Re: [PATCH v2 0/9] ARM: remove set_fs callers and implementation

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

 



On Sat, Sep 19, 2020 at 10:19 AM Russell King - ARM Linux admin
<linux@xxxxxxxxxxxxxxx> wrote:
>
> On Fri, Sep 18, 2020 at 02:46:15PM +0200, Arnd Bergmann wrote:
> > Hi Christoph, Russell,
> >
> > Here is an updated series for removing set_fs() from arch/arm,
> > based on the previous feedback.
> >
> > I have tested the oabi-compat changes using the LTP tests for the three
> > modified syscalls using an Armv7 kernel and a Debian 5 OABI user space,
> > and I have lightly tested the get_kernel_nofault infrastructure by
> > loading the test_lockup.ko module after setting CONFIG_DEBUG_SPINLOCK.
>
> I'm not too keen on always saving the syscall number, but for the gain
> of getting rid of set_fs() I think it's worth it. However...
>
> I think there are some things to check - what value do you end up
> with as the first number in /proc/self/syscall when you do:
>
> strace cat /proc/self/syscall
>
> ?

> It should be 3, not 0x900003. I suspect you're getting the latter
> with these changes.  IIRC, task_thread_info(task)->syscall needs to
> be the value _without_ the offset, otherwise tracing will break.

It seems broken in different ways, depending on the combination
of kernel and userland:

1. EABI armv5-versatile kernel, EABI Debian 5:
$ cat /proc/self/syscall
0 0x1500000000003 0x1500000000400 0x1500000000400 0x60000013c7800480
0xc0008668c0112f8c 0xc0112d14c68e1f68 0xbeab06f8 0xb6e80d4c
$ strace -f cat /proc/self/syscall
execve("/bin/cat", ["cat", "/proc/self/syscall"], [/* 16 vars */]) =
-1 EINTR (Interrupted system call)
dup(2)                                  = -1 EINTR (Interrupted system call)
write(2, "strace: exec: Interrupted system "..., 38) = -1 EINTR
(Interrupted system call)
exit_group(1)                           = ?

2. EABI kernel, OABI Debian 5:
$ cat /proc/self/syscall
3 0x1500000000003 0x13ccc00000400 0x1500000000400 0x60000013c7800480
0xc0008de0c0112f8c 0xc0112d14c7313f68 0xbeed27d0 0xb6eab324
$ strace cat /proc/self/syscall
execve("/bin/cat", ["cat", "/proc/self/syscall"], [/* 16 vars */]) = -1090648236
--- SIGILL (Illegal instruction) @ 0 (0) ---
+++ killed by SIGILL +++

3. OABI kernel, OABI Debian 5:
 cat /proc/self/syscall
9437187 0x1500000000003 0x13ccc00000400 0x1500000000400 0x100060000013
0x15000c72cff6c 0xc72cfe9000000000 0xbece27d0 0xb6f2f324
$ strace cat /proc/self/syscall
execve("/bin/cat", ["cat", "/proc/self/syscall"], [/* 16 vars */]) = -1095141548
--- SIGILL (Illegal instruction) @ 0 (0) ---
+++ killed by SIGILL +++

I suspect the OABI strace in Debian is broken since it crashes on
both kernels. I'll look into fixing the output without strace first then.

       Arnd



[Index of Archives]     [Linux Kernel]     [Kernel Newbies]     [x86 Platform Driver]     [Netdev]     [Linux Wireless]     [Netfilter]     [Bugtraq]     [Linux Filesystems]     [Yosemite Discussion]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Samba]     [Device Mapper]

  Powered by Linux