ppc64le: Tests failed (2): <no-mmap-inval.t> <reg-fd-only.t> #1207

vt-alt · 2024-08-17T11:44:10Z

JFYI, two tests fail on ppc64 on Linux v6.6.46 for liburing-2.7. They succeed or skipped on x86_64.

+ make runtests
...
[00:01:36] Running test no-mmap-inval.t                                        Got -2, wanted -EFAULT
[00:01:36] Test no-mmap-inval.t failed with ret 1
...
[00:01:49] Running test reg-fd-only.t                                          ring setup failed
[00:01:49] test 8 failed
[00:01:49] Test reg-fd-only.t failed with ret 1

On x86_64:

[00:01:22] Running test no-mmap-inval.t                                        0 sec
...
[00:01:29] Running test reg-fd-only.t                                          Enable huge pages to test big rings
[00:01:29] Skipped

The text was updated successfully, but these errors were encountered:

vt-alt · 2024-08-19T14:29:02Z

Also, the same version suddenly got on aarch64:

[00:00:40] Running test accept-non-empty.t                                     Test accept-non-empty.t failed with ret 78
...
[00:02:40] Tests failed (1): <accept-non-empty.t>

Test for IORING_FEAT_RECVSEND_BUNDLE before setting up the connection thread, or we could be tearing down pthread data at inopportune moments leading to odd behavior. Link: #1207 Fixes: 184e6ec ("test/accept-non-empty: add accept IORING_CQE_F_SOCK_NONEMPTY test") Signed-off-by: Jens Axboe <axboe@kernel.dk>

axboe · 2024-08-19T15:03:24Z

Pushed a fix for accept-non-empty, that was a bug in the test.

For ppc64, the -ENOENT for mmap-no-inval is very (very) odd. For reg-fd-only, I'll push a commit to dump 'ret'. Can you try and re-run it? I'm wondering what it's returning. Maybe both are the same arch odditiy and it'll be -ENOENT?!

Link: #1207 Signed-off-by: Jens Axboe <axboe@kernel.dk>

vt-alt · 2024-08-19T15:16:55Z

Thanks. Now with ebd6c8f it failed only on ppc64le (and i586 where is usually exclude sqpoll-sleep.t). Kernel is also changed to v6.6.47.

ppc64le :

[00:01:36] Running test no-mmap-inval.t                                        Got -2, wanted -EFAULT
[00:01:36] Test no-mmap-inval.t failed with ret 1
...
[00:01:50] Running test reg-fd-only.t                                          ring setup failed: -2
[00:01:50] test 8 failed
[00:01:50] Test reg-fd-only.t failed with ret 1
...
[00:01:52] Running test send-zerocopy.t                                        invalid cqe->res -90 expected 65536
[00:01:52] send failed fixed buf 0, conn 0, addr 1, cork 0
[00:01:52] test_inet_send() failed (defer_taskrun 0)
[00:01:52] Test send-zerocopy.t failed with ret 1
...
[00:02:12] Tests failed (3): <no-mmap-inval.t> <reg-fd-only.t> <send-zerocopy.t>

i586:

[00:02:01] Running test sqpoll-sleep.t                                         Test sqpoll-sleep.t failed with ret 1

isilence · 2024-08-19T16:04:29Z

Thanks. Now with ebd6c8f it failed only on ppc64le (and i586 where is usually exclude sqpoll-sleep.t). Kernel is also changed to v6.6.47.
* ppc64le :

...

[00:01:52] Running test send-zerocopy.t invalid cqe->res -90 expected 65536
[00:01:52] send failed fixed buf 0, conn 0, addr 1, cork 0

It's UDP for which we "expect" 65536 bytes in a datagram, more than usually supported by UDP. Looks the test wasn't prepared for 16K pages.

if (!tcp && len > 4 * page_sz)
	continue; // skip test

vt-alt · 2024-10-31T00:19:33Z

For 2.8 test failures on ppc64le on Linux 6.11.5:

[00:01:39] Running test no-mmap-inval.t                                        Got -2, wanted -EFAULT
[00:01:39] Test no-mmap-inval.t failed with ret 1

[00:02:01] Running test reg-fd-only.t                                          ring setup failed: -2
[00:02:01] test 8 failed
[00:02:01] Test reg-fd-only.t failed with ret 1

[00:02:03] Running test recvsend_bundle.t                                      failed recv cqe: -105
[00:02:03] test d failed
[00:02:03] TCP test case (classic=0) failed
[00:02:03] Test recvsend_bundle.t failed with ret 1

