());
>>
>> - /* Do not grab anything on ro-mounted fs. */
>> - if (rofs_super(ctx->super)) {
>> - ctx->grab_enabled = 0;
>> - return 0;
>> - }
>> -
>> sbinfo = get_super_private(ctx->super);
>>
>> spin_lock_reiser4_super(sbinfo);
>> @@ -300,9 +294,6 @@ reiser4_grab(reiser4_context * ctx, __u64 count, reiser4_ba_flags_t flags)
>>
>> assert("nikita-2986", reiser4_check_block_counters(ctx->super));
>>
>> - /* disable grab space in current context */
>> - ctx->grab_enabled = 0;
>> -
>> unlock_and_ret:
>> spin_unlock_reiser4_super(sbinfo);
>>
>> @@ -321,6 +312,12 @@ int reiser4_grab_space(__u64 count, reiser4_ba_flags_t flags)
>> if (!(flags & BA_FORCE) && !is_grab_enabled(ctx))
>> return 0;
>>
>> + /* Do not grab anything on ro-mounted fs. */
>> + if (rofs_super(ctx->super)) {
>> + ctx->grab_enabled = 0;
>> + return 0;
>> + }
>> +
>> ret = reiser4_grab(ctx, count, flags);
>> if (ret == -ENOSPC) {
>>
>> @@ -328,10 +325,15 @@ int reiser4_grab_space(__u64 count, reiser4_ba_flags_t flags)
>> present */
>> if (flags & BA_CAN_COMMIT) {
>> txnmgr_force_commit_all(ctx->super, 0);
>> - ctx->grab_enabled = 1;
>> ret = reiser4_grab(ctx, count, flags);
>> }
> Here, txnmgr_force_commit_all() performs space grabbing (with BA_FORCE flag)
> and as such indirectly sets ctx->grab_enabled to 0.
>
> Initial problem.
> Situation: ctx->grab_enabled == 0
> Call: reiser4_grab_space(toomany, BA_FORCE | BA_CAN_COMMIT)
> * first reiser4_grab() returns -ENOSPC
> * txnmgr_force_commit_all() is called
> * ctx->grab_enabled is set to 1 by us
> * second reiser4_grab() returns -ENOSPC as well
> Result: ctx->grab_enabled == 1
>
> Problem with this patch.
> Situation: ctx->grab_enabled == 1
> Call: reiser4_grab_space(toomany, BA_CAN_COMMIT)
> * first reiser4_grab() returns -ENOSPC
> * txnmgr_force_commit_all() is called and sets ctx->grab_enabled to 0
> * second reiser4_grab() returns -ENOSPC as well
> Result: no grab performed, but ctx->grab_enabled == 0.
>
> Solution 1:
> save and restore ctx->grab_enabled across call to txnmgr_force_commit_all();
>
> Solution 2:
> make txnmgr_force_commit_all() not touch ctx->grab_enabled,
> i. e. set ctx->grab_enabled to 0 only if !BA_FORCE.
>
> What would you suggest?
Could you please remind, what problem
this patch series (sanitize grab_enabled) fixes?
Thanks,
Edward.
>
next prev parent reply other threads:[~2014-11-16 9:53 UTC|newest]
Thread overview: 17+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-10-23 7:18 [PATCHv2 0/3] reiser4: space grabbing fixes Ivan Shapovalov
2014-10-23 7:18 ` [PATCHv2 1/3] reiser4: block_alloc: improve error handling in reiser4_grab_reserved() Ivan Shapovalov
2014-10-23 7:18 ` [PATCHv2 2/3] reiser4: block_alloc: sanitize grab_enabled modifications Ivan Shapovalov
2014-11-16 5:38 ` Ivan Shapovalov
2014-11-16 9:53 ` Edward Shishkin [this message]
2014-11-16 10:45 ` Edward Shishkin
2014-11-16 11:49 ` Edward Shishkin
2014-11-16 10:45 ` Edward Shishkin
2014-11-19 1:39 ` Ivan Shapovalov
2014-10-23 7:18 ` [PATCHv2 3/3] reiser4: do not mess with grab_enabled; instead, use BA_FORCE Ivan Shapovalov
2014-10-23 7:20 ` [PATCHv2 0/3] reiser4: space grabbing fixes Ivan Shapovalov
2014-12-10 12:27 ` Ivan Shapovalov
2014-12-10 12:52 ` Edward Shishkin
2014-12-10 13:30 ` Ivan Shapovalov
2014-12-10 21:51 ` Edward Shishkin
2014-12-12 22:19 ` Ivan Shapovalov
2014-12-13 23:41 ` Edward Shishkin
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=546873FF.6040804@gmail.com \
--to=edward.shishkin@gmail.com \
--cc=intelfx100@gmail.com \
--cc=reiserfs-devel@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).