Re: [PATCH mdadm 1/2] mdadm: Fix hang race condition in wait_for_zero_forks()
From: Paul Menzel <hidden>
Date: 2024-06-04 21:35:30
Dear Logan, Thank you for the patch. Am 04.06.24 um 18:38 schrieb Logan Gunthorpe:
Running a create operation with --write-zeros can randomly hang forever waiting for child processes. This happens roughly on in ten runs with when running with small (20MB) loop devices. The bug is caused by the fact that signals can be coallesced into one if they are not read by signalfd quick enough. So if two children finish at exactly the same time, only one SIGCHLD will be received by the parent. To fix this, wait on all processes with WNOHANG every time a SIGCHLD is recieved and exit when all processes have been waited on.
received
quoted hunk ↗ jump to hunk
Reported-by: Xiao Ni <redacted> Signed-off-by: Logan Gunthorpe <logang@deltatee.com> --- Create.c | 28 +++++++++++++++------------- 1 file changed, 15 insertions(+), 13 deletions(-)diff --git a/Create.c b/Create.c index d033eb68f30c..4f992a22b7c9 100644 --- a/Create.c +++ b/Create.c@@ -178,6 +178,7 @@ static int wait_for_zero_forks(int *zero_pids, int count) bool interrupted = false; sigset_t sigset; ssize_t s; + pid_t pid; for (i = 0; i < count; i++) if (zero_pids[i])@@ -196,7 +197,7 @@ static int wait_for_zero_forks(int *zero_pids, int count) return 1; } - while (1) { + while (wait_count) { s = read(sfd, &fdsi, sizeof(fdsi)); if (s != sizeof(fdsi)) { pr_err("Invalid signalfd read: %s\n", strerror(errno));@@ -209,23 +210,24 @@ static int wait_for_zero_forks(int *zero_pids, int count) pr_info("Interrupting zeroing processes, please wait...\n"); interrupted = true; } else if (fdsi.ssi_signo == SIGCHLD) { - if (!--wait_count) - break; + for (i = 0; i < count; i++) { + if (!zero_pids[i]) + continue; + + pid = waitpid(zero_pids[i], &wstatus, WNOHANG); + if (pid <= 0) + continue; + + zero_pids[i] = 0; + if (!WIFEXITED(wstatus) || WEXITSTATUS(wstatus)) + ret = 1; + wait_count--; + } } } close(sfd); - for (i = 0; i < count; i++) { - if (!zero_pids[i]) - continue; - - waitpid(zero_pids[i], &wstatus, 0); - zero_pids[i] = 0; - if (!WIFEXITED(wstatus) || WEXITSTATUS(wstatus)) - ret = 1; - } - if (interrupted) { pr_err("zeroing interrupted!\n"); return 1;
Reviewed-by: Paul Menzel <redacted> Kind regards, Paul