Here is a patch which I had, sometime back. It is on an older version of
LTP.
Cheers
Suzuki
, Suzuki Poulose wrote:
, renxiu liang wrote:
On Sun, 2010-09-19 at 12:06 +0530, Suzuki Poulose wrote:
Hello Renxiu,
Sorry for the late reply. Could you please resend the patch ? I don't
find the patch in the e-mail.
Hi Suzuki
attached failure logs on different archs and patch for you.
I don't understand why there should be ACCERR. The mmap is done with
PROT_WRIT|PROT_READ. Could we get more info on which condition triggers
this ?
You may use GDB to break for the ACCERR case and take a look at the
instruction which triggers this.
Coming back to your patch :
I think we should ensure the si_addr is within the map. (map_address,
map_address+mapsize-1) ? We should not let the other SIGSEGVs go through.
It may be like :
case SIGSEGV:
if ((si_addr >= map_address) &&
(si_addr < (map_address + map_size) ) {
...
..
}
Thanks
Suzuki
thanks for help review this.
BR
Renxiu
Thanks
Suzuki
, renxiu liang wrote:
On Wed, 2010-09-01 at 00:27 -0700, Garrett Cooper wrote:
On Wed, Sep 1, 2010 at 12:26 AM, Garrett Cooper<yaneg...@gmail.com>
wrote:
On Mon, Aug 30, 2010 at 6:51 PM, renxiu liang
<renxiu.li...@windriver.com> wrote:
On Wed, 2010-08-25 at 20:36 -0700, Garrett Cooper wrote:
On Wed, Aug 18, 2010 at 12:33 AM, renxiu liang
<renxiu.li...@windriver.com> wrote:
Hi,
We met the mtest06 failure several times; this is because in
mtest06/mmap1.c, when handling the signal 11 in sig_handler, it
doesn't
cover another two race conditions:
one is si_code equals "SEGV_MAPERR" but si_address does not
equal to
map_address; and one is si_code equals to "SEGV_ACCERR";
see below error log:
<<<test_start>>>
tag=mtest06 stime=1270902396
cmdline=" mmap1 -x 0.05"
contacts=""
analysis=exit
initiation_status="ok"
<<<test_output>>>
mmap1 0 INFO : pid[5456]: map, change contents, unmap files
1000 times
mmap1 0 INFO : created thread[1216369840]
mmap1 0 INFO : pid[5456] - read contents of memory 0x48002000
1000 times
mmap1 0 INFO : page fault occurred at 0x48002000
mmap1 0 INFO : page fault occurred at 0x48002000
mmap1 0 INFO : page fault occurred at 0x48002000
mmap1 0 INFO : page fault occurred at 0x48002000
......
mmap1 0 INFO : page fault occurred at 0x48002000
mmap1 0 INFO : page fault occurred at 0x48002000
caught unexpected signal - 11 --- exiting
<<<execution_status>>>
duration=1 termination_type=exited termination_id=255 corefile=no
cutime=0 cstime=4
<<<test_end>>>
I made a patch to cover all the three race conditions in one
case in
sig_handler(),
then run mtest06 repeatedly on different archs, test will not
fail by
catching signal 11.
Could you print out what the value of info->si_code and
info->si_addr
are at the time of the fault? Also, which architecture(s) are you
running into this issue on?
Hi Garrett,
Sorry for the late reply; See attached mtest06 logs, I printed
out the
values of si_code, si_addr, map_address;
it is reproducible on arm, x86, ppc and mips; though it is not
reproducible every time, but can be reproducible if run the case
repeatedly.
Nothing is jumping out at me as being incorrect, so I think it's best
that I bring in a better pair of eyes, just in case.
Suzuki-san,
Could you please help review this patch?
Hi Garrett,
No reply from Suzuki-san after almost three weeks; I'm afraid
Suzuki-san
didn't catch the mail, could you find someone else to review the patch?
thanks!
BR,
Renxiu
CCing Suzuki-san...
-Garrett
diff -Naur ltp-full-20091231.orig/testcases/kernel/mem/mtest06/mmap1.c
ltp-full-20091231/testcases/kernel/mem/mtest06/mmap1.c
--- ltp-full-20091231.orig/testcases/kernel/mem/mtest06/mmap1.c 2009-10-13
19:30:46.000000000 +0530
+++ ltp-full-20091231/testcases/kernel/mem/mtest06/mmap1.c 2010-03-15
16:30:33.000000000 +0530
@@ -138,6 +138,7 @@
int verbose_print = FALSE;/* when called with -v print more info */
caddr_t *map_address; /* address of the file mapped. */
sigjmp_buf jmpbuf; /* argument to sigsetjmp and siglongjmp */
+long map_len;
char *TCID = "mmap1";
int TST_TOTAL = 1;
@@ -168,14 +169,22 @@
_exit(0);
case SIGSEGV:
- if (info->si_code == SEGV_MAPERR &&
- info->si_addr == map_address)
+ if ( info->si_addr == map_address ||
+ (map_address && (info->si_addr >= (void*) map_address &&
+ info->si_addr < (void *)((long)map_address +
map_len))
+ )
+ )
{
tst_resm(TINFO,
"page fault occurred at %p",
map_address);
longjmp(jmpbuf, 1);
- }
+ } else {
+ fprintf(stderr, "caught SIGSEGV at %p while mmap is (%p,%p) -
exiting\n",
+ info->si_addr, map_address,
(void*)((long)map_address + map_len));
+ _exit(-1);
+ }
+
default:
fprintf(stderr, "caught unexpected signal - %d --- exiting\n",
signal);
@@ -293,6 +302,9 @@
(long *)args;
long exit_val = 0; /* exit value of the pthread */
+ /* The size of the map */
+ map_len = mwuargs[1];
+
tst_resm(TINFO, "pid[%d]: map, change contents, unmap files %d times",
getpid(), (int)mwuargs[2]);
if (verbose_print)
------------------------------------------------------------------------------
Start uncovering the many advantages of virtual appliances
and start using them to simplify application deployment and
accelerate your shift to cloud computing.
http://p.sf.net/sfu/novell-sfdev2dev
_______________________________________________
Ltp-list mailing list
Ltp-list@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/ltp-list
