Re: copy_from_user/syscall got me stock

[Gregory]

I think I have one of the latest versions of everything, but
copy_from_usr() / syscall() still behave randomly.  I think I missing
something obvious perhaps related to the locking since it works fine 9 out
of 10 times.

nargs[] was reduced to 12B   (when copy_from_user() fails it returns 4)
Kernel is 2.2.16
gcc version egcs-2.91.66 19990314/Linux (egcs-1.1.2 release)
glibc [devel] is 2.1.3-15

Is going to assembly is my only alternative to figure out why
copy_from_user() is failing?

Another strange thing is that my client program which is calling syscall()
generated core dumps few times but that is much more rare then Oops in a
log file.  I'll try to gdb it the next time this happens which has not
happened in a while.  Besides this code that I am trying to add, the
system is very stable.

Gregory

On Wed, 28 Jun 2000 [e-mail] wrote:
> Zitiere Gregory <[e-mail]>:
> > 
> > Jun 24 16:57:07 thinker kernel: Unable to handle
> > kernel NULL pointer dereference at
> > virtual address 00000276
> 
> You caused a stack overrun and overwrote your own 
> return address. 
> 
> * double check that nargs[] is really < 24
> * verify the generated asm code: copy_from_user is a 
> complex macro, perhaps gcc miscompiled that line.
> 
> Which gcc version do you use? Which kernel?
>