git@vger.kernel.org mailing list mirror (one of many)
 help / color / mirror / code / Atom feed
From: "Martin Ågren" <martin.agren@gmail.com>
To: git@vger.kernel.org
Cc: "Michael Haggerty" <mhagger@alum.mit.edu>,
	"Jeff King" <peff@peff.net>, "Paul Tan" <pyokagan@gmail.com>,
	"Christian Couder" <christian.couder@gmail.com>,
	"Nguyễn Thái Ngọc Duy" <pclouds@gmail.com>,
	"Junio C Hamano" <gitster@pobox.com>
Subject: [PATCH v3 11/12] read-cache: leave lock in right state in `write_locked_index()`
Date: Fri,  6 Oct 2017 22:12:13 +0200	[thread overview]
Message-ID: <ac3c5dd103de8364b0289a240d38782eafca2739.1507320376.git.martin.agren@gmail.com> (raw)
In-Reply-To: <cover.1507320376.git.martin.agren@gmail.com>

If the original version of `write_locked_index()` returned with an
error, it didn't roll back the lockfile unless the error occured at the
very end, during closing/committing. See commit 03b866477 (read-cache:
new API write_locked_index instead of write_index/write_cache,
2014-06-13).

In commit 9f41c7a6b (read-cache: close index.lock in do_write_index,
2017-04-26), we learned to close the lock slightly earlier in the
callstack. That was mostly a side-effect of lockfiles being implemented
using temporary files, but didn't cause any real harm.

Recently, commit 076aa2cbd (tempfile: auto-allocate tempfiles on heap,
2017-09-05) introduced a subtle bug. If the temporary file is deleted
(i.e., the lockfile is rolled back), the tempfile-pointer in the `struct
lock_file` will be left dangling. Thus, an attempt to reuse the
lockfile, or even just to roll it back, will induce undefined behavior
-- most likely a crash.

Besides not crashing, we clearly want to make things consistent. The
guarantees which the lockfile-machinery itself provides is A) if we ask
to commit and it fails, roll back, and B) if we ask to close and it
fails, do _not_ roll back. Let's do the same for consistency.

Do not delete the temporary file in `do_write_index()`. One of its
callers, `write_locked_index()` will thereby avoid rolling back the
lock. The other caller, `write_shared_index()`, will delete its
temporary file anyway. Both of these callers will avoid undefined
behavior (crashing).

Teach `write_locked_index(..., COMMIT_LOCK)` to roll back the lock
before returning. If we have already succeeded and committed, it will be
a noop. Simplify the existing callers where we now have a superfluous
call to `rollback_lockfile()`. That should keep future readers from
wondering why the callers are inconsistent.

Signed-off-by: Martin Ågren <martin.agren@gmail.com>
---
v3: Maybe the commit message wasn't too long, but it was loong. For
example, some of it just duplicated stuff from the previous patch.

 builtin/difftool.c |  1 -
 cache.h            |  4 ++++
 merge.c            |  4 +---
 read-cache.c       | 14 ++++++++------
 sequencer.c        |  1 -
 5 files changed, 13 insertions(+), 11 deletions(-)

diff --git a/builtin/difftool.c b/builtin/difftool.c
index b2d3ba753..bcc79d188 100644
--- a/builtin/difftool.c
+++ b/builtin/difftool.c
@@ -616,7 +616,6 @@ static int run_dir_diff(const char *extcmd, int symlinks, const char *prefix,
 			if (hold_lock_file_for_update(&lock, buf.buf, 0) < 0 ||
 			    write_locked_index(&wtindex, &lock, COMMIT_LOCK)) {
 				ret = error("could not write %s", buf.buf);
-				rollback_lock_file(&lock);
 				goto finish;
 			}
 			changed_files(&wt_modified, buf.buf, workdir);
diff --git a/cache.h b/cache.h
index 21a6856c5..0e26224b9 100644
--- a/cache.h
+++ b/cache.h
@@ -616,6 +616,10 @@ extern int read_index_unmerged(struct index_state *);
  * split index to the lockfile. If the temporary file for the shared
  * index cannot be created, fall back to the behavior described in
  * the previous paragraph.
+ *
+ * With `COMMIT_LOCK`, the lock is always committed or rolled back.
+ * Without it, the lock is closed, but neither committed nor rolled
+ * back.
  */
 extern int write_locked_index(struct index_state *, struct lock_file *lock, unsigned flags);
 
diff --git a/merge.c b/merge.c
index a18a452b5..e5d796c9f 100644
--- a/merge.c
+++ b/merge.c
@@ -91,9 +91,7 @@ int checkout_fast_forward(const struct object_id *head,
 	}
 	if (unpack_trees(nr_trees, t, &opts))
 		return -1;
-	if (write_locked_index(&the_index, &lock_file, COMMIT_LOCK)) {
-		rollback_lock_file(&lock_file);
+	if (write_locked_index(&the_index, &lock_file, COMMIT_LOCK))
 		return error(_("unable to write new index file"));
-	}
 	return 0;
 }
diff --git a/read-cache.c b/read-cache.c
index c7aa3632a..0d8d2dede 100644
--- a/read-cache.c
+++ b/read-cache.c
@@ -2182,9 +2182,8 @@ static int has_racy_timestamp(struct index_state *istate)
 void update_index_if_able(struct index_state *istate, struct lock_file *lockfile)
 {
 	if ((istate->cache_changed || has_racy_timestamp(istate)) &&
-	    verify_index(istate) &&
-	    write_locked_index(istate, lockfile, COMMIT_LOCK))
-		rollback_lock_file(lockfile);
+	    verify_index(istate))
+		write_locked_index(istate, lockfile, COMMIT_LOCK);
 }
 
