XXX Not sure if testing tp->t_template is the right way to discern
this -- I just reached for it because the downstream crash is a panic
on tp->t_template == NULL in tcp_output.
XXX In principle this could try connecting to the address, except
it's not passed down from the logic in uipc_socket.c to tcp_sendoob.
Reported-by: syzbot+a01f4cfec72790855ce2@syzkaller.appspotmail.com
- To enable stand still mode (robot is not moving but waiting in place),
specify a scorefile on the command line called <directory>/stand_still
- To enable pattern roll mode (robot is changing which move to look first)
specify a scorefile on the command line called <directory>/pattern_roll
strlcpy does not; strncpy does. However, the member must be
NUL-terminated, because kdump uses it, e.g., with printf("%s"), so
expicitly set the last byte to NUL.
Possible fix for a host of sanitizer complaints in syzkaller.
Return false if timespecadd or timespecsub with the same arguments
would overflow (possibly in an intermediate calculation), true if OK.
Typical usage:
sys_wotsit(...)
{
...
if (!timespecsubok(x, y))
return EINVAL;
timespecub(x, y, xydelta);
...
}
Optimizing for non-strict-alignment architectures -- without falling
afoul of alignment sanitizers or overeager compilers -- is left as an
exercise for the reader.
PR kern/56894
Discovered by code inspection. Remarkably, a combination of errors
made this fail to be a stack buffer overrun. Verified by booting
with ARMv8.0-AES disabled and with the self-test artificially made to
fail.
by GCC for HPPA to hold the return address where the signal trampoline
will resume. XXX Same treatment is needed for HPPA64, but not done as
part of this commit.
Thanks to thorpej for help with this. (ages ago)
For NetBSD/m68k, program header is erroneously readable from core dump,
although a page containing it is missing. This spoils relocation for
the main executable, and debugging with core dumps becomes impossible,
as described in the PR.
In order to avoid this failure, add consistency check for program header;
for NetBSD, 1st entry of program header refers program header itself. If
this is not the case, we should be reading random garbage from core dump.
Yes, this is a kind of the hackest hack. But, unable to debug with core
dump is critical. Therefore, I commit this workaround temporarily until
the real cause is found and fixed.
Note that this is not a recent regression for GDB nor kernel, as
mentioned in the PR. These from netbsd-8 and -9 fail in the same way
without this hack.
no longer build...
> Part of PR port-powerpc/56899
>
> Set default OS ABI to NetBSD when target is *-*-netbsd*.
>
> This fixes default behaviors of GDB when no executable is loaded
> (== OS ABI is not determined yet).