[PATCH] meson: improve CPU affinity routines check

[Martin Kletzander]

On Mon, Nov 22, 2021 at 03:22:11PM +0400, Roman Bogorodskiy wrote:
>  Martin Kletzander wrote:
>
>> On Sun, Nov 21, 2021 at 07:58:55PM +0400, Roman Bogorodskiy wrote:
>> >Recently, FreeBSD has got sched_get/setaffinity(3) implementations and
>> >the sched.h header as well [1]. To make these routines visible,
>> >users have to define _WITH_CPU_SET_T.
>> >
>> >This breaks current detection. Specifically, meson sees the
>> >sched_getaffinity() symbol and defines WITH_SCHED_GETAFFINITY. This
>> >define unlocks Linux implementation of virProcessSetAffinity() and other
>> >functions, which fails to build on FreeBSD because cpu_set_t is not
>> >visible as _WITH_CPU_SET_T is not defined.
>> >
>> >For now, change detection to the following:
>> >
>> > - Instead of checking sched_getaffinity(), check if 'cpu_set_t' is
>> >   available through sched.h
>> > - Explicitly check the sched.h header instead of assuming its presence
>> >   if WITH_SCHED_SETSCHEDULER is defined
>> >
>>
>> Makes sense, although even the sched_getaffinity() is guarded by the new
>> _WITH_CPU_SET_T so there should be no error with the current code, am I
>
>Nope, as I tried to explain in the commit message, the current code is
>not working properly as it's using cc.has_function() for 'sched_getaffinity'
>and this check passes because it's checking the specified function in
>the standard library, so it works regardless of _WITH_CPU_SET_T. And as
>it finds the sched_getaffinity(), the build still fails because we don't
>set _WITH_CPU_SET_T.
>

Looking at the patches you linked to it seems like sched_getaffinity()
is hidden under #ifdef _WITH_CPU_SET_T, which made me think that
cc.has_function() should not find it.

Anyway the code is fine as it is, I was just wondering.

Reviewed-by: Martin Kletzander <mkletzan at redhat.com>

>> right?  And if I understand correctly we cannot use the FreeBSD
>> implementation as is because they do not have all the CPU_* macros we
>> need, right?
>
>Frankly speaking, I haven't yet tested this implementation. Currently,
>virprocess.c is using the original FreeBSD interface: cpuset_(get|set)affinity.
>Generally, it would be a good thing to use the same implementation for
>both Linux and FreeBSD, however, this Linux-compatible interface in
>FreeBSD is not yet available in any FreeBSD release, so we'll have to keep
>the old code for some time anyway, and also I need to test if it
>actually works for us and figure out how to better detect and pass this _WITH_CPU_SET_T
>with meson.
>

Maybe one day =)

>> >1:
>> >https://cgit.freebsd.org/src/commit/?id=43736b71dd051212d5c55be9fa21c45993017fbb
>> >https://cgit.freebsd.org/src/commit/?id=160b4b922b6021848b6b48afc894d16b879b7af2
>> >https://cgit.freebsd.org/src/commit/?id=90fa9705d5cd29cf11c5dc7319299788dec2546a
>> >
>> >Signed-off-by: Roman Bogorodskiy <bogorodskiy at gmail.com>
>> >---
>> > meson.build           | 4 +++-
>> > src/util/virprocess.c | 8 ++++----
>> > 2 files changed, 7 insertions(+), 5 deletions(-)
>> >
>> >diff --git a/meson.build b/meson.build
>> >index 9022bcfdc9..e4f36e8574 100644
>> >--- a/meson.build
>> >+++ b/meson.build
>> >@@ -553,7 +553,6 @@ functions = [
>> >   'posix_fallocate',
>> >   'posix_memalign',
>> >   'prlimit',
>> >-  'sched_getaffinity',
>> >   'sched_setscheduler',
>> >   'setgroups',
>> >   'setns',
>> >@@ -602,6 +601,7 @@ headers = [
>> >   'net/if.h',
>> >   'pty.h',
>> >   'pwd.h',
>> >+  'sched.h',
>> >   'sys/auxv.h',
>> >   'sys/ioctl.h',
>> >   'sys/mount.h',
>> >@@ -671,6 +671,8 @@ symbols = [
>> >
>> >   # Check for BSD approach for setting MAC addr
>> >   [ 'net/if_dl.h', 'link_addr', '#include <sys/types.h>\n#include <sys/socket.h>' ],
>> >+
>> >+  [ 'sched.h', 'cpu_set_t' ],
>> > ]
>> >
>> > if host_machine.system() == 'linux'
>> >diff --git a/src/util/virprocess.c b/src/util/virprocess.c
>> >index 6de3f36f52..81de90200e 100644
>> >--- a/src/util/virprocess.c
>> >+++ b/src/util/virprocess.c
>> >@@ -35,7 +35,7 @@
>> > # include <sys/time.h>
>> > # include <sys/resource.h>
>> > #endif
>> >-#if WITH_SCHED_SETSCHEDULER
>> >+#if WITH_SCHED_H
>> > # include <sched.h>
>> > #endif
>> >
>> >@@ -480,7 +480,7 @@ int virProcessKillPainfully(pid_t pid, bool force)
>> >     return virProcessKillPainfullyDelay(pid, force, 0, false);
>> > }
>> >
>> >-#if WITH_SCHED_GETAFFINITY
>> >+#if WITH_DECL_CPU_SET_T
>> >
>> > int virProcessSetAffinity(pid_t pid, virBitmap *map, bool quiet)
>> > {
>> >@@ -626,7 +626,7 @@ virProcessGetAffinity(pid_t pid)
>> >     return ret;
>> > }
>> >
>> >-#else /* WITH_SCHED_GETAFFINITY */
>> >+#else /* WITH_DECL_CPU_SET_T */
>> >
>> > int virProcessSetAffinity(pid_t pid G_GNUC_UNUSED,
>> >                           virBitmap *map G_GNUC_UNUSED,
>> >@@ -646,7 +646,7 @@ virProcessGetAffinity(pid_t pid G_GNUC_UNUSED)
>> >                          _("Process CPU affinity is not supported on this platform"));
>> >     return NULL;
>> > }
>> >-#endif /* WITH_SCHED_GETAFFINITY */
>> >+#endif /* WITH_DECL_CPU_SET_T */
>> >
>> >
>> > int virProcessGetPids(pid_t pid, size_t *npids, pid_t **pids)
>> >--
>> >2.33.1
>> >
>
>
>
>Roman Bogorodskiy


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: not available
URL: <http://listman.redhat.com/archives/libvir-list/attachments/20211123/626cb1d1/attachment-0001.sig>