Re: LTP failure
Bryan Kadzban wrote: On Wed, Mar 21, 2007 at 10:31:29AM -0500, Bruce Dubbs wrote: Thanks Bryan. It's still a mystery to me. I tried a 2.6.20.3 kernel last night and it still failed for me. Well, I can try installing 2.6.20.3 to see if it also fails for me; perhaps 2.6.19.1 is an oddball? I know it has some other issues, so I should upgrade it anyway. 2.6.20.3 also passes the test for me (fails the call). I suspect it's because it's 64-bit though. mmap seems likely to be affected by the architecture (more than some other syscalls, anyway). mincore does too. If I still have my 32-bit partition laying around, I can try to upgrade its kernel and see whether 2.6.20.3 fails for me in 32-bit mode. But it'll take a while. signature.asc Description: OpenPGP digital signature -- http://linuxfromscratch.org/mailman/listinfo/lfs-dev FAQ: http://www.linuxfromscratch.org/faq/ Unsubscribe: See the above information page
Re: LTP failure
Bryan Kadzban wrote: Bryan Kadzban wrote: On Wed, Mar 21, 2007 at 10:31:29AM -0500, Bruce Dubbs wrote: Thanks Bryan. It's still a mystery to me. I tried a 2.6.20.3 kernel last night and it still failed for me. Well, I can try installing 2.6.20.3 to see if it also fails for me; perhaps 2.6.19.1 is an oddball? I know it has some other issues, so I should upgrade it anyway. 2.6.20.3 also passes the test for me (fails the call). I suspect it's because it's 64-bit though. mmap seems likely to be affected by the architecture (more than some other syscalls, anyway). mincore does too. If I still have my 32-bit partition laying around, I can try to upgrade its kernel and see whether 2.6.20.3 fails for me in 32-bit mode. But it'll take a while. Thanks Bryan. I've gotten a couple of reports of failures in 32 bit systems but all reports about 64 bit systems pass. I've personally tried 2.4.x, 2.6.12.5, and 2.6.20.3 kernels and all have the same incorrect behavior. -- Bruce -- http://linuxfromscratch.org/mailman/listinfo/lfs-dev FAQ: http://www.linuxfromscratch.org/faq/ Unsubscribe: See the above information page
Re: LTP failure
Bruce Dubbs wrote: I've been trying to figure out why the attached file fails on my LFS systems. It does not fail on FC or RHEL kernels. Neither mincore01 nor mincore02 fail on my (C)LFS system either. But it's not exactly LFS, so it's probably not a great test. It's a multilib x86-64 setup (with a 64-bit kernel, of course). Kernel version is 2.6.19.1 (which I need to upgrade one of these days). I wonder if there was maybe some different behavior added in the kernel since 2.6.16.27? Also, when I run your test program (after changing the (unsigned int) cast to an (unsigned long int) cast and changing the printf format specifier to match, since pointers on my setup are 64 bits wide), I don't get a failure whether I pass an argument to min or not. But again, this system is probably too different from standard LFS to be a great test. The 64-bit stuff especially seems like it'd affect this test, at least a bit. signature.asc Description: OpenPGP digital signature -- http://linuxfromscratch.org/mailman/listinfo/lfs-dev FAQ: http://www.linuxfromscratch.org/faq/ Unsubscribe: See the above information page
Re: LTP failure
Bryan Kadzban wrote: Bruce Dubbs wrote: I've been trying to figure out why the attached file fails on my LFS systems. It does not fail on FC or RHEL kernels. Neither mincore01 nor mincore02 fail on my (C)LFS system either. But it's not exactly LFS, so it's probably not a great test. It's a multilib x86-64 setup (with a 64-bit kernel, of course). Kernel version is 2.6.19.1 (which I need to upgrade one of these days). I wonder if there was maybe some different behavior added in the kernel since 2.6.16.27? Also, when I run your test program (after changing the (unsigned int) cast to an (unsigned long int) cast and changing the printf format specifier to match, since pointers on my setup are 64 bits wide), I don't get a failure whether I pass an argument to min or not. But again, this system is probably too different from standard LFS to be a great test. The 64-bit stuff especially seems like it'd affect this test, at least a bit. Thanks Bryan. It's still a mystery to me. I tried a 2.6.20.3 kernel last night and it still failed for me. Experimenting showed that the printf needs to be before the mmap function to make a difference, so I am hypothesizing that the issue is in mmap. -- Bruce -- http://linuxfromscratch.org/mailman/listinfo/lfs-dev FAQ: http://www.linuxfromscratch.org/faq/ Unsubscribe: See the above information page