[00:03:11] Running test timeout.t                                              child failed 0
[00:03:11] test_timeout_link_cancel failed
[00:03:11] Test timeout.t failed with ret 1

[00:03:17] Tests failed (4): <no-mmap-inval.t> <reg-fd-only.t> <recvsend_bundle.t> <timeout.t>

Additionally, 1 test fail on i586:

[00:02:32] Running test sqpoll-sleep.t                                         Test sqpoll-sleep.t failed with ret 1

[00:02:49] Tests failed (1): <sqpoll-sleep.t>

Temporary build logs:
https://git.altlinux.org/tasks/361211/build/100/ppc64le/log
https://git.altlinux.org/tasks/361211/build/100/i586/log

On x86_64 and aarch64, with same build env and same kernel version tests do not fail.

vt-alt · 2024-10-31T00:39:03Z

I additionally tested on 6.12-rc5, and the list of failed tests is identical across all architectures.

[00:03:44] Test run complete, kernel: 6.12.0-6.12-alt0.rc5 #1 SMP Sun Oct 27 23:47:43 UTC 2024
[00:03:44] Tests failed (5): <no-mmap-inval.t> <reg-fd-only.t> <recvsend_bundle.t> <recvsend_bundle-inc.t> <timeout.t>

[00:03:15] Test run complete, kernel: 6.12.0-6.12-alt0.rc5 #1 SMP PREEMPT_DYNAMIC Sun Oct 27 23:46:51 UTC 2024
[00:03:15] Tests failed (1): <sqpoll-sleep.t>

Temporary build logs:
https://git.altlinux.org/tasks/361212/build/100/ppc64le/log
https://git.altlinux.org/tasks/361212/build/100/i586/log

axboe · 2024-10-31T00:54:08Z

Thanks for running these. I'll check x86, but I don't have any powerpc to test on... Oh maybe this is a page size thing. What page size is your ppc box running?

vt-alt · 2024-10-31T00:57:35Z

$ getconf PAGESIZE
65536

100 usec / 1 ms is a bit short, make it 100x larger in the hopes that it'll fix the test case on some boxes. Link: #1207 Signed-off-by: Jens Axboe <axboe@kernel.dk>

axboe · 2024-10-31T00:58:21Z

Can you test sqpoll-sleep after the commit I just made?

axboe · 2024-10-31T00:59:33Z

Can you try and strace no-mmap-inval on ppc and attach it here? It should be using page size dependent code already.

This is most likely because there are no huge pages available, so just skip the test in that case. Link: #1207 Signed-off-by: Jens Axboe <axboe@kernel.dk>

We should loop until WIFEXITED() is true. Link: #1207 Signed-off-by: Jens Axboe <axboe@kernel.dk>

axboe · 2024-10-31T01:05:29Z

Pushed some fixes, hopefully fixing some of them.

vt-alt · 2024-10-31T01:24:47Z

Thanks. After updates applied, up to 59c0cb3 2024-10-30 test/timeout: properly loop around waitpid() status.
i586 still have failure:

[00:04:23] Tests failed (1): <sqpoll-sleep.t>

ppc64le:

[00:03:52] Tests failed (3): <no-mmap-inval.t> <recvsend_bundle.t> <recvsend_bundle-inc.t>

