Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 27 Feb 2012 13:49:28 +0100
From:      Bengt Ahlgren <bengta@sics.se>
To:        Konstantin Belousov <kostikbel@gmail.com>
Cc:        emulation@freebsd.org
Subject:   Re: [Bengt Ahlgren] 8.3-PRERELEASE panic in linux emulation
Message-ID:  <uh762es5tlz.fsf@P142.sics.se>
In-Reply-To: <20120225011228.GB55074@deviant.kiev.zoral.com.ua> (Konstantin Belousov's message of "Sat, 25 Feb 2012 03:12:28 %2B0200")
References:  <uh762ewxvpv.fsf@P142.sics.se> <20120224142940.GR55074@deviant.kiev.zoral.com.ua> <uh739a0w39k.fsf@P142.sics.se> <uh7fwe054u4.fsf@P142.sics.se> <20120225011228.GB55074@deviant.kiev.zoral.com.ua>

next in thread | previous in thread | raw e-mail | index | archive | help
Konstantin Belousov <kostikbel@gmail.com> writes:

> On Fri, Feb 24, 2012 at 09:55:31PM +0100, Bengt Ahlgren wrote:
>> Bengt Ahlgren <bengta@sics.se> writes:
>> 
>> > Konstantin Belousov <kostikbel@gmail.com> writes:
>> >
>> >> On Fri, Feb 24, 2012 at 01:27:24PM +0100, Bengt Ahlgren wrote:
>> >>> Hello!
>> >>> 
>> >>> Perhaps emulation@ is a better place to report this problem?
>> >>> 
>> >>> Bengt
>> >>> 
>> >>
>> >>> From: Bengt Ahlgren <bengta@sics.se>
>> >>> To: stable@freebsd.org
>> >>> Subject: 8.3-PRERELEASE panic in linux emulation
>> >>> Date: Thu, 23 Feb 2012 17:26:32 +0100
>> >>> 
>> >>> Hi!
>> >>> 
>> >>> I get a consistent panic when starting acroread after updating to
>> >>> 8.3-PRERELEASE.  An 8.2-STABLE from Feb 4th was OK.  Can provide more
>> >>> info if needed.
>> >>> 
>> >>> Bengt
>> >>> 
>> >>> FreeBSD xx.yy.zz 8.3-PRERELEASE FreeBSD 8.3-PRERELEASE #13 r231999: Wed Feb 22 21:01:38 CET 2012     bengta@P142.sics.se:/usr/obj/usr/src/sys/P142-82  i386
>> >>> 
>> >>> Fatal trap 12: page fault while in kernel mode
>> >>> fault virtual address   = 0xbfbfdffc
>> >>> fault code              = supervisor write, page not present
>> >>> instruction pointer     = 0x20:0xc50b396c
>> >>> stack pointer           = 0x28:0xe7481a6c
>> >>> frame pointer           = 0x28:0xe7481a90
>> >>> code segment            = base 0x0, limit 0xfffff, type 0x1b
>> >>>                         = DPL 0, pres 1, def32 1, gran 1
>> >>> processor eflags        = interrupt enabled, resume, IOPL = 0
>> >>> current process         = 1997 (bash)
>> >>> trap number             = 12
>> >>> panic: page fault
>> >>> KDB: stack backtrace:
>> >>> db_trace_self_wrapper(c091af2a,70797420,78302065,a0d6231,c5d6b600,...) at db_trace_self_wrapper+0x26
>> >>> kdb_backtrace(c0919061,c09b49a0,c0900251,e7481910,e7481910,...) at kdb_backtrace+0x2a
>> >>> panic(c0900251,c0941cab,c5c246e8,1,1,...) at panic+0xaf
>> >>> trap_fatal(c0670d02,0,e7481964,80400,c5c24580,...) at trap_fatal+0x353
>> >>> trap_pfault(e74819cc,bfbfe190,c5c24580,202,c5cecac0,...) at trap_pfault+0x87
>> >>> trap(e7481a2c) at trap+0x453
>> >>> calltrap() at calltrap+0x6
>> >>> --- trap 0xc, eip = 0xc50b396c, esp = 0xe7481a6c, ebp = 0xe7481a90 ---
>> >>> elf_linux_fixup(e7481c0c,e7481b98,c065ca92,c60ffce8,100000,...) at elf_linux_fixup+0x33c
>> >>> kern_execve(c5c24580,e7481c3c,0,8112710,8116cd8,...) at kern_execve+0x7d6
>> >>> linux_execve(c5c24580,e7481cec,c,c,c,...) at linux_execve+0xa7
>> >>> syscall(e7481d28) at syscall+0x372
>> >>> Xint0x80_syscall() at Xint0x80_syscall+0x21
>> >>> --- syscall (11, Linux ELF, linux_execve), eip = 0x281e0d4a, esp = 0xbfbfd644, ebp = 0xbfbfd7e8 ---
>> >>> 
>> >>> #0  doadump () at pcpu.h:244
>> >>> #1  0xc05de609 in boot (howto=260) at /usr/src/sys/kern/kern_shutdown.c:441
>> >>> #2  0xc05de84f in panic (fmt=Variable "fmt" is not available.
>> >>> ) at /usr/src/sys/kern/kern_shutdown.c:614