 /*
@@ -2321,7 +2320,6 @@ static int do_write_index(struct index_state *istate, struct tempfile *tempfile,
 		return -1;
 	if (close_tempfile_gently(tempfile)) {
 		error(_("could not close '%s'"), tempfile->filename.buf);
-		delete_tempfile(&tempfile);
 		return -1;
 	}
 	if (stat(tempfile->filename.buf, &st))
@@ -2501,7 +2499,8 @@ int write_locked_index(struct index_state *istate, struct lock_file *lock,
 	    (istate->cache_changed & ~EXTMASK)) {
 		if (si)
 			hashclr(si->base_sha1);
-		return do_write_locked_index(istate, lock, flags);
+		ret = do_write_locked_index(istate, lock, flags);
+		goto out;
 	}
 
 	if (getenv("GIT_TEST_SPLIT_INDEX")) {
@@ -2517,7 +2516,7 @@ int write_locked_index(struct index_state *istate, struct lock_file *lock,
 	if (new_shared_index) {
 		ret = write_shared_index(istate, lock, flags);
 		if (ret)
-			return ret;
+			goto out;
 	}
 
 	ret = write_split_index(istate, lock, flags);
@@ -2526,6 +2525,9 @@ int write_locked_index(struct index_state *istate, struct lock_file *lock,
 	if (!ret && !new_shared_index)
 		freshen_shared_index(sha1_to_hex(si->base_sha1), 1);
 
+out:
+	if (flags & COMMIT_LOCK)
+		rollback_lock_file(lock);
 	return ret;
 }
 
diff --git a/sequencer.c b/sequencer.c
index 60636ce54..d56c38081 100644
--- a/sequencer.c
+++ b/sequencer.c
@@ -1183,7 +1183,6 @@ static int read_and_refresh_cache(struct replay_opts *opts)
 	refresh_index(&the_index, REFRESH_QUIET|REFRESH_UNMERGED, NULL, NULL, NULL);
 	if (the_index.cache_changed && index_fd >= 0) {
 		if (write_locked_index(&the_index, &index_lock, COMMIT_LOCK)) {
-			rollback_lock_file(&index_lock);
 			return error(_("git %s: failed to refresh the index"),
 				_(action_name(opts)));
 		}
-- 
2.15.0.rc0


  parent reply	other threads:[~2017-10-06 20:13 UTC|newest]

