Re: [PATCH] manual: Add note to process.texi about MT-process fork (Bug 30159)

[Carlos O'Donell <carlos@redhat.com>]

On 2/23/23 16:32, Florian Weimer wrote:
> * Carlos O'Donell via Libc-alpha:
> 
>> The forked child of an MT-process may only call AS-safe functions.
>> ---
>>  manual/process.texi | 6 ++++++
>>  1 file changed, 6 insertions(+)
>>
>> diff --git a/manual/process.texi b/manual/process.texi
>> index 9307379194..dfdf0923b4 100644
>> --- a/manual/process.texi
>> +++ b/manual/process.texi
>> @@ -316,6 +316,12 @@ The child doesn't inherit alarms set by the parent process.
>>  The set of pending signals (@pxref{Delivery of Signal}) for the child
>>  process is cleared.  (The child process inherits its mask of blocked
>>  signals and signal actions from the parent process.)
>> +
>> +@item
>> +In the case of a multi-threaded parent, the child process may only
>> +safely call asynchronous-signal-safe functions until the process is
>> +replaced with a call to an @code{exec} family function.
> 
> We support calling malloc after fork as an extension.  For example,
> calling opendir after fork to find file descriptors in /proc/self/fd is
> fairly common.

We support calling a lot of things in reality, but I didn't want to over commit
with this change to the manual. I like your narrowly worded addition.

> Perhaps add:
> 
> As an extension relative to POSIX, @theglibc{} supports calling
> @code{malloc} and other allocation functions after @code{fork}.  This
> only applies to the standard @code{malloc}.  Whether this extension is
> supported by replacement @code{malloc} implementations depends on the
> replacement.  @xref{Replacing malloc}.

v2 posted:
https://patchwork.sourceware.org/project/glibc/patch/20230223223728.423748-1-carlos@redhat.com/

Checked pdf/html/texi builds and the xref and macros look good.

-- 
Cheers,
Carlos.