Re: INFO: rcu detected stall in n_tty_receive_char_special

[Tetsuo Handa]

On 2018/04/22 0:34, Tetsuo Handa wrote:
> Dmitry Vyukov wrote:
>> This looks somewhat similar to "INFO: rcu detected stall in __process_echoes":
>> https://syzkaller.appspot.com/bug?id=17f23b094cd80df750e5b0f8982c521ee6bcbf40
>> But I am not sure because stall stacks are somewhat different.
> 
> Yes, this will be the same one which can be reproduced using same reproducer
> provided by syzbot. You can try below one like "while :; do echo -n .; ./a.out; done".
> 
> ----------------------------------------
> #include <fcntl.h>
> #include <stdio.h>
> #include <sys/ioctl.h>
> #include <sys/stat.h>
> #include <unistd.h>
> 
> int main(int argc, char *argv[])
> {
> 	int fd[2];
> 	unsigned int zero = 0;
> 	char buf[128];
> 	int ptyno = 0;
> 	char c = 0;
> 	char data[64] = { };
> 	while (1) {
> 		fd[0] = open("/dev/ptmx", O_RDWR);
> 		ioctl(fd[0], TIOCSPTLCK, &zero);
> 		write(fd[0], data, sizeof(data));
> 		if (ioctl(fd[0], TIOCGPTN, &ptyno))
> 			break;
> 		sprintf(buf, "/dev/pts/%d", ptyno);
> 		fd[1] = open(buf, O_RDONLY);
> 		if (fd[1] == EOF)
> 			break;
> 		ioctl(fd[1], TIOCSTI, &c);
> 	}
> 	return 1;
> }
> ----------------------------------------

Bisection reached to commit e052c6d15c61cc4c ("tty: Use unbound workqueue
for all input workers") added in 4.4-rc1.

Reverting that commit on top of 4.4 solved the lockup. But unfortunately,
reverting that commit on top of 4.16 only helped surviving the test longer
than vanilla 4.16; still stalls at __process_echoes(). There might be similar
changes which need to be reverted as well.



# bad: [8005c49d9aea74d382f474ce11afbbc7d7130bec] Linux 4.4-rc1
# good: [6a13feb9c82803e2b815eca72fa7a9f5561d7861] Linux 4.3
# good: [64291f7db5bd8150a74ad2036f1037e6a0428df2] Linux 4.2
# good: [b953c0d234bc72e8489d3bf51a276c5c4ec85345] Linux 4.1
# good: [39a8804455fb23f09157341d3ba7db6d7ae6ee76] Linux 4.0
# good: [bfa76d49576599a4b9f9b7a71f23d73d6dcff735] Linux 3.19
# good: [b2776bf7149bddd1f4161f14f79520f17fc1d71d] Linux 3.18
# good: [bfe01a5ba2490f299e1d2d5508cbbbadd897bbe9] Linux 3.17
# good: [19583ca584d6f574384e17fe7613dfaeadcdc4a6] Linux 3.16
# good: [1860e379875dfe7271c649058aeddffe5afd9d0d] Linux 3.15
# good: [455c6fdbd219161bd09b1165f11699d6d73de11c] Linux 3.14
# good: [d8ec26d7f8287f5788a494f56e8814210f0e64be] Linux 3.13
# good: [5e01dc7b26d9f24f39abace5da98ccbd6a5ceb52] Linux 3.12
git bisect start 'HEAD' 'v4.3' 'v4.2' 'v4.1' 'v4.0' 'v3.19' 'v3.18' 'v3.17' 'v3.16' 'v3.15' 'v3.14' 'v3.13' 'v3.12' '--' 'drivers/tty/'
# good: [cc32382d9fd22dc8eebba4a245f50417267bda8e] serial: imx: make setup_ufcr more useful
git bisect good cc32382d9fd22dc8eebba4a245f50417267bda8e
# good: [32ede4a51754cb62b0d43d91cb7c4e3c57069a9c] serial: tegra: Add helper function for handling RX buffer
git bisect good 32ede4a51754cb62b0d43d91cb7c4e3c57069a9c
# good: [4b41b9539a1e9531f942ededfcdcff372317d2e7] tty: Prevent tty teardown during tty_write_message()
git bisect good 4b41b9539a1e9531f942ededfcdcff372317d2e7
# bad: [fd0d351de7bbd718bc2b34d5846854831aa2b88c] Merge tag 'tty-4.4-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty
git bisect bad fd0d351de7bbd718bc2b34d5846854831aa2b88c
# good: [14c79092909a52b6fd6394b6ad5e7756c4f9565e] Merge branch 'parisc-4.3-2' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux
git bisect good 14c79092909a52b6fd6394b6ad5e7756c4f9565e
# good: [41ecf1404b34d9975eb97f5005d9e4274eaeb76a] Merge tag 'for-linus-4.4-rc0-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip
git bisect good 41ecf1404b34d9975eb97f5005d9e4274eaeb76a
# bad: [e052c6d15c61cc4caff2f06cbca72b183da9f15e] tty: Use unbound workqueue for all input workers
git bisect bad e052c6d15c61cc4caff2f06cbca72b183da9f15e
# good: [e176058f0de53c2346734e5254835e0045364001] tty: Abstract tty buffer work
git bisect good e176058f0de53c2346734e5254835e0045364001
# first bad commit: [e052c6d15c61cc4caff2f06cbca72b183da9f15e] tty: Use unbound workqueue for all input workers