git/builtin
Jeff King ab791dd138 index-pack: fix race condition with duplicate bases
When we are resolving deltas in an indexed pack, we do it by
first selecting a potential base (either one stored in full
in the pack, or one created by resolving another delta), and
then resolving any deltas that use that base.  When we
resolve a particular delta, we flip its "real_type" field
from OBJ_{REF,OFS}_DELTA to whatever the real type is.

We assume that traversing the objects this way will visit
each delta only once. This is correct for most packs; we
visit the delta only when we process its base, and each
object (and thus each base) appears only once. However, if a
base object appears multiple times in the pack, we will try
to resolve any deltas based on it once for each instance.

We can detect this case by noting that a delta we are about
to resolve has already had its real_type field flipped, and
we already do so with an assert().  However, if multiple
threads are in use, we may race with another thread on
comparing and flipping the field. We need to synchronize the
access.

The right mechanism for doing this is a compare-and-swap (we
atomically "claim" the delta for our own and find out
whether our claim was successful). We can implement this
in C by using a pthread mutex to protect the operation. This
is not the fastest way of doing a compare-and-swap; many
processors provide instructions for this, and gcc and other
compilers provide builtins to access them. However, some
experiments showed that lock contention does not cause a
significant slowdown here. Adding c-a-s support for many
compilers would increase the maintenance burden (and we
would still end up including the pthread version as a
fallback).

Note that we only need to touch the OBJ_REF_DELTA codepath
here. An OBJ_OFS_DELTA object points to its base using an
offset, and therefore has only one base, even if another
copy of that base object appears in the pack (we do still
touch it briefly because the setting of real_type is
factored out of resolve_data).

