| Message ID | 18ea7470-f507-6424-b21a-00ac0f5673ba@genexis.eu |
|---|---|
| State | New |
| Headers | show |
| Series | [procd] Fix retriggering of init.d-scripts. | expand |
The sender domain has a DMARC Reject/Quarantine policy which disallows sending mailing list messages using the original "From" header. To mitigate this problem, the original message has been wrapped automatically by the mailing list software. Em 28/08/2023 11:12, Markus Gothe escreveu: > Fix retriggering of init.d-scripts which calls > commands dependent on functional STDIN/STDOUT/STDERR. Yeah, it is no fun when some library decides to become allergic to one of FD0-2 being closed, and a O.S. update suddenly exposes this issue. Been there, suffered through it, learned from it. So, this is right up there in the "Just don't do it" list, right after calling execve() with NULL for argv[0]. Avoid closed FDs 0-2 at program start like the plague: fix it first thing. Never execve() anything with closed FDs 0-2 in the general case. It is also true for shell scripts: don't close FDs 0-2, redirect them to/from /dev/null instead. Don't repurpose them, either. Data corruption (due to unexpected writes from just about anything) is a possible outcome of running stuff with FDs 0-2 closed or repurposed, BTW. Consider yourself lucky when things just crash or exit with an error, instead. IMHO, if they're not doing it already, ubus and all its helpers, such as rpcd, should be defensive about FDs 0-2 and ensure they are always properly open (to /dev/null if need be), preferably at program start, and certainly when they are about to execve() anything. PS: I have attached an example of how one could do generic self-healing of the standard low FDs in C in a POSIX environment where fcntl() and dup2() are available. Note that FD 2 is expected to be R/W, according to the C specification for stdio.h streams.
diff --git a/service/trigger.c b/service/trigger.c index 4af1474..324d78e 100644 --- a/service/trigger.c +++ b/service/trigger.c @@ -118,9 +118,16 @@ static void trigger_command_run(struct runqueue *q, struct runqueue_task *t) } if (debug < 3) { - close(STDIN_FILENO); - close(STDOUT_FILENO); - close(STDERR_FILENO); + int devnull = open("/dev/null", O_RDWR); + + if (devnull >= 0) { + dup2(devnull, STDIN_FILENO); + dup2(devnull, STDOUT_FILENO); + dup2(devnull, STDERR_FILENO); + + if (devnull > STDERR_FILENO) + close(devnull); + } }
Fix retriggering of init.d-scripts which calls commands dependent on functional STDIN/STDOUT/STDERR. If we just close these file descriptors those commands will not work as expected leading to unwanted consequences. If we instead redirect the file descriptors to /dev/null, we will end up the same end-result and these commands will work as expected. Signed-off-by: Markus Gothe <markus.gothe@genexis.eu> --- service/trigger.c | 13 ++++++++++--- 1 file changed, 10 insertions(+), 3 deletions(-) blobmsg_for_each_attr(cur, cmd->data, rem)