>> >>> #3  0xc08b22c3 in trap_fatal (frame=0xe7481a2c, eva=3217022972) at /usr/src/sys/i386/i386/trap.c:981
>> >>> #4  0xc08b2357 in trap_pfault (frame=0xe7481a2c, usermode=0, eva=3217022972) at /usr/src/sys/i386/i386/trap.c:843
>> >>> #5  0xc08b3133 in trap (frame=0xe7481a2c) at /usr/src/sys/i386/i386/trap.c:562
>> >>> #6  0xc089bedc in calltrap () at /usr/src/sys/i386/i386/exception.s:168
>> >>> #7  0xc50b396c in elf_linux_fixup (stack_base=0xe7481c0c, imgp=0xe7481b98) at /usr/src/sys/modules/linux/../../i386/linux/linux_sysvec.c:288
>> >>> #8  0xc05ac636 in kern_execve (td=0xc5c24580, args=0xe7481c3c, mac_p=0x0) at /usr/src/sys/kern/kern_exec.c:551
>> >>> #9  0xc50ab387 in linux_execve (td=0xc5c24580, args=0xe7481cec) at /usr/src/sys/modules/linux/../../i386/linux/linux_machdep.c:143
>> >>> #10 0xc08b2902 in syscall (frame=0xe7481d28) at subr_syscall.c:114
>> >>> #11 0xc089bf41 in Xint0x80_syscall () at /usr/src/sys/i386/i386/exception.s:266
>> >>> #12 0x00000033 in ?? ()
>> >>> 
>> >> I am not sure if this is the real cause of your panic, but the line from
>> >> the backtrace indeed has a bug. Please try the change below.
>> >>
>> >> diff --git a/sys/i386/linux/linux_sysvec.c b/sys/i386/linux/linux_sysvec.c
>> >> index 7634138..d4e23e1 100644
>> >> --- a/sys/i386/linux/linux_sysvec.c
>> >> +++ b/sys/i386/linux/linux_sysvec.c
>> >> @@ -227,11 +227,11 @@ linux_fixup(register_t **stack_base, struct image_params *imgp)
>> >>  	argv = *stack_base;
>> >>  	envp = *stack_base + (imgp->args->argc + 1);
>> >>  	(*stack_base)--;
>> >> -	**stack_base = (intptr_t)(void *)envp;
>> >> +	suword(*stack_base, (intptr_t)(void *)envp);
>> >>  	(*stack_base)--;
>> >> -	**stack_base = (intptr_t)(void *)argv;
>> >> +	suword(*stack_base, (intptr_t)(void *)argv);
>> >>  	(*stack_base)--;
>> >> -	**stack_base = imgp->args->argc;
>> >> +	suword(*stack_base, imgp->args->argc);
>> >>  	return (0);
>> >>  }
>> >>  
>> >> @@ -286,7 +286,7 @@ elf_linux_fixup(register_t **stack_base, struct image_params *imgp)
>> >>  	imgp->auxargs = NULL;
>> >>  
>> >>  	(*stack_base)--;
>> >> -	**stack_base = (register_t)imgp->args->argc;
>> >> +	suword(*stack_base, (register_t)imgp->args->argc);
>> >>  	return (0);
>> >>  }
>> >>  
>> >
>> > Thanks for the response!  I will try the patch, but that file has not
>> > been touched since June 2011.  I was suspecting the changes in r231146
>> > and r231148.  If there is no change with your patch I will roll back
>> > those to see what happens.
> This is very unlikely. fadvise() has nothing to do with image activators.
>
>> 
>> No panics so far.  That patch does indeed seem to solve the problem!  I
>> also verified with going back to the old kernel, which again
>> consistently paniced.
> I will commit the change in minutes. Kernel must not access usermode
> addresses directly.
>
> But, does the application that used to panic the system, behave properly ?

Yes, it (acroread8) does behave properly with the patch.  Have not
tested extensively, however.

>> Thanks very much for good work!
>> 
>> I'm a but puzzled though, because that bug must have been there for
>> quite some time without triggering the panic.
> The panic with unpatched kernel looks puzzling. Do you have some
> non-default stack limit ? Can you look at the resource limit values
> for the process initiated the panic ?

Not that I'm aware of.  Unless the acroread launch script does this.  I
don't know how to check this for a running process, but "limits|grep
stack" in a regular shell gives me:

  stacksize               65536 kB

Or, do you mean that I can dig that out of the crash dump?  If so, I'll
need some help with how to.

Bengt



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?uh762es5tlz.fsf>