Re: LTP failure
On Wed, Mar 21, 2007 at 10:31:29AM -0500, Bruce Dubbs wrote: Thanks Bryan. It's still a mystery to me. I tried a 2.6.20.3 kernel last night and it still failed for me. Well, I can try installing 2.6.20.3 to see if it also fails for me; perhaps 2.6.19.1 is an oddball? I know it has some other issues, so I should upgrade it anyway. However, I've done some other testing on a really-old LFS system (pre 6.2) that's still running kernel 2.4 (2.4.33.6 plus grsecurity), and it passes the test. I've also tested a Debian-testing box running kernel 2.6.18-1-686 (whatever that corresponds to), and it passes as well. Both of these machines run 32-bit kernels. Experimenting showed that the printf needs to be before the mmap function to make a difference, so I am hypothesizing that the issue is in mmap. Entirely possible; there have been a few issues with mmap over the past couple years that I can remember. Maybe one of the fixes broke this test? What happens if you lseek() back to the beginning of the file before calling mmap? I don't think that should matter, but who knows. pgpNOwhnNcYcQ.pgp Description: PGP signature -- http://linuxfromscratch.org/mailman/listinfo/lfs-dev FAQ: http://www.linuxfromscratch.org/faq/ Unsubscribe: See the above information page
Re: LTP failure
results here from an older system (specifically an LFS dev version from around mar 2006) mincore011 PASS : expected failure: errno = 22 (Invalid argument) mincore012 PASS : expected failure: errno = 14 (Bad address) mincore013 FAIL : call succeeded unexpectedly mincore014 PASS : expected failure: errno = 12 (Cannot allocate memory) current kernel version is (recently rebuilt to add USB support): Linux lfsbuild 2.6.15.6-LFS #7 Fri Mar 2 00:55:33 NZDT 2007 i686 i686 i386 GNU/Linux On 3/21/07, Bruce Dubbs [EMAIL PROTECTED] wrote: I've been trying to figure out why the attached file fails on my LFS systems. It does not fail on FC or RHEL kernels. I'd appreciate it if some of you could run the attached program ans send me the results. You just run the binary. If you don't want to do that, the full source to is at http://prdownloads.sourceforge.net/ltp/ltp-full-20070228.tgz?download -- Bruce -- http://linuxfromscratch.org/mailman/listinfo/lfs-dev FAQ: http://www.linuxfromscratch.org/faq/ Unsubscribe: See the above information page -- -- - Steve Crosby -- http://linuxfromscratch.org/mailman/listinfo/lfs-dev FAQ: http://www.linuxfromscratch.org/faq/ Unsubscribe: See the above information page
LTP failure
I've been trying to figure out why the attached file fails on my LFS systems. It does not fail on FC or RHEL kernels. I'd appreciate it if some of you could run the attached program ans send me the results. You just run the binary. If you don't want to do that, the full source to is at http://prdownloads.sourceforge.net/ltp/ltp-full-20070228.tgz?download -- Bruce mincore01 Description: Binary data -- http://linuxfromscratch.org/mailman/listinfo/lfs-dev FAQ: http://www.linuxfromscratch.org/faq/ Unsubscribe: See the above information page
Re: LTP failure
On 3/20/07, Bruce Dubbs [EMAIL PROTECTED] wrote: I've been trying to figure out why the attached file fails on my LFS systems. It does not fail on FC or RHEL kernels. I'd appreciate it if some of you could run the attached program ans send me the results. I got all passes on my system. $ ./mincore01 mincore011 PASS : expected failure: errno = 22 (Invalid argument) mincore012 PASS : expected failure: errno = 14 (Bad address) mincore013 PASS : expected failure: errno = 12 (Cannot allocate memory) mincore014 PASS : expected failure: errno = 12 (Cannot allocate memory) This is on linux-2.6.20.1/glibc-2.3.6. I have some of the DIY tweaks for the toolchain build. -- Dan -- http://linuxfromscratch.org/mailman/listinfo/lfs-dev FAQ: http://www.linuxfromscratch.org/faq/ Unsubscribe: See the above information page
Re: LTP failure
Dan Nicholson wrote:
On 3/20/07, Bruce Dubbs [EMAIL PROTECTED] wrote:
I've been trying to figure out why the attached file fails on my LFS
systems. It does not fail on FC or RHEL kernels.