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2014-08-29 14:50:43 -07:00
..
add.c add: don't complain when adding empty project root 2013-12-26 10:46:26 -08:00
annotate.c annotate: use argv_array 2014-07-16 11:10:11 -07:00
apply.c Replace deprecated OPT_BOOLEAN by OPT_BOOL 2013-08-05 11:32:19 -07:00
archive.c pkt-line: provide a LARGE_PACKET_MAX static buffer 2013-02-20 13:42:22 -08:00
bisect--helper.c Replace deprecated OPT_BOOLEAN by OPT_BOOL 2013-08-05 11:32:19 -07:00
blame.c Remove the line length limit for graft files 2013-12-27 16:46:25 -08:00
branch.c Merge branch 'jx/branch-vv-always-compare-with-upstream' 2013-11-18 12:24:49 -08:00
bundle.c
cat-file.c cat-file: handle --batch format with missing type/size 2013-12-12 11:31:25 -08:00
check-attr.c Merge branch 'sb/parseopt-boolean-removal' 2013-09-04 12:39:03 -07:00
check-ignore.c Merge branch 'dw/check-ignore-sans-index' 2013-09-20 12:37:32 -07:00
check-mailmap.c builtin: add git-check-mailmap command 2013-07-13 10:19:37 -07:00
check-ref-format.c
checkout-index.c Merge branch 'nd/lift-path-max' 2013-10-30 12:10:56 -07:00
checkout.c Merge branch 'mm/checkout-auto-track-fix' into maint 2013-11-07 14:36:59 -08:00
clean.c enums: remove trailing ',' after last item in enum 2014-07-02 12:37:05 -07:00
clone.c Merge branch 'nd/transport-positive-depth-only' into maint 2014-02-05 13:58:52 -08:00
column.c i18n: column: mark parseopt strings for translation 2012-08-20 12:23:16 -07:00
commit-tree.c commit-tree: document -S option consistently 2013-03-25 15:01:22 -07:00
commit.c commit -v: strip diffs and submodule shortlogs from the commit message 2013-12-05 14:39:11 -08:00
config.c Merge branch 'jk/config-int-range-check' 2013-09-12 14:41:00 -07:00
count-objects.c count-objects: add -H option to humanize sizes 2013-04-10 13:27:26 -07:00
credential.c git credential fill: output the whole 'struct credential' 2012-06-25 11:56:24 -07:00
describe.c describe: trivial style fixes 2013-10-31 13:47:35 -07:00
diff-files.c convert read_cache_preload() to take struct pathspec 2013-07-15 10:56:08 -07:00
diff-index.c convert read_cache_preload() to take struct pathspec 2013-07-15 10:56:08 -07:00
diff-tree.c
diff.c Merge branch 'nd/magic-pathspec' into maint 2013-12-17 11:21:34 -08:00
fast-export.c Merge branch 'fc/fast-export' 2013-09-17 11:42:31 -07:00
fetch-pack.c git fetch-pack: add --diag-url 2013-12-09 14:54:47 -08:00
fetch.c Merge branch 'nd/transport-positive-depth-only' into maint 2014-02-05 13:58:52 -08:00
fmt-merge-msg.c Merge branch 'rt/commentchar-fmt-merge-msg' into maint 2013-04-26 11:10:47 -07:00
for-each-ref.c for-each-ref: remove unused variable 2013-12-30 12:23:51 -08:00
fsck.c Merge branch 'jk/free-tree-buffer' 2013-09-17 11:37:33 -07:00
gc.c gc: notice gc processes run by other users 2014-01-02 16:15:29 -08:00
grep.c Merge branch 'mg/more-textconv' 2013-10-23 13:21:31 -07:00
hash-object.c hash-object: replace stdin parsing OPT_BOOLEAN by OPT_COUNTUP 2013-08-07 08:30:55 -07:00
help.c Merge branch 'ph/builtin-srcs-are-in-subdir-these-days' into maint 2013-07-21 22:51:29 -07:00
index-pack.c index-pack: fix race condition with duplicate bases 2014-08-29 14:50:43 -07:00
init-db.c i18n: init-db: mark parseopt strings for translation 2012-08-20 12:23:17 -07:00
log.c Merge branch 'mg/more-textconv' 2013-10-23 13:21:31 -07:00
ls-files.c Merge branch 'jc/ls-files-killed-optim' into maint 2013-10-23 13:33:08 -07:00
ls-remote.c ls-remote: document the '--get-url' option 2012-09-07 10:58:35 -07:00
ls-tree.c Merge branch 'jl/submodule-mv' 2013-09-09 14:36:15 -07:00
mailinfo.c Merge branch 'jc/same-encoding' into maint 2012-12-07 14:10:56 -08:00
mailsplit.c mailsplit: sort maildir filenames more cleverly 2013-03-02 22:52:44 -08:00
