Current-Users archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: arc4random abort -- possibly caused by recent pthread_atfork() change?




> On Mar 1, 2025, at 2:52 AM, Greg A. Woods <woods%planix.ca@localhost> wrote:
> 
> Ah ha!
> 
> So I just encountered this with a new build from today's sources too.
> 
> I was running sysinst from the install image and it barfed a pax.core
> while trying to upgrade the target system, immediately on trying to
> extract the first set (i.e. the kernel set).
> 
> The core reveals:
> 
> Reading symbols from pax...
> Reading symbols from /home/woods/pax.debug...
> [New process 1664]
> Core was generated by `pax'.
> Program terminated with signal SIGABRT, Aborted.
> #0  0x000000000047903a in _lwp_kill ()
> (gdb) bt
> #0  0x000000000047903a in _lwp_kill ()
> #1  0x00000000004df714 in abort ()
> #2  0x00000000004df67a in arc4random_prng_get[cold] ()
> #3  0x000000000046e57b in arc4random_uniform ()
> #4  0x000000000046d5c7 in __gettemp ()
> #5  0x000000000046d4b6 in mkstemp ()
> #6  0x0000000000406375 in file_creat (arcn=0x51fea0 <archd>, write_to_hardlink=<optimized out>)
>    at /Volumes/work/woods/g-NetBSD-src/bin/pax/file_subs.c:165
> #7  0x000000000040420b in extract () at /Volumes/work/woods/g-NetBSD-src/bin/pax/ar_subs.c:487
> #8  0x00000000004eac25 in main (argc=<optimized out>, argv=<optimized out>) at /Volumes/work/woods/g-NetBSD-src/bin/pax/pax.c:309
> (gdb)
> 
> 
> The static-linked pax binary on the target system still works fine.
> That was built from Feb 25 sources, i.e. just a few days ago.
> 
> I re-linked pax again creating a link map to make sure I knew for sure
> where those arc4random* calls were loaded from and noticed a call from
> arc4random.o to pthread_atfork() and I remembered a recent change I saw
> in what I pulled to update from Feb 25 to today (this is from git, of
> course):
> 
> 	commit 7b183bad5a4e5b2b85152895ad68754a9c7d4090
> 	Author: christos <christos%NetBSD.org@localhost>
> 	Date:   Fri Feb 28 16:00:26 2025 +0000
> 
> 	    PR/59112: Martin Husemann: switch to using mmap instead of malloc
> 
> Perhaps this could be causing the problem?  Christos?

Yes, looks like it. Let me try to reproduce it.

christos

Attachment: signature.asc
Description: Message signed with OpenPGP



Home | Main Index | Thread Index | Old Index