I'd appreciate it if some of you could run the attached program ans send
me the results.
I got all passes on my system.
$ ./mincore01
mincore011 PASS : expected failure: errno = 22 (Invalid argument)
mincore012 PASS : expected failure: errno = 14 (Bad address)
mincore013 PASS : expected failure: errno = 12 (Cannot allocate memory)
mincore014 PASS : expected failure: errno = 12 (Cannot allocate memory)
This is on linux-2.6.20.1/glibc-2.3.6. I have some of the DIY tweaks
for the toolchain build.
Thanks Dan. I'm really puzzled about this. I pared the code down to
what I have below. compile with `gcc min.c -o min`.
For me, if I do ./min, I get:
PAGESIZE=4096
global_len=8192
global_pointer=0xb7fda000
alloc=2
Result=0
global_vec: 01 01 01 01 01 01
Which is a FAIL. Bit if I do `./min anything`, I get:
argc=2
PAGESIZE=4096
global_len=8192
global_pointer=0xb7f78000
alloc=2
Result=-1
global_vec: 01 01 00 00 00 00
which is a PASS.
The operation we are testing is mincore which is a direct call to the
kernel. The code is in mm/mincore.c.
Its like something is not getting initialized until the printf is run.
My kernel is 2.6.16.27-lfs62-a. I don't see anything in the config that
would make a difference. It's a mystery to me.
-- Bruce
--
cat EOF min.c
#include fcntl.h
#include errno.h
#include sys/mman.h
#include stdlib.h
#include unistd.h
#include sys/types.h
#include sys/stat.h
#include string.h
#include stdio.h
static intPAGESIZE;
static char file_name[]= fooXX;
static char* global_pointer = NULL;
unsigned char*global_vec = NULL;
static intglobal_len = 0;
static intfile_desc = 0;
int main(int ac, char **av)
{
int i;
char* buf;
unsigned char vect[6] = {0,0,0,0,0,0};
if ( ac 1 ) printf(argc=%d\n, ac);
PAGESIZE = getpagesize();
/* global_pointer will point to a mmapped area of global_len bytes */
global_len = PAGESIZE*2;
buf = (char*)malloc(global_len);
memset(buf,42,global_len);
/* create a temporary file */
file_desc = mkstemp(file_name);
/* fill the temporary file with two pages of data */
write(file_desc,buf,global_len);
free(buf);
/* map the file in memory */
global_pointer = (char *)mmap(NULL,global_len,
PROT_READ|PROT_WRITE|PROT_EXEC,MAP_SHARED,file_desc,0);
i = mincore((void*)global_pointer, (size_t)(global_len*5), vect);
printf(PAGESIZE=%d\n, PAGESIZE);
printf(global_len=%d\n, global_len);
printf(global_pointer=0x%x\n, (unsigned int)global_pointer);
//printf(global_vec=0x%x\n, (unsigned int)global_vec);
printf(alloc=%d\n, (global_len+PAGESIZE-1) / PAGESIZE );
printf(Result=%d\n, i);
printf(global_vec: %02x %02x %02x %02x %02x %02x\n,
vect[0], vect[1], vect[2], vect[3], vect[4], vect[5]);
}
EOF
--
http://linuxfromscratch.org/mailman/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page