Thread overview: 69+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-10-01 14:56 [PATCH 00/11] various lockfile-leaks and -fixes Martin Ågren
2017-10-01 14:56 ` [PATCH 01/11] sha1_file: do not leak `lock_file` Martin Ågren
2017-10-02  5:26   ` Jeff King
2017-10-02 10:15     ` Martin Ågren
2017-10-01 14:56 ` [PATCH 02/11] treewide: prefer lockfiles on the stack Martin Ågren
2017-10-02  3:37   ` Junio C Hamano
2017-10-02  4:12     ` Martin Ågren
2017-10-02  5:34   ` Jeff King
2017-10-01 14:56 ` [PATCH 03/11] lockfile: fix documentation on `close_lock_file_gently()` Martin Ågren
2017-10-02  5:35   ` Jeff King
2017-10-01 14:56 ` [PATCH 04/11] tempfile: fix documentation on `delete_tempfile()` Martin Ågren
2017-10-02  5:38   ` Jeff King
2017-10-01 14:56 ` [PATCH 05/11] cache-tree: simplify locking logic Martin Ågren
2017-10-02  3:40   ` Junio C Hamano
2017-10-02  5:41   ` Jeff King
2017-10-01 14:56 ` [PATCH 06/11] apply: move lockfile into `apply_state` Martin Ågren
2017-10-02  5:48   ` Jeff King
2017-10-01 14:56 ` [PATCH 07/11] apply: remove `newfd` from `struct apply_state` Martin Ågren
2017-10-02  5:50   ` Jeff King
2017-10-01 14:56 ` [PATCH 08/11] cache.h: document `write_locked_index()` Martin Ågren
2017-10-01 14:56 ` [PATCH 09/11] read-cache: require flags for `write_locked_index()` Martin Ågren
2017-10-02  3:49   ` Junio C Hamano
2017-10-02  4:14     ` Martin Ågren
2017-10-02 10:16       ` Martin Ågren
2017-10-02  6:00   ` Jeff King
2017-10-01 14:56 ` [PATCH 10/11] read-cache: don't leave dangling pointer in `do_write_index()` Martin Ågren
2017-10-02  6:15   ` Jeff King
2017-10-02  6:20     ` Jeff King
2017-10-01 14:56 ` [PATCH 11/11] read-cache: roll back lock on error with `COMMIT_LOCK` Martin Ågren
2017-10-02  4:01   ` Junio C Hamano
2017-10-02  2:37 ` [PATCH 00/11] various lockfile-leaks and -fixes Junio C Hamano
2017-10-02  6:22 ` Jeff King
2017-10-02  6:30   ` Junio C Hamano
2017-10-02 10:19     ` Martin Ågren
2017-10-03  6:21       ` Junio C Hamano
2017-10-05 20:32         ` [PATCH v2 00/12] " Martin Ågren
2017-10-05 20:32           ` [PATCH v2 01/12] sha1_file: do not leak `lock_file` Martin Ågren
2017-10-06  1:17             ` Junio C Hamano
2017-10-05 20:32           ` [PATCH v2 02/12] treewide: prefer lockfiles on the stack Martin Ågren
2017-10-05 20:32           ` [PATCH v2 03/12] lockfile: fix documentation on `close_lock_file_gently()` Martin Ågren
2017-10-05 20:32           ` [PATCH v2 04/12] tempfile: fix documentation on `delete_tempfile()` Martin Ågren
2017-10-05 20:32           ` [PATCH v2 05/12] checkout-index: simplify locking logic Martin Ågren
2017-10-06  1:21             ` Junio C Hamano
2017-10-05 20:32           ` [PATCH v2 06/12] cache-tree: " Martin Ågren
2017-10-05 20:32           ` [PATCH v2 07/12] apply: move lockfile into `apply_state` Martin Ågren
2017-10-05 20:32           ` [PATCH v2 08/12] apply: remove `newfd` from `struct apply_state` Martin Ågren
2017-10-05 20:32           ` [PATCH v2 09/12] cache.h: document `write_locked_index()` Martin Ågren
2017-10-05 20:32           ` [PATCH v2 10/12] read-cache: drop explicit `CLOSE_LOCK`-flag Martin Ågren
2017-10-06  1:39             ` Junio C Hamano
2017-10-06 11:02               ` Martin Ågren
2017-10-05 20:32           ` [PATCH v2 11/12] read-cache: leave lock in right state in `write_locked_index()` Martin Ågren
2017-10-06  2:01             ` Junio C Hamano
2017-10-06 11:04               ` Martin Ågren
2017-10-06 12:02                 ` Junio C Hamano
2017-10-06 19:44                   ` Martin Ågren
2017-10-06 20:12                     ` [PATCH v3 00/12] Re: various lockfile-leaks and -fixes Martin Ågren
2017-10-06 20:12                       ` [PATCH v3 01/12] sha1_file: do not leak `lock_file` Martin Ågren
2017-10-06 20:12                       ` [PATCH v3 02/12] treewide: prefer lockfiles on the stack Martin Ågren
2017-10-06 20:12                       ` [PATCH v3 03/12] lockfile: fix documentation on `close_lock_file_gently()` Martin Ågren
2017-10-06 20:12                       ` [PATCH v3 04/12] tempfile: fix documentation on `delete_tempfile()` Martin Ågren
2017-10-06 20:12                       ` [PATCH v3 05/12] checkout-index: simplify locking logic Martin Ågren
2017-10-06 20:12                       ` [PATCH v3 06/12] cache-tree: " Martin Ågren
2017-10-06 20:12                       ` [PATCH v3 07/12] apply: move lockfile into `apply_state` Martin Ågren
2017-10-06 20:12                       ` [PATCH v3 08/12] apply: remove `newfd` from `struct apply_state` Martin Ågren
2017-10-06 20:12                       ` [PATCH v3 09/12] cache.h: document `write_locked_index()` Martin Ågren
2017-10-06 20:12                       ` [PATCH v3 10/12] read-cache: drop explicit `CLOSE_LOCK`-flag Martin Ågren
2017-10-06 20:12                       ` Martin Ågren [this message]
2017-10-06 20:12                       ` [PATCH v3 12/12] read_cache: roll back lock in `update_index_if_able()` Martin Ågren
2017-10-05 20:32           ` [PATCH v2 " Martin Ågren

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

  List information: http://vger.kernel.org/majordomo-info.html

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=ac3c5dd103de8364b0289a240d38782eafca2739.1507320376.git.martin.agren@gmail.com \
    --to=martin.agren@gmail.com \
    --cc=christian.couder@gmail.com \
    --cc=git@vger.kernel.org \
    --cc=gitster@pobox.com \
    --cc=mhagger@alum.mit.edu \
    --cc=pclouds@gmail.com \
    --cc=peff@peff.net \
    --cc=pyokagan@gmail.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.
Code repositories for project(s) associated with this public inbox

	https://80x24.org/mirrors/git.git

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 read-only IMAP folder(s) and NNTP newsgroup(s).