git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: [PATCH] config: use a static lock_file struct

From
Jeff King <peff@peff.net>
Date
Aug 30, 2017, 04:55 UTC
Message-ID
<20170830045555.27xczwo3ql7q4bg3@sigill.intra.peff.net>
In-Reply-To
<20170830043147.culn63luzdsbpuuw@sigill.intra.peff.net>
On Wed, Aug 30, 2017 at 12:31:47AM -0400, Jeff King wrote:
Show 27 quoted lines
> > It was surprisingly hard trying to get that code to do the right thing,
> > non-racily, in every error path. Since `remove_tempfiles()` can be
> > called any time (even from a signal handler), the linked list of
> > `tempfile` objects has to be kept valid at all times but can't use
> > mutexes. I didn't have the energy to keep going and make the lock
> > objects freeable.
> > 
> > I suppose the task could be made a bit easier by using `sigprocmask(2)`
> > or `pthread_sigmask(3)` to make its running environment a little bit
> > less hostile.
> 
> I think there are really two levels of carefulness here:
> 
>   1. Avoiding complicated things during a signal handler that may rely
>      on having a sane state from the rest of the program (e.g.,
>      half-formed entries, stdio locks, etc).
> 
>   2. Being truly race-free in the face of a signal arriving while we're
>      running arbitrary code that might have a tempfile struct in a funny
>      state.
> 
> I feel like right now we meet (1) and not (2). But I think if we keep to
> that lower bar of (1), it might not be that bad. We're assuming now that
> there's no race on the tempfile->active flag, for instance. We could
> probably make a similar assumption about putting items onto or taking
> them off of a linked list (it's not really atomic, but a single pointer
> assignment is probably "atomic enough" for our purposes).

Something like this, which AFAICT is about as safe as the existing code in its list manipulation. It retains the "active" flag as an additional check which I think isn't strictly necessary, but potentially catches some logic errors.

The strbuf_reset() calls become strbuf_release(), since we're promising the caller that they could now free the struct (or let it go out of scope) if they chose. Probably during a signal handler we should skip that (we know the struct is off the list and non-active at that point, but we could possibly hit libc's free() mutex).

diff --git a/tempfile.c b/tempfile.c
index 6843710670..a7d964ebf8 100644
--- a/tempfile.c
+++ b/tempfile.c
@@ -69,7 +69,6 @@ static void remove_tempfiles(int skip_fclose)
 				tempfile_list->fp = NULL;
 			delete_tempfile(tempfile_list);
 		}
-		tempfile_list = tempfile_list->next;
 	}
 }
 
@@ -85,39 +84,54 @@ static void remove_tempfiles_on_signal(int signo)
 	raise(signo);
 }
 
-/*
- * Initialize *tempfile if necessary and add it to tempfile_list.
- */
-static void prepare_tempfile_object(struct tempfile *tempfile)
+static void init_tempfile(struct tempfile *tempfile)
+{
+	tempfile->fd = -1;
+	tempfile->fp = NULL;
+	tempfile->active = 0;
+	tempfile->owner = 0;
+	tempfile->next = NULL;
+	strbuf_init(&tempfile->filename, 0);
+}
+
+static void activate_tempfile(struct tempfile *tempfile)
 {
-	if (!tempfile_list) {
-		/* One-time initialization */
+	static volatile int initialized;
+
+	if (!initialized) {
 		sigchain_push_common(remove_tempfiles_on_signal);
 		atexit(remove_tempfiles_on_exit);
+		initialized = 1;
 	}
 
 	if (tempfile->active)
-		die("BUG: prepare_tempfile_object called for active object");
-	if (!tempfile->on_list) {
-		/* Initialize *tempfile and add it to tempfile_list: */
-		tempfile->fd = -1;
-		tempfile->fp = NULL;
-		tempfile->active = 0;
-		tempfile->owner = 0;
-		strbuf_init(&tempfile->filename, 0);
-		tempfile->next = tempfile_list;
-		tempfile_list = tempfile;
-		tempfile->on_list = 1;
-	} else if (tempfile->filename.len) {
-		/* This shouldn't happen, but better safe than sorry. */
-		die("BUG: prepare_tempfile_object called for improperly-reset object");
+		die("BUG: activate_tempfile called for active object");
+
+	tempfile->next = tempfile_list;
+	tempfile_list = tempfile;
+	tempfile->active = 1;
+}
+
+static void deactivate_tempfile(struct tempfile *tempfile)
+{
+	struct tempfile *volatile *p;
+
+	if (!tempfile->active)
+		return;
+
+	tempfile->active = 0;
+	for (p = &tempfile_list; *p; p = &(*p)->next) {
+		if (*p == tempfile) {
+			*p = tempfile->next;
+			break;
+		}
 	}
 }
 
 /* Make sure errno contains a meaningful value on error */
 int create_tempfile(struct tempfile *tempfile, const char *path)
 {
-	prepare_tempfile_object(tempfile);
+	init_tempfile(tempfile);
 
 	strbuf_add_absolute_path(&tempfile->filename, path);
 	tempfile->fd = open(tempfile->filename.buf,
@@ -127,11 +141,11 @@ int create_tempfile(struct tempfile *tempfile, const char *path)
 		tempfile->fd = open(tempfile->filename.buf,
 				    O_RDWR | O_CREAT | O_EXCL, 0666);
 	if (tempfile->fd < 0) {
-		strbuf_reset(&tempfile->filename);
+		strbuf_release(&tempfile->filename);
 		return -1;
 	}
 	tempfile->owner = getpid();
-	tempfile->active = 1;
+	activate_tempfile(tempfile);
 	if (adjust_shared_perm(tempfile->filename.buf)) {
 		int save_errno = errno;
 		error("cannot fix permission bits on %s", tempfile->filename.buf);
@@ -144,25 +158,25 @@ int create_tempfile(struct tempfile *tempfile, const char *path)
 
