linux-arm-kernel.lists.infradead.org archive mirror
 help / color / mirror / Atom feed
From: "Madhavan T. Venkataraman" <madvenka@linux.microsoft.com>
To: Mark Rutland <mark.rutland@arm.com>
Cc: broonie@kernel.org, jpoimboe@redhat.com, ardb@kernel.org,
	nobuta.keiya@fujitsu.com, sjitindarsingh@gmail.com,
	catalin.marinas@arm.com, will@kernel.org, jmorris@namei.org,
	pasha.tatashin@soleen.com, jthierry@redhat.com,
	linux-arm-kernel@lists.infradead.org,
	live-patching@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [RFC PATCH v6 1/3] arm64: Improve the unwinder return value
Date: Thu, 29 Jul 2021 08:54:58 -0500	[thread overview]
Message-ID: <52686cb6-573c-03ca-06c2-67ae07c91243@linux.microsoft.com> (raw)
In-Reply-To: <20210728165635.GA47345@C02TD0UTHF1T.local>

Thanks for the review. Responses inline...

On 7/28/21 11:56 AM, Mark Rutland wrote:
> On Wed, Jun 30, 2021 at 05:33:54PM -0500, madvenka@linux.microsoft.com wrote:
>> From: "Madhavan T. Venkataraman" <madvenka@linux.microsoft.com>
>>
>> Currently, the unwinder returns a tri-state return value:
>>
>> 	0		means "continue with the unwind"
>> 	-ENOENT		means "successful termination of the stack trace"
>> 	-EINVAL		means "fatal error, abort the stack trace"
>>
>> This is confusing. To fix this, define an enumeration of different return
>> codes to make it clear. Handle the return codes in all of the unwind
>> consumers.
> 
> I agree the tri-state is confusing, and I also generally agree that
> enums are preferabel to a set of error codes. However, I don't think
> this is quite the right abstraction; more on that below.
> 

OK.

>>
>> Signed-off-by: Madhavan T. Venkataraman <madvenka@linux.microsoft.com>
>> ---
>>  arch/arm64/include/asm/stacktrace.h | 14 ++++++--
>>  arch/arm64/kernel/perf_callchain.c  |  5 ++-
>>  arch/arm64/kernel/process.c         |  8 +++--
>>  arch/arm64/kernel/return_address.c  | 10 ++++--
>>  arch/arm64/kernel/stacktrace.c      | 53 ++++++++++++++++-------------
>>  arch/arm64/kernel/time.c            |  9 +++--
>>  6 files changed, 64 insertions(+), 35 deletions(-)
>>
>> diff --git a/arch/arm64/include/asm/stacktrace.h b/arch/arm64/include/asm/stacktrace.h
>> index eb29b1fe8255..6fcd58553fb1 100644
>> --- a/arch/arm64/include/asm/stacktrace.h
>> +++ b/arch/arm64/include/asm/stacktrace.h
>> @@ -30,6 +30,12 @@ struct stack_info {
>>  	enum stack_type type;
>>  };
>>  
>> +enum unwind_rc {
>> +	UNWIND_CONTINUE,		/* No errors encountered */
>> +	UNWIND_ABORT,			/* Fatal errors encountered */
>> +	UNWIND_FINISH,			/* End of stack reached successfully */
>> +};
> 
> Generally, there are a bunch of properties we might need to check for an
> unwind step relating to reliabiltiy (e.g. as you add
> UNWIND_CONTINUE_WITH_RISK in the next patch), and I'd prefer that we
> check those properties on the struct stackframe, and simplify
> unwind_frame() to return a bool.
> 
> Something akin to the x86 unwinders, where the main loop looks like:
> 
> for (unwind_start(&state, ...);
>      !unwind_done(&state) && !unwind_error(&state);
>      unwind_next_frame(&state) {
> 	...
> }
> 
> That way we don't have to grow the enum to handle every variation that
> we can think of, and it's simple enough for users to check the
> properties with the helpers.
> 

I can do that.

