mmiotrace uses the ftrace framework for dynamic patching, so enabling it
does not cause any performance impact.
Having mmiotrace enabled is very useful for debugging nvidia / nouveau
driver problems, as we often use mmiotraces for nouveau development /
debugging.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Unlike the other joystick modules joydev is not just needed for old analog
joysticks, but also to provide /dev/js? devices with new usb connected
joysticks.
Many games still use the (old) /dev/js interface, and will not work ootb
unless kernel-modules-extra is installed. Moving jodev.ko to the main
kernel package fixes this.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Yes, this is a big uber-patch. But with the git-am apply method, it still
splits them out in the prepped tree. There's really no advantage to listing
78 patches in the spec at this point as kdbus is likely to go all-or-nothing
if/when it is merged.
So. Because of the joys of how perf builds up its LDFLAGS, builds started
failing in rawhide with the latest perl build. See, perf thinks it is a
good idea to look at perl (and python too) and see what ld options were
used to build it, then just slam those into what it uses. That's cute, and
probably kind of works, but with the hardening changes it totally broke
the build. One of the final links failed with:
/usr/bin/ld: perf-in.o: relocation R_X86_64_32 against `.rodata.str1.1' can not be used when making a shared object; recompile with -fPIC
perf-in.o: error adding symbols: Bad value
because it sucked in the hardened build LD options from perl, but didn't
we didn't pass in the required CFLAGS options. Isn't that special?
Fix this by specifying both LDFLAGS (so it isn't magically getting applied
without thought) and EXTRA_CFLAGS for the perf build.
I hate everything.
Josh Boyer
Hans de Goede
Peter Robinson