public inbox for glibc-cvs@sourceware.org
help / color / mirror / Atom feed
* [glibc] hurd: Prevent the final file_exec_paths call from signals
@ 2023-11-20 22:28 Samuel Thibault
0 siblings, 0 replies; only message in thread
From: Samuel Thibault @ 2023-11-20 22:28 UTC (permalink / raw)
To: glibc-cvs
https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=49b308a26e2a9e02ef396f67f59c462ad4171ea4
commit 49b308a26e2a9e02ef396f67f59c462ad4171ea4
Author: Samuel Thibault <samuel.thibault@ens-lyon.org>
Date: Mon Nov 20 19:19:50 2023 +0100
hurd: Prevent the final file_exec_paths call from signals
Otherwise if the exec server started thrashing the old task,
we won't be able to restart the exec.
This notably fixes building ghc.
Diff:
---
hurd/hurdexec.c | 12 ++++++++++++
sysdeps/mach/hurd/spawni.c | 16 +++++++++++++++-
2 files changed, 27 insertions(+), 1 deletion(-)
diff --git a/hurd/hurdexec.c b/hurd/hurdexec.c
index e358d846c8..317d7ea0ad 100644
--- a/hurd/hurdexec.c
+++ b/hurd/hurdexec.c
@@ -362,6 +362,7 @@ retry:
if (!err)
{
int flags;
+ sigset_t old, new;
if (pdp)
{
@@ -420,6 +421,15 @@ retry:
if (__sigismember (&_hurdsig_traced, SIGKILL))
flags |= EXEC_SIGTRAP;
#endif
+
+ /* Avoid getting interrupted while exec(), notably not after the exec
+ server has committed to the exec and started thrashing us.
+
+ TODO Rather add proper interrupt support to the exec server, that
+ avoids interrupts in that period. */
+ __sigfillset (&new);
+ __sigprocmask (SIG_SETMASK, &new, &old);
+
err = __file_exec_paths (file, task, flags,
path ? path : "",
abspath ? abspath : "",
@@ -440,6 +450,8 @@ retry:
ints, INIT_INT_MAX,
please_dealloc, pdp - please_dealloc,
portnames, nportnames);
+
+ __sigprocmask (SIG_SETMASK, &old, NULL);
}
/* Release references to the standard ports. */
diff --git a/sysdeps/mach/hurd/spawni.c b/sysdeps/mach/hurd/spawni.c
index 9516001817..58a89b45f3 100644
--- a/sysdeps/mach/hurd/spawni.c
+++ b/sysdeps/mach/hurd/spawni.c
@@ -812,6 +812,18 @@ retry:
inline error_t exec (file_t file)
{
+ sigset_t old, new;
+
+ /* Avoid getting interrupted while exec(), notably not after the exec
+ server has committed to the exec and started thrashing the task.
+
+ Various issues otherwise show up when building e.g. ghc.
+
+ TODO Rather add proper interrupt support to the exec server, that
+ avoids interrupts in that period. */
+ __sigfillset(&new);
+ __sigprocmask (SIG_SETMASK, &new, &old);
+
error_t err = __file_exec_paths
(file, task,
__sigismember (&_hurdsig_traced, SIGKILL) ? EXEC_SIGTRAP : 0,
@@ -824,7 +836,7 @@ retry:
/* Fallback for backwards compatibility. This can just be removed
when __file_exec goes away. */
if (err == MIG_BAD_ID)
- return __file_exec (file, task,
+ err = __file_exec (file, task,
(__sigismember (&_hurdsig_traced, SIGKILL)
? EXEC_SIGTRAP : 0),
args, argslen, env, envlen,
@@ -833,6 +845,8 @@ retry:
ints, INIT_INT_MAX,
NULL, 0, NULL, 0);
+ __sigprocmask (SIG_SETMASK, &old, NULL);
+
return err;
}
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2023-11-20 22:28 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-11-20 22:28 [glibc] hurd: Prevent the final file_exec_paths call from signals Samuel Thibault
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).