>> +
>>  /*
>>   * A snapshot of a frame record or fp/lr register values, along with some
>>   * accounting information necessary for robust unwinding.
>> @@ -61,7 +67,8 @@ struct stackframe {
>>  #endif
>>  };
>>  
>> -extern int unwind_frame(struct task_struct *tsk, struct stackframe *frame);
>> +extern enum unwind_rc unwind_frame(struct task_struct *tsk,
>> +				   struct stackframe *frame);
>>  extern void walk_stackframe(struct task_struct *tsk, struct stackframe *frame,
>>  			    bool (*fn)(void *, unsigned long), void *data);
>>  extern void dump_backtrace(struct pt_regs *regs, struct task_struct *tsk,
>> @@ -148,8 +155,8 @@ static inline bool on_accessible_stack(const struct task_struct *tsk,
>>  	return false;
>>  }
>>  
>> -static inline void start_backtrace(struct stackframe *frame,
>> -				   unsigned long fp, unsigned long pc)
>> +static inline enum unwind_rc start_backtrace(struct stackframe *frame,
>> +					     unsigned long fp, unsigned long pc)
>>  {
>>  	frame->fp = fp;
>>  	frame->pc = pc;
>> @@ -169,6 +176,7 @@ static inline void start_backtrace(struct stackframe *frame,
>>  	bitmap_zero(frame->stacks_done, __NR_STACK_TYPES);
>>  	frame->prev_fp = 0;
>>  	frame->prev_type = STACK_TYPE_UNKNOWN;
>> +	return UNWIND_CONTINUE;
>>  }
>>  
>>  #endif	/* __ASM_STACKTRACE_H */
>> diff --git a/arch/arm64/kernel/perf_callchain.c b/arch/arm64/kernel/perf_callchain.c
>> index 88ff471b0bce..f459208149ae 100644
>> --- a/arch/arm64/kernel/perf_callchain.c
>> +++ b/arch/arm64/kernel/perf_callchain.c
>> @@ -148,13 +148,16 @@ void perf_callchain_kernel(struct perf_callchain_entry_ctx *entry,
>>  			   struct pt_regs *regs)
>>  {
>>  	struct stackframe frame;
>> +	enum unwind_rc rc;
>>  
>>  	if (perf_guest_cbs && perf_guest_cbs->is_in_guest()) {
>>  		/* We don't support guest os callchain now */
>>  		return;
>>  	}
>>  
>> -	start_backtrace(&frame, regs->regs[29], regs->pc);
>> +	rc = start_backtrace(&frame, regs->regs[29], regs->pc);
>> +	if (rc == UNWIND_FINISH || rc == UNWIND_ABORT)
>> +		return;
>>  	walk_stackframe(current, &frame, callchain_trace, entry);
> 
> As a first step, could we convert this over to arch_stack_walk()?
> 

OK.

>>  }
>>  
>> diff --git a/arch/arm64/kernel/process.c b/arch/arm64/kernel/process.c
>> index 6e60aa3b5ea9..e9c763b44fd4 100644
>> --- a/arch/arm64/kernel/process.c
>> +++ b/arch/arm64/kernel/process.c
>> @@ -573,6 +573,7 @@ unsigned long get_wchan(struct task_struct *p)
>>  	struct stackframe frame;
>>  	unsigned long stack_page, ret = 0;
>>  	int count = 0;
>> +	enum unwind_rc rc;
>>  	if (!p || p == current || p->state == TASK_RUNNING)
>>  		return 0;
>>  
>> @@ -580,10 +581,13 @@ unsigned long get_wchan(struct task_struct *p)
>>  	if (!stack_page)
>>  		return 0;
>>  
>> -	start_backtrace(&frame, thread_saved_fp(p), thread_saved_pc(p));
>> +	rc = start_backtrace(&frame, thread_saved_fp(p), thread_saved_pc(p));
>> +	if (rc == UNWIND_FINISH || rc == UNWIND_ABORT)
>> +		return 0;
>>  
>>  	do {
>> -		if (unwind_frame(p, &frame))
>> +		rc = unwind_frame(p, &frame);
>> +		if (rc == UNWIND_FINISH || rc == UNWIND_ABORT)
>>  			goto out;
>>  		if (!in_sched_functions(frame.pc)) {
>>  			ret = frame.pc;
> 
> Likewise, can we convert this to use arch_stack_walk()?
> 

OK.

>> diff --git a/arch/arm64/kernel/return_address.c b/arch/arm64/kernel/return_address.c
>> index a6d18755652f..1224e043e98f 100644
>> --- a/arch/arm64/kernel/return_address.c
>> +++ b/arch/arm64/kernel/return_address.c
>> @@ -36,13 +36,17 @@ void *return_address(unsigned int level)
>>  {
>>  	struct return_address_data data;
>>  	struct stackframe frame;
>> +	enum unwind_rc rc;
>>  
>>  	data.level = level + 2;
>>  	data.addr = NULL;
>>  
>> -	start_backtrace(&frame,
>> -			(unsigned long)__builtin_frame_address(0),
>> -			(unsigned long)return_address);
>> +	rc = start_backtrace(&frame,
>> +			     (unsigned long)__builtin_frame_address(0),
>> +			     (unsigned long)return_address);
>> +	if (rc == UNWIND_FINISH || rc == UNWIND_ABORT)
>> +		return NULL;
>> +
>>  	walk_stackframe(current, &frame, save_return_addr, &data);
> 
> Likewise, can we convert this to use arch_stack_walk()?
> 

OK.

