Interrupted system call handling

[Ulrich Weigand]

GDB does not handle debugging during interrupted system calls correctly.

Related testsuite failured include:
FAIL: gdb.base/interrupt.exp: continue
FAIL: gdb.base/interrupt.exp: echo data
FAIL: gdb.base/interrupt.exp: Send Control-C, second time
FAIL: gdb.base/interrupt.exp: signal SIGINT (the program is no longer running)
FAIL: gdb.base/interrupt.exp: echo more data (timeout)
FAIL: gdb.base/interrupt.exp: send end of file

Fixing this will likely involve kernel changes as well. It seems the kernel in some cases returns to userspace with a -512 ERESTARTSYS return value -- this should never happen.

[Ulrich Weigand (uweigand)]

This problem also causes failures in gdb.threads/hand-call-in-threads.exp:

FAIL: gdb.threads/hand-call-in-threads.exp: hand call, thread 2
FAIL: gdb.threads/hand-call-in-threads.exp: hand call, thread 3
FAIL: gdb.threads/hand-call-in-threads.exp: hand call, thread 4
FAIL: gdb.threads/hand-call-in-threads.exp: hand call, thread 5

call hand_call()^M
^M
Program received signal SIGSEGV, Segmentation fault.^M
0x00000000 in ?? ()^M

[Ulrich Weigand (uweigand)]

The underlying problem is a kernel issue. If we have a ptrace intercept in an interrupted system call, and GDB changes the PC in order to effect an inferior call, the kernel's do_signal routine still thinks it needs to decrement the PC in order to restart the system call. This causes execution to continue at the wrong address.

This problem was fixed in various platform specific do_signal routines over the years, but apparently not yet on ARM.

[Loïc Minier (lool)]

Arnd, Ulrich told me you'd be looking into this

[Ulrich Weigand (uweigand)]

Re-added gdb-linaro since this affects GDB, and a solution might also require GDB changes in addition to kernel changes.

[Arnd Bergmann (arnd-arndb)]

Ulrich tells me offline that he has an experimental patch for this. Uli, please attach the patch so I can review and/or forward it.

[Ulrich Weigand (uweigand)]

Linux kernel patch that fixes the problem for me.
See patch for detailed description of the problem and proposed solution.

[Ulrich Weigand (uweigand)]

Kernel patch posted to linux-arm-kernel.

[Deepak Saxena (dsaxena-linaro)]

Hi Ulrich, was this patch accepted upstream and has it been posted to Nicolas for inclusion in the Linaro kernel?

[Nicolas Pitre (npitre)]

> Hi Ulrich, was this patch accepted upstream and has it been posted to
> Nicolas for inclusion in the Linaro kernel?

It is in the Linaro kernel. It is not in mainline yet.
Someone should put it in RMK's patch system.

[Ulrich Weigand (uweigand)]

> Someone should put it in RMK's patch system.

Arnd put it in a couple of days ago:
http://www.arm.linux.org.uk/developer/patches/viewpatch.php?id=6892/1

[Ulrich Weigand (uweigand)]

Since with this solution, there is no GDB change required, I'm setting the Linaro GDB project status to Invalid.