merge-base.c Merge branch 'bm/merge-base-octopus-dedup' into maint 2014-02-13 13:38:59 -08:00
merge-file.c Replace deprecated OPT_BOOLEAN by OPT_BOOL 2013-08-05 11:32:19 -07:00
merge-index.c Convert "struct cache_entry *" to "const ..." wherever possible 2013-07-09 09:12:48 -07:00
merge-ours.c
merge-recursive.c
merge-tree.c merge-tree: handle directory/empty conflict correctly 2013-05-06 22:17:00 -07:00
merge.c merge: simplify ff-only option 2013-10-31 11:12:24 -07:00
mktag.c
mktree.c i18n: mktree: mark parseopt strings for translation 2012-08-20 12:23:18 -07:00
mv.c mv: let 'git mv file no-such-dir/' error out on Windows, too 2014-01-10 11:28:12 -08:00
name-rev.c branch, commit, name-rev: ease up boolean conditions 2013-08-07 08:30:30 -07:00
notes.c Replace deprecated OPT_BOOLEAN by OPT_BOOL 2013-08-05 11:32:19 -07:00
pack-objects.c Merge branch 'jc/pack-objects' 2013-10-23 13:21:26 -07:00
pack-redundant.c Fix sizeof usage in get_permutations 2012-12-13 11:13:44 -08:00
pack-refs.c pack-refs: merge code from pack-refs.{c,h} into refs.{c,h} 2013-05-01 15:33:11 -07:00
patch-id.c
prune-packed.c prune-packed: avoid implying "1" is DRY_RUN in prune_packed_objects() 2013-05-28 09:20:54 -07:00
prune.c Merge branch 'nd/prune-packed-dryrun-verbose' 2013-06-06 12:17:52 -07:00
push.c Merge branch 'nd/push-no-thin' 2013-09-11 14:56:59 -07:00
read-tree.c C: have space around && and || operators 2013-10-16 10:26:39 -07:00
receive-pack.c Merge branch 'bk/refs-multi-update' 2013-09-20 12:36:12 -07:00
reflog.c Merge branch 'bk/refs-multi-update' 2013-09-20 12:36:12 -07:00
remote-ext.c
remote-fd.c
remote.c Merge branch 'po/remote-set-head-usage' 2013-10-14 11:07:29 -07:00
repack.c Merge branch 'sb/repack-in-c' into maint 2014-02-13 13:38:09 -08:00
replace.c Merge branch 'cc/replace-with-the-same-type' 2013-09-24 23:35:24 -07:00
rerere.c rerere: convert to use parse_pathspec 2013-07-15 10:56:07 -07:00
reset.c reset: pass real rev name to add--interactive 2013-10-25 14:54:18 -07:00
rev-list.c C: have space around && and || operators 2013-10-16 10:26:39 -07:00
rev-parse.c rev-parse: be more careful with munging arguments 2013-12-09 14:39:16 -08:00
revert.c Merge branch 'hu/cherry-pick-previous-branch' 2013-10-23 13:21:35 -07:00
rm.c rm: re-use parse_pathspec's trailing-slash removal 2013-09-13 12:37:35 -07:00
send-pack.c send-pack: fix parsing of --force-with-lease option 2013-08-02 16:07:45 -07:00
shortlog.c Merge branch 'jk/shortlog-tolerate-broken-commit' into maint 2013-10-28 10:17:31 -07:00
show-branch.c Replace deprecated OPT_BOOLEAN by OPT_BOOL 2013-08-05 11:32:19 -07:00
show-ref.c Replace deprecated OPT_BOOLEAN by OPT_BOOL 2013-08-05 11:32:19 -07:00
stripspace.c builtin/stripspace.c: fix broken indentation 2013-09-06 13:33:17 -07:00
symbolic-ref.c C: have space around && and || operators 2013-10-16 10:26:39 -07:00
tag.c Merge branch 'bk/refs-multi-update' 2013-09-20 12:36:12 -07:00
tar-tree.c tar-tree: remove dependency on sq_quote_print() 2013-07-30 08:10:35 -07:00
unpack-file.c
unpack-objects.c remove superfluous newlines in error messages 2012-04-30 15:45:51 -07:00
update-index.c Merge branch 'jl/submodule-mv' 2013-09-09 14:36:15 -07:00
update-ref.c Merge branch 'bk/refs-multi-update' 2013-09-20 12:36:12 -07:00
update-server-info.c i18n: update-server-info: mark parseopt strings for translation 2012-08-22 10:58:29 -07:00
upload-archive.c pkt-line: provide a LARGE_PACKET_MAX static buffer 2013-02-20 13:42:22 -08:00
var.c ident: rename IDENT_ERROR_ON_NO_NAME to IDENT_STRICT 2012-05-24 17:16:41 -07:00
verify-pack.c i18n: verify-pack: mark parseopt strings for translation 2012-08-22 10:58:29 -07:00
verify-tag.c gpg_interface: allow to request status return 2013-02-14 09:30:04 -08:00
write-tree.c i18n: write-tree: mark parseopt strings for translation 2012-08-22 10:58:29 -07:00