[PATCH v2 0/8] Fix atomicity and avoid fd exhaustion in ref transactions

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



This is v2 of [1], formerly called

    Avoid file descriptor exhaustion in ref_transaction_commit()

As a reminder, this patch series fixes some cases where `git
update-ref --stdin` (and, on 2.4+, `git push --atomic`) could fail to
be atomic, and avoids file descriptor exhaustion while carrying out
ref transactions by closing lockfiles promptly.

Thanks to Peff, Stefan, Junio, and Eric for their comments on v1.

The main change is that this version is rebased onto maint-2.2 as per
Junio's wish [2]. This was not trivial because of the changes in the
area between 2.3 and 2.4, especially 'mh/expire-updateref-fixes',
'mh/refs-have-new', and 'mh/reflog-expire'.

Other changes:

* Remove Stefan's

      bc31f46 refs.c: remove lock_fd from struct ref_lock (2015-04-16)

  from the series, as it is not needed for this fix and was causing
  extra conflicts when backporting and merging forward. This patch is
  still a good idea, but there is no need to backport it.

* Change log message of the last patch to describe both of the bugs
  that it fixes.

* Many other small tweaks.

This patch series is available from my GitHub account [3] as branch
'write-refs-sooner'. Please note that the branch there is applied on
top of a cherry-pick of

      f6786c8 http-push: trim trailing newline from remote symref (2015-01-12)

, which was needed to get test t5540 to pass.

The following other branches, also from my GitHub repo, might be
useful:

* 'write-refs-sooner-2.3' -- suggested merge of the change to 'maint'.

* 'write-refs-sooner-master' -- suggested merge of the change to
  'master'.

* 'write-refs-sooner-rebased-2.3' and
  'write-refs-sooner-rebased-master' -- rebases of 'write-refs-sooner'
  onto 'maint' and 'master' respectively, in case anybody is
  interested to see how the individual patches would look if
  implemented natively on these branches.

[1] http://thread.gmane.org/gmane.comp.version-control.git/267735
[2] http://article.gmane.org/gmane.comp.version-control.git/267799
[3] https://github.com/mhagger/git

Michael Haggerty (6):
  write_ref_to_lockfile(): new function, extracted from write_ref_sha1()
  commit_ref_update(): new function, extracted from write_ref_sha1()
  rename_ref(): inline calls to write_ref_sha1() from this function
  ref_transaction_commit(): inline calls to write_ref_sha1()
  ref_transaction_commit(): remove the local flags variable
  ref_transaction_commit(): fix atomicity and avoid fd exhaustion

Stefan Beller (2):
  update-ref: test handling large transactions properly
  t7004: rename ULIMIT test prerequisite to ULIMIT_STACK_SIZE

 refs.c                | 99 +++++++++++++++++++++++++++++++++++++--------------
 t/t1400-update-ref.sh | 28 +++++++++++++++
 t/t7004-tag.sh        |  4 +--
 3 files changed, 103 insertions(+), 28 deletions(-)

-- 
2.1.4

--
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html




[Index of Archives]     [Linux Kernel Development]     [Gcc Help]     [IETF Annouce]     [DCCP]     [Netdev]     [Networking]     [Security]     [V4L]     [Bugtraq]     [Yosemite]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Linux SCSI]     [Fedora Users]