Re: [PATCH v3 0/5] ARM64: disable irq between breakpoint and step exception

From: Pratyush Anand
Date: Fri Aug 25 2017 - 02:05:53 EST


Hi Mark, James and Will,

What is the take on this series?

Just to summarize:

- James took first three patches of the series and replaced 4th and 5th patch with his 3 patch series [3].
- My patchset disables interrupt while exiting from HW breakpoint/watchpoint and SW breakpoint handlers, and re-enables after single step handling..but it will work reliably for single stepping even if an interrupt is generated during breakpoint/watchpoint handler execution.
- James's approach keeps interrupt enabled, but it might fail if a hardware breakpoint is instrumented on an ISR called between breakpoint and step exception.

@James, I understand there would be other things on priority..do you have a plan to fix the pitfall and send a new series. If not, can I address Peter's concern on patch 1/5 and send v4?

[3] https://www.spinics.net/lists/arm-kernel/msg598214.html

On Monday 31 July 2017 04:10 PM, Pratyush Anand wrote:
v2 -> v3
- Moved step_needed from uapi structure to kernel only structure
- Re-enable interrupt if stepped instruction faults
- Modified register_wide_hw_breakpoint() to accept step_needed arg
v2 was here: http://marc.info/?l=linux-arm-kernel&m=149942910730496&w=2

v1 -> v2:
- patch 1 of v1 has been modified to patch 1-3 of v2.
- Introduced a new event attribute step_needed and implemented
hw_breakpoint_needs_single_step() (patch 1)
- Replaced usage of is_default_overflow_handler() with
hw_breakpoint_needs_single_step(). (patch 2)
- Modified sample test to set set step_needed bit field (patch 3)
v1 was here: http://marc.info/?l=linux-arm-kernel&m=149910958418708&w=2

samples/hw_breakpoint/data_breakpoint.c passes with x86_64 but fails with
ARM64. Even though it has been NAKed previously on upstream [1, 2], I have
tried to come up with patches which can resolve it for ARM64 as well.

I noticed that even perf step exception can go into an infinite loop if CPU
receives an interrupt while executing breakpoint/watchpoint handler. So,
event though we are not concerned about above test, we will have to find a
solution for the perf issue.

This patchset attempts to resolve both the issue. Please review.
Since, it also takes care of SW breakpoint, so I hope kgdb should also be
fine. However, I have not tested that.
@Takahiro: Will it be possible to test these patches for kgdb.

[1] http://marc.info/?l=linux-arm-kernel&m=149580777524910&w=2
[2] http://lists.infradead.org/pipermail/linux-arm-kernel/2016-April/425266.html


Pratyush Anand (5):
hw_breakpoint: Add step_needed event attribute
arm64: use hw_breakpoint_needs_single_step() to decide if step is
needed
register_wide_hw_breakpoint(): modify to accept step_needed arg
arm64: disable irq between breakpoint and step exception
arm64: fault: re-enable irq if it was disabled for single stepping

arch/arm64/kernel/debug-monitors.c | 3 +++
arch/arm64/kernel/hw_breakpoint.c | 10 +++++-----
arch/arm64/mm/fault.c | 28 ++++++++++++++++++++++++----
arch/x86/kernel/kgdb.c | 2 +-
include/linux/hw_breakpoint.h | 10 ++++++++--
include/linux/perf_event.h | 6 ++++++
kernel/events/core.c | 2 ++
kernel/events/hw_breakpoint.c | 4 +++-
samples/hw_breakpoint/data_breakpoint.c | 3 ++-
9 files changed, 54 insertions(+), 14 deletions(-)


--
Regards
Pratyush