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

Re: ssh unaligned trap error...

On 20 Apr 1999, Steve Clark wrote:

> Can anyone identify with this? What is the error telling me?  Should I be
> concerned?  If so, any hints on what I might do to correct it.

Well, it's not an "error" per se, but rather a warning that some part of
the programme or lib that it uses is manipulating "unaligned" data or
structures, meaning that they are not aligned on the boundaries that the
Alpha processor expects. These warnings can be disabled at the kernel
level (modify a small chunk of the kernel...patch is on the axp-list
archives somewhere).

Now, as to whether or not to be concerned about it...that depends on how
good you are at coding and/or gdb.  If you're programming skills are at
the "novice" level, it may not be something you want to tackle.  After
all, the software DOES work and the performance tradeoff is frequently not
THAT bad (the kernel corrects for such unaligned accesses, fyi).  If
you're good, however, dig in there with gdb and find out just what is at
the address that it's giving you in the warning message.  That should tell
you that, somewhere there's either a type mismatch or some type of poor
typecasting done on that function/structure/variable.

> Apr 20 10:26:44 www kernel: VFS: Mounted root (ext2 filesystem) readonly.
> Apr 20 10:26:44 www kernel: Adding Swap: 523640k swap-space (priority -1)
> Apr 20 10:26:44 www kernel: sysctl: ip forwarding enabled
> Apr 20 10:26:46 www kernel: sshd2(219): unaligned trap at 0000000120025458:
> 000000011ffff60c 2d 2
> Apr 20 10:26:46 www kernel: sshd2(219): unaligned trap at 0000000120025458:
> 000000011ffff614 2d 2
> Apr 20 10:26:46 www sshd2[219]: Listener created on port 22.
> Apr 20 10:26:46 www sshd2[220]: Daemon is running.

IMO, the daemon runs and works, so unless you use the console alot and/or
really care if it gets debugged totally, don't mess with it.  If you
manage to get everything debugged, however, please notify the upstream
authors with a patch so that they can fix it there for all of us to enjoy.
Most of the people who write these packages are usually writing on
i386-class Linux machines and don't think about 64-bit archs.  They
usually appreciate the extra work since they rarely have access to such


Reply to: