----- Original Message ----- > From: "Francesco RUNDO" <[email protected]> > To: "Jan Stancek" <[email protected]> > Cc: [email protected], [email protected] > Sent: Friday, 21 February, 2014 11:25:03 AM > Subject: RE: [LTP] [PATCH] madvise/madvise03.c: Changed test exit code > > > >>Why should this case be unsupported when addr1 is backed by shared memory? > >> TEST(madvise(addr1, 4096, MADV_REMOVE)); > > The reported scenario MADV_REMOVE is managed -at kernel side- by > madvise_remove() function on KERNEL_ROOT/mm/madvice.c. > From this file I got (mm/madvice.c): > > == > static long madvise_remove(struct vm_area_struct *vma, > struct vm_area_struct **prev, > unsigned long start, unsigned long end) > { > [snip] > error = do_fallocate(f, FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE, > offset, end - start); > [snip] > } > == > > Now, "do_fallocate()" needs fallocate() op. but on ext3 this is not supported > i.e. ENOTSUP.
This is the "unexpected part" for me. What I expect and get on x86_64 with 3.10 based kernel is call to shmem_fallocate(): madvise(start=0x7fcff6d66000 len_in=0x1000 behavior=0x9) Call Trace: 0xffffffff81150130 : shmem_fallocate+0x0/0x300 [kernel] 0xffffffff8122d056 : shm_fallocate+0x26/0x40 [kernel] 0xffffffff811a0089 : do_fallocate+0xf9/0x190 [kernel] 0xffffffff8115ac95 : SyS_madvise+0x355/0x7a0 [kernel] 0xffffffff815d05d9 : system_call_fastpath+0x16/0x1b [kernel] Do you have possibility to run original testcase with strace? # strace ./madvise03 2>&1 | grep -v abcdefghijklmnopqrstuvwxyz12345 Regards, Jan > This is the reason for which I consider this scenario as not supported and > not ENOSYS as the system call is supported as the remaining scenarios worked > fine (MADV_DONTFORK and MADV_DOFORK). > > Regards > FR > > > > Regards, > > FR > > > > >>Regards, > > >>Jan > > > > > > > > > > > > > Regards, > > > FR > > > > > > -----Original Message----- > > > From: Wanlong Gao [mailto:[email protected]] > > > Sent: Friday, February 21, 2014 9:02 AM > > > To: Francesco RUNDO > > > Cc: [email protected] > > > Subject: Re: [LTP] [PATCH] madvise/madvise03.c: Changed test exit > > > code > > > > > > On 02/21/2014 03:51 PM, Francesco RUNDO wrote: > > > > Changed exit code from TFAIL to TCONF in case of unsupported syscall. > > > > The "madvise" syscall needs kernel support as well as fs support > > > > as -for specific advice- it requires specific filesystem > > > > operations such as fallocate(). > > > > > > Could you paste the running log on not supported system? > > > > > > Thanks, > > > Wanlong Gao > > > > > > > So, if kernel support is missed or fs doesn't provide specific > > > > operation needed for such "advice" scenario of that syscall, the > > > > right exit code (according to LTP defs) is not TFAIL as it is not > > > > a failure but it is TCONF as the syscall is unsupported. > > > > > > > > Signed-off-by: Francesco Rundo <[email protected]> > > > > --- > > > > testcases/kernel/syscalls/madvise/madvise03.c | 17 +++++++++++++---- > > > > 1 files changed, 13 insertions(+), 4 deletions(-) > > > > > > > > diff --git a/testcases/kernel/syscalls/madvise/madvise03.c > > > > b/testcases/kernel/syscalls/madvise/madvise03.c > > > > index b0c17e8..23dd9a6 100644 > > > > --- a/testcases/kernel/syscalls/madvise/madvise03.c > > > > +++ b/testcases/kernel/syscalls/madvise/madvise03.c > > > > @@ -160,10 +160,19 @@ static void cleanup(void) static void > > > > check_and_print(char *advice) { > > > > if (TEST_RETURN == -1) { > > > > - tst_resm(TFAIL, > > > > - "madvise test for %s failed with " > > > > - "return = %ld, errno = %d : %s", > > > > - advice, TEST_RETURN, TEST_ERRNO, > > > > strerror(TEST_ERRNO)); > > > > + if (TEST_ERRNO != ENOTSUP) { > > > > + tst_resm(TFAIL, > > > > + "madvise test for %s failed with " > > > > + "return = %ld, errno = %d : %s", > > > > + advice, TEST_RETURN, TEST_ERRNO, > > > > + strerror(TEST_ERRNO)); > > > > + } else { > > > > + tst_resm(TCONF, > > > > + "madvise test for %s unsupported " > > > > + "return = %ld, errno = %d : %s", > > > > + advice, TEST_RETURN, TEST_ERRNO, > > > > + strerror(TEST_ERRNO)); > > > > + } > > > > } else if (STD_FUNCTIONAL_TEST) { > > > > tst_resm(TPASS, "madvise test for %s PASSED", advice); > > > > } > > > > > > > > > > > > > -------------------------------------------------------------------- > > > -- > > > -------- Managing the Performance of Cloud-Based Applications Take > > > advantage of what the Cloud has to offer - Avoid Common Pitfalls. > > > Read the Whitepaper. > > > http://pubads.g.doubleclick.net/gampad/clk?id=121054471&iu=/4140/ostg. > > > clktrk _______________________________________________ > > > Ltp-list mailing list > > > [email protected] > > > https://lists.sourceforge.net/lists/listinfo/ltp-list > > > > > > ------------------------------------------------------------------------------ Managing the Performance of Cloud-Based Applications Take advantage of what the Cloud has to offer - Avoid Common Pitfalls. Read the Whitepaper. http://pubads.g.doubleclick.net/gampad/clk?id=121054471&iu=/4140/ostg.clktrk _______________________________________________ Ltp-list mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/ltp-list