strace -v -f test/no-mmap-inval.t

[00:00:24] #1 SMP Sun Oct 2++ strace -v -f test/no-mmap-inval.t
[00:00:25] execve("test/no-mmap-inval.t", ["test/no-mmap-inval.t"], ["RPM_PYTHON_COMPILE_INCLUDE=/usr/"..., "RPM_FIXUP_TOPDIR=", "RPM_PYTHON3_SELF_PROV_PATH=", "RPM_SOURCE_DIR=/usr/src/RPM/SOUR"..., "RPM_PKG_CONTENTS_INDEX_BIN=/.hos"..., "RPM_PYTHON_LIB_PATH=", "RPM_PYTHON3_VERSION=unknown", "G_BROKEN_FILENAMES=1", "HISTSIZE=999", "HOSTNAME=localhost.localdomain", "RPM_DEBUGINFO_STRIPPED_TERMINATE"..., "RPM_PYTHON3_SITELIBDIR=/usr/lib6"..., "RPM_PYTHON=/usr/bin/python2.7", "RPM_PYTHON3_COMPILE_EXCLUDE=/usr"..., "RPM_PYTHON_COMPILE_DEEP=20", "RPM_PYTHON3_SITELIBDIR_NOARCH=/u"..., "RPM_PYTHON_COMPILE_SKIP_X=1", "RPM_PYTHON_COMPILE_EXCLUDE=/usr/"..., "RPM_LIB=lib64", "RPM_FIXUP_METHOD=binconfig pkgco"..., "RPM_LD_PRELOAD_python=/usr/lib64"..., "PWD=/usr/src/RPM/BUILD/liburing-"..., "RPM_CLEANUP_TOPDIR=", "SOURCE_DATE_EPOCH=1730337221", "LOGNAME=root", "RPM_PYTHON3_COMPILE_SKIP_X=1", "RPM_VERIFY_ELF_METHOD=strict", "RPM_FILES_TO_LD_PRELOAD_python= "..., "RPM_CLEANUP_SKIPLIST=", "RPM_ARCH=ppc64le", "RPM_PYTHON3_COMPILE_CLEAN=1", "RPM_DATADIR=/usr/share", "HOME=/usr/src", "RPM_FILES_TO_LD_PRELOAD_python3="..., "PERL_USE_UNSAFE_INC=1", "RPM_PERL_REQ_METHOD=normal", "RPM_PYTHON_REQ_METHOD=slight", "RPM_TARGET_ARCH=ppc64le", "RPM_PYTHON3=/usr/bin/python3", "RPM_FINDPROV_LIB_PATH=", "RPM_COMPRESS_TOPDIR=/usr", "TMPDIR=/tmp", "RPM_VERIFY_ELF_TOPDIR=", "RPM_CHECK_CONTENTS_METHOD=defaul"..., "RPM_FINDPROV_TOPDIR=", "RPM_PACKAGE_RELEASE=alt1.test.2", "RPM_DEBUGINFO_SKIPLIST=", "RPM_PYTHON_COMPILE_CLEAN=1", "RPM_OS=linux", "RPM_VERIFY_ELF_SKIPLIST=", "MAKEFLAGS=-w -O PAM_SO_SUFFIX=", "RPM_CHECK_CONTENTS_SKIPLIST=", "TERM=dumb", "RPM_PYTHON3_REQ_METHOD=slight", "USER=root", "RPM_PYTHON3_LIBDIR=/usr/lib64/py"..., "RPM_PYTHON3_PATH=/usr/lib64/pyth"..., "PAM_SO_SUFFIX=", "RPM_TARGET_OS=linux", "SHLVL=3", "RPM_BUILD_DIR=/usr/src/RPM/BUILD", "RPM_FINDREQ_TOPDIR=", "SCRIPT=/usr/src/tmp/vm.yp4UP1yb8"..., "RPM_FIXUP_SKIPLIST=", "PAM_NAME_SUFFIX=", "RPM_PYTHON2_PATH=/usr/lib64/pyth"..., "RPM_PYTHON3_COMPILE_DEEP=20", "RPM_OPT_FLAGS=-pipe -frecord-gcc"..., "RPM_PYTHON3_REQ_HIER=yes", "RPM_FINDREQ_SKIPLIST=/usr/share/"..., "RPM_PYTHON3_IMPORT_PATH=", "RPM_DOC_DIR=/usr/share/doc", "RPM_VERIFY_INFO_METHOD=normal", "RPM_PACKAGE_VERSION=2.8", "RPM_STRICT_INTERDEPS=sisyphus.36"..., "RPM_PYTHON3_COMPILE_INCLUDE=/usr"..., "RPM_PYTHON_MODULE_DECLARED=", "RPM_PYTHON_REQ_SKIP=", "RPM_LIBDIR=/usr/lib64", "RPM_LD_PRELOAD_python3=/usr/lib6"..., "RPM_PYTHON_COMPILE_METHOD=ALL", "RPM_PYTHON3_REQ_SKIP=", "PATH=/usr/src/bin:/usr/bin:/bin:"..., "RPM_CHECK_CONTENTS_TOPDIR=", "HISTFILESIZE=9999", "MAIL=/var/mail/builder", "RPM_FINDPROV_SKIPLIST=/usr/share"..., "RPM_FINDPACKAGE_PATH=", "RPM_COMPRESS_SKIPLIST=", "RPM_PACKAGE_NAME=liburing", "RPM_CLEANUP_METHOD=auto", "RPM_BUILD_ROOT=/usr/src/tmp/libu"..., "OLDPWD=/usr/src/RPM/BUILD/liburi"..., "RPM_COMPRESS_METHOD=auto", "RPM_PYTHON3_LIB_PATH=", "_=/usr/bin/strace"]) = 0
