arm64: ptrace: avoid using HW_BREAKPOINT_EMPTY for disabled events
authorWill Deacon <will.deacon@arm.com>
Tue, 17 Dec 2013 17:09:08 +0000 (17:09 +0000)
committerCatalin Marinas <catalin.marinas@arm.com>
Thu, 19 Dec 2013 17:41:25 +0000 (17:41 +0000)
commitcdc27c27843248ae7eb0df5fc261dd004eaa5670
tree202d38bcf449e2a9edb9e2e8da9384078b1af660
parent319e2e3f63c348a9b66db4667efa73178e18b17d
arm64: ptrace: avoid using HW_BREAKPOINT_EMPTY for disabled events

Commit 8f34a1da35ae ("arm64: ptrace: use HW_BREAKPOINT_EMPTY type for
disabled breakpoints") fixed an issue with GDB trying to zero breakpoint
control registers. The problem there is that the arch hw_breakpoint code
will attempt to create a (disabled), execute breakpoint of length 0.

This will fail validation and report unexpected failure to GDB. To avoid
this, we treated disabled breakpoints as HW_BREAKPOINT_EMPTY, but that
seems to have broken with recent kernels, causing watchpoints to be
treated as TYPE_INST in the core code and returning ENOSPC for any
further breakpoints.

This patch fixes the problem by prioritising the `enable' field of the
breakpoint: if it is cleared, we simply update the perf_event_attr to
indicate that the thing is disabled and don't bother changing either the
type or the length. This reinforces the behaviour that the breakpoint
control register is essentially read-only apart from the enable bit
when disabling a breakpoint.

Cc: <stable@vger.kernel.org>
Reported-by: Aaron Liu <liucy214@gmail.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
arch/arm64/kernel/ptrace.c