| Message ID | 20190906060115.9460-1-mark-pk.tsai@mediatek.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | perf/hw_breakpoint: Fix arch_hw_breakpoint use-before-initialization | expand |
Hi, On Thu, Sep 5, 2019 at 11:01 PM Mark-PK Tsai <mark-pk.tsai@mediatek.com> wrote: > > If we disable the compiler's auto-initialization feature > (-fplugin-arg-structleak_plugin-byref or -ftrivial-auto-var-init=pattern) > is disabled, arch_hw_breakpoint may be used before initialization after > the change 9a4903dde2c86. > (perf/hw_breakpoint: Split attribute parse and commit) > > On our arm platform, the struct step_ctrl in arch_hw_breakpoint, which > used to be zero-initialized by kzalloc, may be used in > arch_install_hw_breakpoint without initialization. > > Signed-off-by: Mark-PK Tsai <mark-pk.tsai@mediatek.com> > Cc: YJ Chiang <yj.chiang@mediatek.com> > Cc: Alix Wu <alix.wu@mediatek.com> > --- > kernel/events/hw_breakpoint.c | 4 ++-- > 1 file changed, 2 insertions(+), 2 deletions(-) Stable should pick this up, please. It landed in mainline as commit 310aa0a25b33 ("perf/hw_breakpoint: Fix arch_hw_breakpoint use-before-initialization"). * I have confirmed that it cleanly applies to and fixes a kernel based on v4.19.75, so picking it back to kernels 4.19+ is the easiest. * I have confirmed that my test shows that hardware breakpoints fail on my arm32 test machine on v4.18.20 and on v4.17.0. They last worked on 4.16. Picking this patch alone is not sufficient to make 4.17 and 4.18 work again. Bisecting shows that the first breakage was the merge resolution that happened in commit 2d074918fb15 ("Merge branch 'perf/urgent' into perf/core"). Specifically both parents of that merge passed my test but the result of the merge didn't pass my test. If anyone cares about 4.17 and 4.18 at this point, I will leave it as an exercise to them to try to get them working again. -Doug
On Thu, Oct 10, 2019 at 10:44:13AM -0700, Doug Anderson wrote: > Hi, > > On Thu, Sep 5, 2019 at 11:01 PM Mark-PK Tsai <mark-pk.tsai@mediatek.com> wrote: > > > > If we disable the compiler's auto-initialization feature > > (-fplugin-arg-structleak_plugin-byref or -ftrivial-auto-var-init=pattern) > > is disabled, arch_hw_breakpoint may be used before initialization after > > the change 9a4903dde2c86. > > (perf/hw_breakpoint: Split attribute parse and commit) > > > > On our arm platform, the struct step_ctrl in arch_hw_breakpoint, which > > used to be zero-initialized by kzalloc, may be used in > > arch_install_hw_breakpoint without initialization. > > > > Signed-off-by: Mark-PK Tsai <mark-pk.tsai@mediatek.com> > > Cc: YJ Chiang <yj.chiang@mediatek.com> > > Cc: Alix Wu <alix.wu@mediatek.com> > > --- > > kernel/events/hw_breakpoint.c | 4 ++-- > > 1 file changed, 2 insertions(+), 2 deletions(-) > > Stable should pick this up, please. It landed in mainline as commit > 310aa0a25b33 ("perf/hw_breakpoint: Fix arch_hw_breakpoint > use-before-initialization"). > > * I have confirmed that it cleanly applies to and fixes a kernel based > on v4.19.75, so picking it back to kernels 4.19+ is the easiest. > > * I have confirmed that my test shows that hardware breakpoints fail > on my arm32 test machine on v4.18.20 and on v4.17.0. They last worked > on 4.16. Picking this patch alone is not sufficient to make 4.17 and > 4.18 work again. Bisecting shows that the first breakage was the > merge resolution that happened in commit 2d074918fb15 ("Merge branch > 'perf/urgent' into perf/core"). Specifically both parents of that > merge passed my test but the result of the merge didn't pass my test. > If anyone cares about 4.17 and 4.18 at this point, I will leave it as > an exercise to them to try to get them working again. Now queued up to 4.19.y, thanks. greg k-h
diff --git a/kernel/events/hw_breakpoint.c b/kernel/events/hw_breakpoint.c index c5cd852fe86b..8fb842394924 100644 --- a/kernel/events/hw_breakpoint.c +++ b/kernel/events/hw_breakpoint.c @@ -413,7 +413,7 @@ static int hw_breakpoint_parse(struct perf_event *bp, int register_perf_hw_breakpoint(struct perf_event *bp) { - struct arch_hw_breakpoint hw; + struct arch_hw_breakpoint hw = {0}; int err; err = reserve_bp_slot(bp); @@ -461,7 +461,7 @@ int modify_user_hw_breakpoint_check(struct perf_event *bp, struct perf_event_attr *attr, bool check) { - struct arch_hw_breakpoint hw; + struct arch_hw_breakpoint hw = {0}; int err; err = hw_breakpoint_parse(bp, attr, &hw);
If we disable the compiler's auto-initialization feature (-fplugin-arg-structleak_plugin-byref or -ftrivial-auto-var-init=pattern) is disabled, arch_hw_breakpoint may be used before initialization after the change 9a4903dde2c86. (perf/hw_breakpoint: Split attribute parse and commit) On our arm platform, the struct step_ctrl in arch_hw_breakpoint, which used to be zero-initialized by kzalloc, may be used in arch_install_hw_breakpoint without initialization. Signed-off-by: Mark-PK Tsai <mark-pk.tsai@mediatek.com> Cc: YJ Chiang <yj.chiang@mediatek.com> Cc: Alix Wu <alix.wu@mediatek.com> --- kernel/events/hw_breakpoint.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) -- 2.18.0