[00:00:25] brk(NULL)                               = 0x10016f70000
[00:00:25] access("/etc/ld.so.preload", R_OK)      = -1 ENOENT (No such file or directory)
[00:00:25] openat(AT_FDCWD, "/etc/ld.so.cache", O_RDONLY|O_CLOEXEC) = 3
[00:00:25] newfstatat(3, "", {st_dev=makedev(0, 0x14), st_ino=53631, st_mode=S_IFREG|0644, st_nlink=1, st_uid=1031, st_gid=1031, st_blksize=196608, st_blocks=128, st_size=8599, st_atime=1730337271 /* 2024-10-31T01:14:31.422995818+0000 */, st_atime_nsec=422995818, st_mtime=1730337271 /* 2024-10-31T01:14:31.421995880+0000 */, st_mtime_nsec=421995880, st_ctime=1730337271 /* 2024-10-31T01:14:31.421995880+0000 */, st_ctime_nsec=421995880}, AT_EMPTY_PATH) = 0
[00:00:25] mmap(NULL, 8599, PROT_READ, MAP_PRIVATE, 3, 0) = 0x7fffaf200000
[00:00:25] close(3)                                = 0
[00:00:25] openat(AT_FDCWD, "/lib64/libc.so.6", O_RDONLY|O_CLOEXEC) = 3
[00:00:25] read(3, "\177ELF\2\1\1\3\0\0\0\0\0\0\0\0\3\0\25\0\1\0\0\0\240\243\2\0\0\0\0\0"..., 832) = 832
[00:00:25] newfstatat(3, "", {st_dev=makedev(0, 0x14), st_ino=2969, st_mode=S_IFREG|0755, st_nlink=1, st_uid=1031, st_gid=1031, st_blksize=196608, st_blocks=4864, st_size=2439000, st_atime=1730337245 /* 2024-10-31T01:14:05.859595745+0000 */, st_atime_nsec=859595745, st_mtime=1714366800 /* 2024-04-29T05:00:00+0000 */, st_mtime_nsec=0, st_ctime=1730337243 /* 2024-10-31T01:14:03.321754580+0000 */, st_ctime_nsec=321754580}, AT_EMPTY_PATH) = 0
[00:00:25] mmap(NULL, 2482960, PROT_READ|PROT_EXEC, MAP_PRIVATE|MAP_DENYWRITE, 3, 0) = 0x7fffaefa0000
[00:00:25] mmap(0x7fffaf1e0000, 131072, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0x240000) = 0x7fffaf1e0000
[00:00:25] close(3)                                = 0
[00:00:25] set_tid_address(0x7fffaf2a2e10)         = 176
[00:00:25] set_robust_list(0x7fffaf2a2e20, 24)     = 0
[00:00:25] rseq(0x7fffaf2a3460, 0x20, 0, 0xfe5000b) = 0
[00:00:25] mprotect(0x7fffaf1e0000, 65536, PROT_READ) = 0
[00:00:25] mprotect(0x100ad0000, 65536, PROT_READ) = 0
[00:00:25] mprotect(0x7fffaf290000, 65536, PROT_READ) = 0
[00:00:25] prlimit64(0, RLIMIT_STACK, NULL, {rlim_cur=8192*1024, rlim_max=RLIM64_INFINITY}) = 0
[00:00:25] munmap(0x7fffaf200000, 8599)            = 0
[00:00:25] getrandom("\x57\x15\x16\x31\xd6\xa0\x58\xe8", 8, GRND_NONBLOCK) = 8
[00:00:25] brk(NULL)                               = 0x10016f70000
[00:00:25] brk(0x10016fb0000)                      = 0x10016fb0000
[00:00:25] mmap(NULL, 65536, PROT_READ|PROT_WRITE, MAP_SHARED|MAP_ANONYMOUS, -1, 0) = 0x7fffaf200000
[00:00:25] mmap(NULL, 2097152, PROT_READ|PROT_WRITE, MAP_SHARED|MAP_ANONYMOUS|MAP_HUGETLB, -1, 0) = -1 ENOENT (No such file or directory)
[00:00:25] munmap(0x7fffaf200000, 1)               = 0
[00:00:25] write(2, "Got -2, wanted -EFAULT\n", 23Got -2, wanted -EFAULT
[00:00:25] ) = 23
[00:00:25] exit_group(1)                           = ?
[00:00:25] +++ exited with 1 +++

