n the attached file are two new tests and some patche sto remove
according to the response to my question I have used both
possibilities. It was just the easiest way to create the tests.
A new test to check for EFAULT. I copied a file and made the neccesary
In this case it was easier to modify readlink03.c for the additional
removes warnings in kernel.io.stress_cd and kernel.syscalls.rmdir
/ __/____/ / Caldera (Deutschland) GmbH
/ /_/ __ / /__ Naegelsbachstr. 49c, 91052 Erlangen
/_____//_/ /____/ http://www.caldera.de
==== /_____/ ====== Dipl. Inform. Ihno Krumreich, email: ihno@...
Caldera Open Linux Tel: ++49 9131 71923-27, fax: ++49 9131 71923-99
Ihno Krumreich wrote:
>A new test to check for EFAULT. I copied a file and made the neccesary
Doesn't this cause a segfault? In the past, the stat functions had
libc-side accomodations which kept the user buffer passed to the stat
call from being written to by the kernel. Instead, the kernel wrote to
a buffer established by libc. It would just fault when libc tried to
write to the user buffer. Maybe this has changed.
If you look at kernel/syscalls/stat/stat06.c, there is some code there
to accomodate this problem using setjmp/longjmp.
Get latest updates about Open Source Projects, Conferences and News.