linux-user/main.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
From: Helge Deller <deller@gmx.de>
Qemu user mode does not properly flushes error messages related to bad
arguments when exiting (at least when the output is piped to a file
instead of running on a terminal).
Ensure that we always flush by using exit() instead of _exit().
Reported by: Tobias Bergkvist
Resolves: https://gitlab.com/qemu-project/qemu/-/work_items/2544
Signed-off-by: Helge Deller <deller@gmx.de>
---
linux-user/main.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/linux-user/main.c b/linux-user/main.c
index c49d1e91d2..84e110dfe9 100644
--- a/linux-user/main.c
+++ b/linux-user/main.c
@@ -767,7 +767,7 @@ int main(int argc, char **argv, char **envp)
execfd = open(exec_path, O_RDONLY);
if (execfd < 0) {
printf("Error while loading %s: %s\n", exec_path, strerror(errno));
- _exit(EXIT_FAILURE);
+ exit(EXIT_FAILURE);
}
}
--
2.53.0
On Sat, 25 Apr 2026 at 15:49, Helge Deller <deller@kernel.org> wrote:
>
> From: Helge Deller <deller@gmx.de>
>
> Qemu user mode does not properly flushes error messages related to bad
> arguments when exiting (at least when the output is piped to a file
> instead of running on a terminal).
> Ensure that we always flush by using exit() instead of _exit().
>
> Reported by: Tobias Bergkvist
> Resolves: https://gitlab.com/qemu-project/qemu/-/work_items/2544
> Signed-off-by: Helge Deller <deller@gmx.de>
> ---
> linux-user/main.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/linux-user/main.c b/linux-user/main.c
> index c49d1e91d2..84e110dfe9 100644
> --- a/linux-user/main.c
> +++ b/linux-user/main.c
> @@ -767,7 +767,7 @@ int main(int argc, char **argv, char **envp)
> execfd = open(exec_path, O_RDONLY);
> if (execfd < 0) {
> printf("Error while loading %s: %s\n", exec_path, strerror(errno));
> - _exit(EXIT_FAILURE);
> + exit(EXIT_FAILURE);
> }
> }
How about the other _exit() further down in the function
for the case where loader_exec() fails ? That seems like
it would have the same problem.
-- PMM
On 4/26/26 15:01, Peter Maydell wrote:
> On Sat, 25 Apr 2026 at 15:49, Helge Deller <deller@kernel.org> wrote:
>>
>> From: Helge Deller <deller@gmx.de>
>>
>> Qemu user mode does not properly flushes error messages related to bad
>> arguments when exiting (at least when the output is piped to a file
>> instead of running on a terminal).
>> Ensure that we always flush by using exit() instead of _exit().
>>
>> Reported by: Tobias Bergkvist
>> Resolves: https://gitlab.com/qemu-project/qemu/-/work_items/2544
>> Signed-off-by: Helge Deller <deller@gmx.de>
>> ---
>> linux-user/main.c | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/linux-user/main.c b/linux-user/main.c
>> index c49d1e91d2..84e110dfe9 100644
>> --- a/linux-user/main.c
>> +++ b/linux-user/main.c
>> @@ -767,7 +767,7 @@ int main(int argc, char **argv, char **envp)
>> execfd = open(exec_path, O_RDONLY);
>> if (execfd < 0) {
>> printf("Error while loading %s: %s\n", exec_path, strerror(errno));
>> - _exit(EXIT_FAILURE);
>> + exit(EXIT_FAILURE);
>> }
>> }
>
> How about the other _exit() further down in the function
> for the case where loader_exec() fails ? That seems like
> it would have the same problem.
I'll send a seperate patch for this one. Will then include in
next pull request series.
Thanks!
Helge
On Sat, Apr 25, 2026 at 8:50 AM Helge Deller <deller@kernel.org> wrote: > From: Helge Deller <deller@gmx.de> > > Qemu user mode does not properly flushes error messages related to bad > arguments when exiting (at least when the output is piped to a file > instead of running on a terminal). > Ensure that we always flush by using exit() instead of _exit(). > > Reported by: Tobias Bergkvist > Resolves: https://gitlab.com/qemu-project/qemu/-/work_items/2544 > Signed-off-by: Helge Deller <deller@gmx.de> > --- > linux-user/main.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > Reviewed-by: Warner Losh <imp@bsdimp.com> Note: I've added the linux-user maintainers to the CC list for visibility. bsd-user has the same problem. Warner
On 4/25/26 17:02, Warner Losh wrote: > > > On Sat, Apr 25, 2026 at 8:50 AM Helge Deller <deller@kernel.org <mailto:deller@kernel.org>> wrote: > > From: Helge Deller <deller@gmx.de <mailto:deller@gmx.de>> > > Qemu user mode does not properly flushes error messages related to bad > arguments when exiting (at least when the output is piped to a file > instead of running on a terminal). > Ensure that we always flush by using exit() instead of _exit(). > > Reported by: Tobias Bergkvist > Resolves: https://gitlab.com/qemu-project/qemu/-/work_items/2544 <https://gitlab.com/qemu-project/qemu/-/work_items/2544> > Signed-off-by: Helge Deller <deller@gmx.de <mailto:deller@gmx.de>> > --- > linux-user/main.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > > Reviewed-by: Warner Losh <imp@bsdimp.com <mailto:imp@bsdimp.com>> Thanks! > Note: I've added the linux-user maintainers to the CC list for visibility. Ok. > bsd-user has the same problem. I hesitate to change things I don't not test at all :-( Even this patch here is not widely tested, but I think we need to add it. If things break we see reports anyway. Helge
[[ Oops, my last message cc'd Pierrick's old address, maybe my tree is old
or it needs to update in MAINTAINERS ]]
On Sat, Apr 25, 2026 at 9:17 AM Helge Deller <deller@gmx.de> wrote:
> On 4/25/26 17:02, Warner Losh wrote:
> >
> >
> > On Sat, Apr 25, 2026 at 8:50 AM Helge Deller <deller@kernel.org <mailto:
> deller@kernel.org>> wrote:
> >
> > From: Helge Deller <deller@gmx.de <mailto:deller@gmx.de>>
> >
> > Qemu user mode does not properly flushes error messages related to
> bad
> > arguments when exiting (at least when the output is piped to a file
> > instead of running on a terminal).