 void register_tempfile(struct tempfile *tempfile, const char *path)
 {
-	prepare_tempfile_object(tempfile);
+	init_tempfile(tempfile);
 	strbuf_add_absolute_path(&tempfile->filename, path);
 	tempfile->owner = getpid();
-	tempfile->active = 1;
+	activate_tempfile(tempfile);
 }
 
 int mks_tempfile_sm(struct tempfile *tempfile,
 		    const char *template, int suffixlen, int mode)
 {
-	prepare_tempfile_object(tempfile);
+	init_tempfile(tempfile);
 
 	strbuf_add_absolute_path(&tempfile->filename, template);
 	tempfile->fd = git_mkstemps_mode(tempfile->filename.buf, suffixlen, mode);
 	if (tempfile->fd < 0) {
-		strbuf_reset(&tempfile->filename);
+		strbuf_release(&tempfile->filename);
 		return -1;
 	}
 	tempfile->owner = getpid();
-	tempfile->active = 1;
+	activate_tempfile(tempfile);
 	return tempfile->fd;
 }
 
@@ -171,7 +185,7 @@ int mks_tempfile_tsm(struct tempfile *tempfile,
 {
 	const char *tmpdir;
 
-	prepare_tempfile_object(tempfile);
+	init_tempfile(tempfile);
 
 	tmpdir = getenv("TMPDIR");
 	if (!tmpdir)
@@ -180,11 +194,11 @@ int mks_tempfile_tsm(struct tempfile *tempfile,
 	strbuf_addf(&tempfile->filename, "%s/%s", tmpdir, template);
 	tempfile->fd = git_mkstemps_mode(tempfile->filename.buf, suffixlen, mode);
 	if (tempfile->fd < 0) {
-		strbuf_reset(&tempfile->filename);
+		strbuf_release(&tempfile->filename);
 		return -1;
 	}
 	tempfile->owner = getpid();
-	tempfile->active = 1;
+	activate_tempfile(tempfile);
 	return tempfile->fd;
 }
 
@@ -293,8 +307,8 @@ int rename_tempfile(struct tempfile *tempfile, const char *path)
 		return -1;
 	}
 
-	tempfile->active = 0;
-	strbuf_reset(&tempfile->filename);
+	deactivate_tempfile(tempfile);
+	strbuf_release(&tempfile->filename);
 	return 0;
 }
 
@@ -305,7 +319,7 @@ void delete_tempfile(struct tempfile *tempfile)
 
 	if (!close_tempfile(tempfile)) {
 		unlink_or_warn(tempfile->filename.buf);
-		tempfile->active = 0;
-		strbuf_reset(&tempfile->filename);
+		deactivate_tempfile(tempfile);
+		strbuf_release(&tempfile->filename);
 	}
 }
diff --git a/tempfile.h b/tempfile.h
index 2f0038decd..df96f82e84 100644
--- a/tempfile.h
+++ b/tempfile.h
@@ -85,7 +85,6 @@ struct tempfile {
 	volatile int fd;
 	FILE *volatile fp;
 	volatile pid_t owner;
-	char on_list;
 	struct strbuf filename;
 };
 
Previous: Michael HaggertyNext: Jeff King
Message 22 of 43 in “pkt-line: re-'static'-ify buffer in packet_write_fmt_1()”
  1. pkt-line: re-'static'-ify buffer in packet_write_fmt_1()Martin Ågren, Aug 27, 2017
  2. Jeff KingAug 27, 2017
  3. Jeff KingAug 27, 2017
  4. Lars SchneiderAug 27, 2017
  5. Martin ÅgrenAug 27, 2017
  6. Jeff KingAug 27, 2017
  7. Lars SchneiderAug 27, 2017
  8. Jeff KingAug 27, 2017
  9. Martin ÅgrenAug 28, 2017
  10. Stefan BellerAug 28, 2017
  11. Jeff KingAug 28, 2017
  12. Junio C HamanoSep 5, 2017
  13. Lars SchneiderAug 29, 2017
  14. Jeff KingAug 29, 2017
  15. config: use a static lock_file structJeff King, Aug 29, 2017
  16. Brandon WilliamsAug 29, 2017
  17. Brandon WilliamsAug 29, 2017
  18. Jeff KingAug 29, 2017
  19. Michael HaggertyAug 30, 2017
  20. Jeff KingAug 30, 2017
  21. Michael HaggertyAug 30, 2017
  22. Jeff KingAug 30, 2017
  23. Jeff KingAug 30, 2017
  24. Michael HaggertyAug 30, 2017
  25. Jeff KingSep 2, 2017
  26. Jeff KingSep 2, 2017
  27. Michael HaggertyAug 30, 2017
  28. Jeff KingAug 30, 2017
  29. Brandon WilliamsAug 30, 2017
  30. Jeff KingAug 30, 2017
  31. Brandon WilliamsAug 30, 2017
  32. Jeff KingAug 31, 2017
  33. Junio C HamanoSep 6, 2017
  34. Jeff KingSep 6, 2017
  35. Martin ÅgrenAug 29, 2017
  36. Jeff KingAug 29, 2017
  37. Jeff KingAug 30, 2017
  38. Junio C HamanoSep 5, 2017
  39. 0/2 Do not call cmd_*() as a subroutineJunio C Hamano, Oct 10, 2017
  40. 1/2 describe: do not use cmd_*() as a subroutineJunio C Hamano, Oct 10, 2017
  41. SZEDER GáborOct 10, 2017
  42. Junio C HamanoOct 11, 2017
  43. 2/2 merge-ours: do not use cmd_*() as a subroutineJunio C Hamano, Oct 10, 2017

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.