Thanks.

Madhavan

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

  reply	other threads:[~2021-07-29 13:57 UTC|newest]

Thread overview: 28+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
     [not found] <3f2aab69a35c243c5e97f47c4ad84046355f5b90>
2021-06-30 22:33 ` [RFC PATCH v6 0/3] arm64: Implement stack trace reliability checks madvenka
2021-06-30 22:33   ` [RFC PATCH v6 1/3] arm64: Improve the unwinder return value madvenka
2021-07-28 16:56     ` Mark Rutland
2021-07-29 13:54       ` Madhavan T. Venkataraman [this message]
2021-06-30 22:33   ` [RFC PATCH v6 2/3] arm64: Introduce stack trace reliability checks in the unwinder madvenka
2021-06-30 22:33   ` [RFC PATCH v6 3/3] arm64: Create a list of SYM_CODE functions, check return PC against list madvenka
2021-07-28 17:25     ` Mark Rutland
2021-07-29 14:06       ` Madhavan T. Venkataraman
2021-07-29 14:52         ` Mark Brown
2021-07-29 17:07           ` Madhavan T. Venkataraman
2021-07-29 15:48         ` Mark Rutland
2021-07-29 16:27           ` Mark Brown
2021-07-29 17:09           ` Madhavan T. Venkataraman
2021-07-26 13:49   ` [RFC PATCH v6 0/3] arm64: Implement stack trace reliability checks Madhavan T. Venkataraman
2021-08-12 13:24 ` [RFC PATCH v7 0/4] arm64: Reorganize the unwinder and implement " madvenka
2021-08-12 13:24   ` [RFC PATCH v7 1/4] arm64: Make all stack walking functions use arch_stack_walk() madvenka
2021-08-12 15:23     ` Mark Brown
2021-08-12 16:30       ` Madhavan T. Venkataraman
2021-08-12 13:24   ` [RFC PATCH v7 2/4] arm64: Reorganize the unwinder code for better consistency and maintenance madvenka
2021-08-12 13:24   ` [RFC PATCH v7 3/4] arm64: Introduce stack trace reliability checks in the unwinder madvenka
2021-08-12 13:24   ` [RFC PATCH v7 4/4] arm64: Create a list of SYM_CODE functions, check return PC against list madvenka
2021-08-12 18:31   ` [RFC PATCH v7 0/4] arm64: Reorganize the unwinder and implement stack trace reliability checks Madhavan T. Venkataraman
2021-08-12 18:45     ` Madhavan T. Venkataraman
2021-08-12 18:35 ` madvenka
2021-08-12 18:35   ` [RFC PATCH v7 1/4] arm64: Make all stack walking functions use arch_stack_walk() madvenka
2021-08-12 18:35   ` [RFC PATCH v7 2/4] arm64: Reorganize the unwinder code for better consistency and maintenance madvenka
2021-08-12 18:35   ` [RFC PATCH v7 3/4] arm64: Introduce stack trace reliability checks in the unwinder madvenka
2021-08-12 18:35   ` [RFC PATCH v7 4/4] arm64: Create a list of SYM_CODE functions, check return PC against list madvenka

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=52686cb6-573c-03ca-06c2-67ae07c91243@linux.microsoft.com \
    --to=madvenka@linux.microsoft.com \
    --cc=ardb@kernel.org \
    --cc=broonie@kernel.org \
    --cc=catalin.marinas@arm.com \
    --cc=jmorris@namei.org \
    --cc=jpoimboe@redhat.com \
    --cc=jthierry@redhat.com \
    --cc=linux-arm-kernel@lists.infradead.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=live-patching@vger.kernel.org \
    --cc=mark.rutland@arm.com \
    --cc=nobuta.keiya@fujitsu.com \
    --cc=pasha.tatashin@soleen.com \
    --cc=sjitindarsingh@gmail.com \
    --cc=will@kernel.org \
    --subject='Re: [RFC PATCH v6 1/3] arm64: Improve the unwinder return value' \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).