> > Ensure that we always flush by using exit() instead of _exit().
> >
> > Reported by: Tobias Bergkvist
> > Resolves: https://gitlab.com/qemu-project/qemu/-/work_items/2544 <
> https://gitlab.com/qemu-project/qemu/-/work_items/2544>
> > Signed-off-by: Helge Deller <deller@gmx.de <mailto:deller@gmx.de>>
> > ---
> > linux-user/main.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> >
> > Reviewed-by: Warner Losh <imp@bsdimp.com <mailto:imp@bsdimp.com>>
>
> Thanks!
>
> > Note: I've added the linux-user maintainers to the CC list for
> visibility.
>
> Ok.
>
> > bsd-user has the same problem.
> I hesitate to change things I don't not test at all :-(
>
I can, since I do a lot with bsd-user.
> Even this patch here is not widely tested, but I think we need to add it.
> If things break we see reports anyway.
>
I don't think this will break anything. The printf can't possibly work,
either
we must change it to a write or do the exit. exit is done everywhere else.
This has been wrong since:
commit 03cfd8faa7ffb7201e2949b99c2f35b1fef7078b
Author: Laurent Vivier <laurent@vivier.eu>
Date: Fri Aug 30 01:46:44 2013 +0200
linux-user: add support of binfmt_misc 'O' flag
...
+ if (execfd < 0) {
+ printf("Error while loading %s: %s\n", filename,
strerror(-execfd));
+ _exit(1);
+ }
...
but the full commit log doesn't say why. Laaurent was one of the people I
cc'd before
I chased this back to this commit. :).
I'll go ahead and make the bsd-user/main.c change (the error message is a
little different, and there's
a pull request against the bsd-user upstream I need to land in the area
around it, IIRC).
Warner
On Sat, 25 Apr 2026 at 16:51, Warner Losh <imp@bsdimp.com> wrote:
> I don't think this will break anything. The printf can't possibly work, either
> we must change it to a write or do the exit. exit is done everywhere else.
> This has been wrong since:
>
> commit 03cfd8faa7ffb7201e2949b99c2f35b1fef7078b
> Author: Laurent Vivier <laurent@vivier.eu>
> Date: Fri Aug 30 01:46:44 2013 +0200
>
> linux-user: add support of binfmt_misc 'O' flag
> ...
> + if (execfd < 0) {
> + printf("Error while loading %s: %s\n", filename, strerror(-execfd));
> + _exit(1);
> + }
> ...
> but the full commit log doesn't say why.
My guess is that the "print an error and exit" if() block was
simply copied from the bit further down that does:
if (ret != 0) {
printf("Error while loading %s: %s\n", exec_path, strerror(-ret));
_exit(EXIT_FAILURE);
}
for the loader_exec failure case. That one dates back to
commit 74cd30b811cc2f in 2003, which changes some exit() calls
to _exit(). I have no idea why, and the commit message is cryptic
and useless. But it seems unlikely that the reason is still
relevant today.
-- PMM
On 26/4/26 15:06, Peter Maydell wrote:
> On Sat, 25 Apr 2026 at 16:51, Warner Losh <imp@bsdimp.com> wrote:
>> I don't think this will break anything. The printf can't possibly work, either
>> we must change it to a write or do the exit. exit is done everywhere else.
>> This has been wrong since:
>>
>> commit 03cfd8faa7ffb7201e2949b99c2f35b1fef7078b
>> Author: Laurent Vivier <laurent@vivier.eu>
>> Date: Fri Aug 30 01:46:44 2013 +0200
>>
>> linux-user: add support of binfmt_misc 'O' flag
>> ...
>> + if (execfd < 0) {
>> + printf("Error while loading %s: %s\n", filename, strerror(-execfd));
>> + _exit(1);
>> + }
>> ...
>> but the full commit log doesn't say why.
>
> My guess is that the "print an error and exit" if() block was
> simply copied from the bit further down that does:
>
> if (ret != 0) {
> printf("Error while loading %s: %s\n", exec_path, strerror(-ret));
> _exit(EXIT_FAILURE);
> }
>
> for the loader_exec failure case. That one dates back to
> commit 74cd30b811cc2f in 2003, which changes some exit() calls
> to _exit(). I have no idea why, and the commit message is cryptic
> and useless. But it seems unlikely that the reason is still
> relevant today.
Could it be do discard any atexit() callback?
© 2016 - 2026 Red Hat, Inc.