On Mon, Jul 4, 2011 at 11:26 AM, Carlos O'Donell
<carlos@xxxxxxxxxxxxxxxx> wrote:
> On Mon, Jul 4, 2011 at 11:21 AM, Carlos O'Donell
> <carlos@xxxxxxxxxxxxxxxx> wrote:
>> pthread_create: ALLOCATE_STACK failed.
>>
>> timer_helper_thread: pthread_create ret 22
>> ~~~
>>
>> The second threads stack fails to allocate.
>
> 1309792564.144002 set_robust_list(0x4000b4f0, 0xc) = 0
> 1309792564.144002 rt_sigtimedwait([RTMIN], {si_signo=SIGRTMIN,
> si_code=SI_TIMER, si_pid=4, si_uid=0, si_value={int=90456,
> ptr=0x16158}}, NULL, 8) = 37
> 1309792564.368004 write(1, "timer_helper_thread: creating th"..., 40) = 40
> 1309792564.368004 write(1, "\n", 1) = 1
> 1309792564.368004 mmap(NULL, 8388608, PROT_READ|PROT_WRITE|PROT_EXEC,
> MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x41349000
> 1309792564.368004 mprotect(0x41b47000, 4096, PROT_NONE) = 0
> 1309792564.368004 write(1, "do_clone: ARCH_CLONE...\n", 24) = 24
> 1309792564.368004 write(1, "\n", 1) = 1
> 1309792564.372004 clone(child_stack=0x41349040,
> flags=CLONE_VM|CLONE_FS|CLONE_FILES|CLONE_SIGHAND|CLONE_THREAD|CLONE_SYSVSEM|CLONE_SETTLS|CLONE_PARENT_SETTID|CLONE_CHILD_CLEARTID,
> parent_tidptr=0x41b484e8, tls=0x41b48900, child_tidptr=0x41b484e8) =
> 21795
> 1309792564.372004 write(1, "timer_helper_thread: pthread_cre"..., 42) = 42
> 1309792564.372004 rt_sigtimedwait([RTMIN], {si_signo=SIGRTMIN,
> si_code=SI_TIMER, si_pid=5, si_uid=0, si_value={int=90520,
> ptr=0x16198}}, NULL, 8) = 37
> 1309792564.476005 write(1, "timer_helper_thread: creating th"..., 40) = 40
> 1309792564.476005 write(1, "\n", 1) = 1
> 1309792564.480005 mprotect(0x41b47480, 4096,
> PROT_READ|PROT_WRITE|PROT_EXEC) = -1 EINVAL (Invalid argument)
> 1309792564.480005 munmap(0x41349000, 8388608) = 0
> 1309792564.480005 write(1, "pthread_create: ALLOCATE_STACK f"..., 39) = 39
>
> Why would this mprotect fail with -EINVAL? It's not page aligned?
>
> You can see the earlier mmap/mprotect from thread one's stack looks
> page aligned.
>
> I'll look into this.
Alright, the documetnation for mprotect definately says "addr must be
aligned to a page boundary."
And matches the return failure:
"EINVAL addr is not a valid pointer, or not a multiple of the system page size."
So it would seem that this just works in the _S_G_D case because of
the way addresses line up.
I'm instrumenting the stack allocation code and guard setup to see if
I missed anything.
This part of glibc has been heavily patched by me to enable _S_G_U,
and it's probably the ugliest part of the thread startup code.
Cheers,
Carlos.
--
To unsubscribe from this list: send the line "unsubscribe linux-parisc" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
