On Mon, Apr 30, 2018 at 10:20:08AM -0700, Greg Kroah-Hartman wrote:
On Mon, Apr 30, 2018 at 02:11:55PM -0300, Murilo Opsfelder Araujo wrote:
On 04/11/2018 05:21 PM, Murilo Opsfelder Araújo wrote:
On 04/11/2018 04:33 PM, Greg Kroah-Hartman wrote:
On Wed, Apr 11, 2018 at 03:41:35PM -0300, Murilo Opsfelder Araujo wrote:
Using linux-3.18.y branch, perf build fails with the following:
$ make -s -j16 -C tools/perf V=1 WERROR=0 NO_LIBUNWIND=1 HAVE_CPLUS_DEMANGLE=1 NO_GTK2=1 NO_LIBNUMA=1 NO_STRLCPY=1 NO_BIONIC=1 prefix=/usr DESTDIR=/tmp/builddir/build all [...] util/session.c: In function ‘__perf_session__process_pipe_events’: util/session.c:1093:36: error: ‘oe’ undeclared (first use in this function) ordered_events__set_copy_on_queue(oe, true); ^ util/session.c:1093:36: note: each undeclared identifier is reported only once for each function it appears inThis patch fixes it for linux-3.18.y branch.
Why is this failing now? Has it always been broken? Is there an upstream patch that fixed this instead?
Hi, Greg.
We've caught this build issue this week. I'm not sure since when it's failing.
The upstream patch that fixes this is:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?i...
commit fa713a4eb9cebe5dec71b1bd11429603e17d841d Author: Arnaldo Carvalho de Melo acme@redhat.com Date: Tue Mar 3 11:48:12 2015 -0300
perf ordered_events: Untangle from perf_sessionBut it doesn't apply straightforwardly. At fa713a4eb9cebe5dec71b1bd11429603e17d841d, perf API and perf_evlist struct are slightly different from linux-3.18.y.
A list of upstream patches for a clean backport (I think) would be:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?i...
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?i...
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?i...
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?i...
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?i...
Do you prefer all above patches backported to linux-3.18.y or just the one-line fix I sent?
Fixes: 95b33b99cdd6 ("perf inject: Copy events when reordering events in pipe mode")
Was this just a bad backport?
Some patches were left behind, I guess.
thanks,
greg k-h
Hi, Greg.
I've realized that perf build on v3.18.107 is still broken, actually, it got a bit worse than v3.18.104.
After applying my patch to fix the undeclared 'oe', perf build fails with the following errors:
$ make -s -j16 -C tools/perf V=1 WERROR=0 NO_LIBUNWIND=1 HAVE_CPLUS_DEMANGLE=1 NO_GTK2=1 NO_LIBNUMA=1 NO_STRLCPY=1 NO_BIONIC=1 prefix=/usr DESTDIR=/tmp/builddir/build all [...] tests/code-reading.c: In function ‘read_object_code’: tests/code-reading.c:145:19: error: ‘KMOD_DECOMP_LEN’ undeclared (first use in this function) char decomp_name[KMOD_DECOMP_LEN]; ^ tests/code-reading.c:145:19: note: each undeclared identifier is reported only once for each function it appears in tests/code-reading.c:208:2: warning: implicit declaration of function ‘dso__needs_decompress’ [-Wimplicit-function-declaration] if (dso__needs_decompress(al.map->dso)) { ^ tests/code-reading.c:208:2: warning: nested extern declaration of ‘dso__needs_decompress’ [-Wnested-externs] tests/code-reading.c:209:3: warning: implicit declaration of function ‘dso__decompress_kmodule_path’ [-Wimplicit-function-declaration] if (dso__decompress_kmodule_path(al.map->dso, objdump_name, ^ tests/code-reading.c:209:3: warning: nested extern declaration of ‘dso__decompress_kmodule_path’ [-Wnested-externs] tests/code-reading.c:145:7: warning: unused variable ‘decomp_name’ [-Wunused-variable] char decomp_name[KMOD_DECOMP_LEN]; ^
Apparently, this was introduced by commit 091b3b4ba55daf8b28f40794aefdaa0bdb5af7a1:
commit 091b3b4ba55daf8b28f40794aefdaa0bdb5af7a1 Author: Namhyung Kim namhyung@kernel.org Date: Thu Jun 8 16:31:07 2017 +0900
perf tests: Decompress kernel module before objdumpNamhyung, are you fixing this?
Ugh, I can fix this, I just did so on 4.4.y...
Nope, this is a mess. When did perf last build on 3.18.y? It really looks like no one cares about it...
greg k-h