Temporary build logs
https://git.altlinux.org/tasks/361214/build/100/i586/log
https://git.altlinux.org/tasks/361214/build/100/ppc64le/log

See commit becdca8 for details. Link: #1207 Signed-off-by: Jens Axboe <axboe@kernel.dk>

axboe · 2024-10-31T01:27:57Z

The bundle one needs a bit more investigation. no-mmap-inval should skip now too on ppc. I'll check the sqpoll-sleep on x86, that's very odd.

Rewrite this test to be a bit better: - Read wakeup properly with IO_URING_READ_ONCE() - Check if wakeup has been seen - Check elapsed time before wakeup flag is seen - Prepare and push a nop request first, to ensure the thread is up and running Hopefully this will fix the quirks with this test. Link: #1207 Signed-off-by: Jens Axboe <axboe@kernel.dk>

axboe · 2024-11-01T16:51:09Z

Pushed another fix for sqpoll-sleep, can you give it a spin on x86?

axboe added a commit that referenced this issue Aug 19, 2024

test/reg-fd-only: dump 'ret' if ring setup fails unexpectedly

ebd6c8f

Link: #1207 Signed-off-by: Jens Axboe <axboe@kernel.dk>

axboe added a commit that referenced this issue Oct 31, 2024

test/sqpoll-sleep: use coarser time for sleep checking

3fd36a9

100 usec / 1 ms is a bit short, make it 100x larger in the hopes that it'll fix the test case on some boxes. Link: #1207 Signed-off-by: Jens Axboe <axboe@kernel.dk>

axboe added a commit that referenced this issue Oct 31, 2024

test/reg-fd-only: skip on -ENOENT return from mmap

becdca8

This is most likely because there are no huge pages available, so just skip the test in that case. Link: #1207 Signed-off-by: Jens Axboe <axboe@kernel.dk>

axboe added a commit that referenced this issue Oct 31, 2024

test/timeout: properly loop around waitpid() status

59c0cb3

We should loop until WIFEXITED() is true. Link: #1207 Signed-off-by: Jens Axboe <axboe@kernel.dk>

axboe added a commit that referenced this issue Oct 31, 2024

test/no-mmap-inval: skip for -ENOENT mmap return

5123e95

See commit becdca8 for details. Link: #1207 Signed-off-by: Jens Axboe <axboe@kernel.dk>

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

ppc64le: Tests failed (2): <no-mmap-inval.t> <reg-fd-only.t> #1207

ppc64le: Tests failed (2): <no-mmap-inval.t> <reg-fd-only.t> #1207

vt-alt commented Aug 17, 2024

vt-alt commented Aug 19, 2024

axboe commented Aug 19, 2024

vt-alt commented Aug 19, 2024

isilence commented Aug 19, 2024 •

edited

Loading

vt-alt commented Oct 31, 2024

vt-alt commented Oct 31, 2024

axboe commented Oct 31, 2024

vt-alt commented Oct 31, 2024

axboe commented Oct 31, 2024

axboe commented Oct 31, 2024

axboe commented Oct 31, 2024

vt-alt commented Oct 31, 2024 •

edited

Loading

axboe commented Oct 31, 2024

axboe commented Nov 1, 2024

ppc64le: Tests failed (2): <no-mmap-inval.t> <reg-fd-only.t> #1207

ppc64le: Tests failed (2): <no-mmap-inval.t> <reg-fd-only.t> #1207

Comments

vt-alt commented Aug 17, 2024

vt-alt commented Aug 19, 2024

axboe commented Aug 19, 2024

vt-alt commented Aug 19, 2024

isilence commented Aug 19, 2024 • edited Loading

vt-alt commented Oct 31, 2024

vt-alt commented Oct 31, 2024

axboe commented Oct 31, 2024

vt-alt commented Oct 31, 2024

axboe commented Oct 31, 2024

axboe commented Oct 31, 2024

axboe commented Oct 31, 2024

vt-alt commented Oct 31, 2024 • edited Loading

axboe commented Oct 31, 2024

axboe commented Nov 1, 2024

isilence commented Aug 19, 2024 •

edited

Loading

vt-alt commented Oct 31, 2024 •

edited

Loading