Re: [uml-devel] UML 2.6.10-rc2-mm3 + incremental crashes

[Bodo Stroesser <bstroesser@fujitsu-siemens.com>]

[-- Attachment #1: Type: text/plain, Size: 2339 bytes --]

Blaisorblade wrote:
> On Thursday 25 November 2004 19:21, Bodo Stroesser wrote:
> 
>>Have started to test with the new 2.6.10-rc2-mm3 + incrementals tree.
>>
>>First, it didn't compile, because there are inconsistent definitions
>>for __pgd_alloc in include/linux/mm.h and arch/um/kernel/mem.c.
>>I added "fastcall" in arch/um/kernel/mem.c, then the build succeeded.
>>
>>But now the kernel crashes on startup. Please ignore the error messages
>>regarding ext3.ko. For the "first shot" I didn't update the initrd.
>>But the kernel should handle the fs in ext2-mode anyway.
>>
>>Bodo
> 
> It seems (I've just read it in the ML, not very clearly) that this is 
> addressed by:
Yes. That caker pointed out on #uml.
But there is another crash, that happens sometimes only, since it's dependend
on timing. To fix this other one, you need the appended patch too.
> 
> http://marc.theaimsgroup.com/?l=linux-kernel&m=110117212115156&w=2
> 
> I.e. it's a host kernel bug with quotas and DevFS (or whatever). The marked 
> portion of the traces match.
> 
>>Error messages from SKAS:
> 
> 
>>Call Trace:
> 
> [...]
> 
>>  [<a0073939>] invalidate_bh_lru+0x19/0x40
>>  [<a007396a>] invalidate_bh_lrus+0xa/0x10
> 
> **
> 
>>  [<a009e3e8>] vfs_quota_off+0xa8/0x220
>>  [<a008d231>] do_umount+0x1b1/0x1d0
>>  [<a008d2b7>] sys_umount+0x67/0x90
> 
> **
> 
>>  [<a00182bb>] handle_page_fault+0x16b/0x1f0
>>  [<a0018465>] segv+0x85/0x1e0
>>  [<a001d53a>] execute_syscall_skas+0xda/0x120
>>  [<a00172cc>] record_syscall_start+0x4c/0x60
>>  [<a001d5b0>] handle_syscall+0x30/0x60
>>  [<a001c7e0>] handle_trap+0x30/0x130
>>  [<a001cc38>] userspace+0x188/0x1e0
>>  [<a001d190>] fork_handler+0x0/0x80
>>  [<a01ab228>] __restore+0x0/0x8
>>  [<a001d190>] fork_handler+0x0/0x80
>>  [<a01ab3e1>] __kill+0x11/0x20
>>
>>
>>
>>Error messages from TT:
> 
> 
>>Call Trace:
> 
> [...]
> **
> 
>>  [<a009e3e8>] vfs_quota_off+0xa8/0x220
>>  [<a008d231>] do_umount+0x1b1/0x1d0
>>  [<a008d2b7>] sys_umount+0x67/0x90
> 
> **
> 
>>  [<a001170f>] sys_execve+0x3f/0x50
>>  [<a01ab474>] sigemptyset+0x24/0x40
>>  [<a001aaba>] execute_syscall_tt+0xda/0x120
>>  [<a00172cc>] record_syscall_start+0x4c/0x60
>>  [<a001ab3f>] syscall_handler_tt+0x3f/0x70
>>  [<a001bd6d>] sig_handler_common_tt+0x8d/0x110
>>  [<a002e2ae>] sig_handler+0x1e/0x40
>>  [<a01ab228>] __restore+0x0/0x8
> 
> 
> 


[-- Attachment #2: timer-context.patch --]
[-- Type: text/x-diff, Size: 491 bytes --]

--- a/arch/um/kernel/time_kern.c	2004-11-29 16:10:48.766132965 +0100
+++ b/arch/um/kernel/time_kern.c	2004-11-29 16:11:04.423088221 +0100
@@ -155,7 +155,7 @@ void __const_udelay(unsigned long usecs)
 void timer_handler(int sig, union uml_pt_regs *regs)
 {
 	local_irq_disable();
-	update_process_times(user_context(UPT_SP(regs)));
+	update_process_times(CHOOSE_MODE(user_context(UPT_SP(regs)), (regs)->skas.is_user));
 	local_irq_enable();
 	if(current_thread->cpu == 0)
 		timer_irq(regs);