Skip to contentSkip to conversation

suspend.c: make sure stdin, stdout and stderr are open

PATCH In review Stefan Rompf 3 participants 6 messages Mar 24, 2006 – Apr 10, 2006

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 suspend.c: make sure stdin, stdout and stderr are open

Hi,

recently I patched my 2.6.16 kernel to use uswsusp. Nice work, but when I installed /usr/local/sbin/suspend into the Suse 9.3 powermanagement scripts, I've stumbled over a bad interaction: powersaved closes all filedescriptors before calling scripts. suspend then opens snapshot and swap device, allocating file descriptors 0 and 1, but printfs() afterwards. Luckily, I did not lose any data.

For robustness, suspend must make sure that these fds are open. Simple patch to achieve this is attached. It loses the descriptors, but I don't think that's an issue for the short running program.

Stefan

Index: suspend.c
===================================================================
RCS file: /cvsroot/suspend/suspend/suspend.c,v
retrieving revision 1.33

suspend.c +4 -0
diff -u -p -r1.33 suspend.c
@@ -831,6 +830,10 @@ int main(int argc, char *argv[])
 	dev_t resume_dev;
 	int orig_loglevel, ret = 0;
 
+	/* Easy way to make sure that stdin, stdout and stderr are open
+	   so that later printfs wont go to our suspend partition */
+	while((unsigned)open("/dev/null", O_RDWR)<2);
+
 	if (get_config("suspend", argc, argv, PARAM_NO, parameters, resume_dev_name))
 		return EINVAL;
 	if (compute_checksum != 'y' && compute_checksum != 'Y')
#1 lore raw

Hi,

Quoted 10 lines · Stefan Rompf

recently I patched my 2.6.16 kernel to use uswsusp. Nice work, but when I installed /usr/local/sbin/suspend into the Suse 9.3 powermanagement scripts, I've stumbled over a bad interaction: powersaved closes all filedescriptors before calling scripts. suspend then opens snapshot and swap device, allocating file descriptors 0 and 1, but printfs() afterwards. Luckily, I did not lose any data.

For robustness, suspend must make sure that these fds are open. Simple patch to achieve this is attached. It loses the descriptors, but I don't think that's an issue for the short running program.

Thanks a lot for the patch, I'll have a look at it shortly.

BTW, it follows from what you said that you compiled suspend on SuSE 9.3. Was it on i386?

Rafael

#2 lore raw
Quoted 7 lines · Rafael J. Wysocki
Quoted 6 lines · Stefan Rompf

recently I patched my 2.6.16 kernel to use uswsusp. Nice work, but when I installed /usr/local/sbin/suspend into the Suse 9.3 powermanagement scripts, I've stumbled over a bad interaction: powersaved closes all filedescriptors before calling scripts. suspend then opens snapshot and swap device, allocating file descriptors 0 and 1, but printfs() afterwards. Luckily, I did not lose any data.

Fortunately it drops the 0, 1 and 2 descriptors afterwards in prepare_console(). :-)

Still this is a bug and has to be fixed.

I've redone your patch a bit, could you please test it?

Rafael

---
 suspend.c |   15 ++++++++++++++-
 1 files changed, 14 insertions(+), 1 deletion(-)
suspend.c +14 -1
Index: suspend/suspend.c
===================================================================
@@ -856,7 +856,7 @@ int main(int argc, char *argv[])
 	struct stat stat_buf;
 	int resume_fd, snapshot_fd, vt_fd, orig_vc = -1, suspend_vc = -1;
 	dev_t resume_dev;
-	int orig_loglevel, ret = 0;
+	int orig_loglevel, ret;
 
 	if (get_config("suspend", argc, argv, PARAM_NO, parameters, resume_dev_name))
 		return EINVAL;
@@ -875,6 +875,18 @@ int main(int argc, char *argv[])
 	if (s2ram != 'y' && s2ram != 'Y')
 		s2ram = 0;
 
+	/* Make sure the 0, 1, 2 descriptors are open before opening the
+	 * snapshot and resume devices
+	 */
+	do {
+		ret = open("/dev/null", O_RDWR);
+		if (ret < 0) {
+			fprintf(stderr, "suspend: Could not open /dev/null\n");
+			return errno;
+		}
+	} while (ret <= 2);
+	close(ret);
+
 	setvbuf(stdout, NULL, _IONBF, 0);
 	setvbuf(stderr, NULL, _IONBF, 0);
 
@@ -898,6 +910,7 @@ int main(int argc, char *argv[])
 	}
 	resume_dev = stat_buf.st_rdev;
 
+	ret = 0;
 	if (stat(snapshot_dev_name, &stat_buf)) {
 		fprintf(stderr, "suspend: Could not stat the snapshot device file\n");
 		ret = ENODEV;
#3 lore raw

Am Freitag 24 März 2006 23:55 schrieb Rafael J. Wysocki:

Quoted 1 line

I've redone your patch a bit, could you please test it?

Works nicely, thanks, but why did you move the loop downwards, so that it is executed after reading config and calling into libcrypto?

As you asked for the distribution I use, it is SuSE 9.3 on i386, 2.6.16 with some patches (uswsusp, RFC2863-operstate, avoid "udev vc" bloat).

Stefan

#4 lore raw
Quoted 6 lines · Stefan Rompf

Am Freitag 24 März 2006 23:55 schrieb Rafael J. Wysocki:

Quoted 1 line

I've redone your patch a bit, could you please test it?

Works nicely, thanks, but why did you move the loop downwards, so that it is executed after reading config and calling into libcrypto?

For no particular reason. You're right it should go earlier, I'll change that.

Quoted 2 lines

As you asked for the distribution I use, it is SuSE 9.3 on i386, 2.6.16 with some patches (uswsusp, RFC2863-operstate, avoid "udev vc" bloat).

Thanks. I'm having problems with compiling vbetool on x86_64 SuSE 9.3, but it's good to know i386 is fine.

Greetings,
Rafael

#5 lore raw
In reply to Stefan Rompf
Quoted 6 lines · Stefan Rompf

Hi,

recently I patched my 2.6.16 kernel to use uswsusp. Nice work, but when I installed /usr/local/sbin/suspend into the Suse 9.3 powermanagement scripts, I've stumbled over a bad interaction: powersaved closes all filedescriptors before calling scripts. suspend then opens snapshot and swap device,

FWIW, i will also adress this in powersaved. More programs will run into the same trap probably ;-)

Signature · 3 lines
Stefan Seyfried                  \ "I didn't want to write for pay. I
QA / R&D Team Mobile Devices      \ wanted to be paid for what I write."
SUSE LINUX Products GmbH, Nürnberg \                    -- Leonard Cohen