* Re: Any tracing mechanism can track the executed instructions of a user process in the kernel?
@ 2021-10-18 8:41 ` Dongliang Mu
0 siblings, 0 replies; 18+ messages in thread
From: Dongliang Mu @ 2021-10-18 8:41 UTC (permalink / raw)
To: FMDF; +Cc: Greg KH, Pavel Skripkin, linux-kernel, Dan Carpenter, kernelnewbies
On Mon, Oct 18, 2021 at 4:07 PM FMDF <fmdefrancesco@gmail.com> wrote:
>
> On Mon, 18 Oct 2021, 04:46 Dongliang Mu, <mudongliangabcd@gmail.com> wrote:
>>
>> Hi all,
>>
>> I am writing to kindly ask one question: is there any tracing
>> mechanism in Linux kernel that can trace all the executed instructions
>> of a user process? If this user process is run on different
>> processors, traces of this process on different processors should be
>> also recorded.
>
>
> You've not explained in detail what is the purpose of the tracing that you want to do. Missing this information I can only provide you a list of links to various tools and methods. Take a look by yourself and try to figure out what is better suited for your needs...
I want to log all the executed instructions of a user process (e.g.,
poc.c in syzkaller) in the kernel mode and then would like to leverage
backward analysis to capture the root cause of kernel panic/crash.
Therefore, I need the instruction-level tracing mechanisms or tools.
>
> https://lwn.net/Kernel/Index/#Development_tools-Kernel_tracing
> trace-cmd: https://trace-cmd.org
Ftrace is working at the function level. Therefore, ftrace and tracers
based on ftrace are not suitable for my scenario.
> perf ftrace: <Linux source>/tools/perf/Documentation/perf-trace.txt
> perf-tools: https://github.com/brendangregg/perf-tools
I am not familiar with the perf tool and cannot verify if it works for
my scenario.
>
> In particular, the "function_graph" tracer of ftrace seems to be suited for your purposes:
>
> https://man7.org/linux/man-pages/man1/trace-cmd.1.html
> https://man7.org/linux/man-pages/man1/trace-cmd-record.1.html
> https://man7.org/linux/man-pages/man1/trace-cmd-report.1.html
>
> Obviously, gdb (and friends) is still invaluable (but it may be difficult to use, depending on your knowledge and experience, so use it if you really know that you need it):
>
> https://www.kernel.org/doc/html/latest/dev-tools/gdb-kernel-debugging.html
> https://www.kernel.org/doc/html/latest/dev-tools/kgdb.html
>
> Regards,
>
> Fabio M. De Francesco
>
>>
>> Any comment is welcome.
>>
>> --
>> My best regards to you.
>>
>> No System Is Safe!
>> Dongliang Mu
>>
>> _______________________________________________
>> Kernelnewbies mailing list
>> Kernelnewbies@kernelnewbies.org
>> https://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
_______________________________________________
Kernelnewbies mailing list
Kernelnewbies@kernelnewbies.org
https://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: Any tracing mechanism can track the executed instructions of a user process in the kernel?
2021-10-18 8:41 ` Dongliang Mu
@ 2021-10-18 8:48 ` Greg KH
-1 siblings, 0 replies; 18+ messages in thread
From: Greg KH @ 2021-10-18 8:48 UTC (permalink / raw)
To: Dongliang Mu
Cc: FMDF, Pavel Skripkin, linux-kernel, Dan Carpenter, kernelnewbies
On Mon, Oct 18, 2021 at 04:41:14PM +0800, Dongliang Mu wrote:
> On Mon, Oct 18, 2021 at 4:07 PM FMDF <fmdefrancesco@gmail.com> wrote:
> >
> > On Mon, 18 Oct 2021, 04:46 Dongliang Mu, <mudongliangabcd@gmail.com> wrote:
> >>
> >> Hi all,
> >>
> >> I am writing to kindly ask one question: is there any tracing
> >> mechanism in Linux kernel that can trace all the executed instructions
> >> of a user process? If this user process is run on different
> >> processors, traces of this process on different processors should be
> >> also recorded.
> >
> >
> > You've not explained in detail what is the purpose of the tracing that you want to do. Missing this information I can only provide you a list of links to various tools and methods. Take a look by yourself and try to figure out what is better suited for your needs...
>
> I want to log all the executed instructions of a user process (e.g.,
> poc.c in syzkaller) in the kernel mode and then would like to leverage
> backward analysis to capture the root cause of kernel panic/crash.
>
> Therefore, I need the instruction-level tracing mechanisms or tools.
Then use a userspace debugger like gdb, that is what they are designed
for.
good luck!
greg k-h
^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: Any tracing mechanism can track the executed instructions of a user process in the kernel?
@ 2021-10-18 8:48 ` Greg KH
0 siblings, 0 replies; 18+ messages in thread
From: Greg KH @ 2021-10-18 8:48 UTC (permalink / raw)
To: Dongliang Mu
Cc: Pavel Skripkin, FMDF, linux-kernel, Dan Carpenter, kernelnewbies
On Mon, Oct 18, 2021 at 04:41:14PM +0800, Dongliang Mu wrote:
> On Mon, Oct 18, 2021 at 4:07 PM FMDF <fmdefrancesco@gmail.com> wrote:
> >
> > On Mon, 18 Oct 2021, 04:46 Dongliang Mu, <mudongliangabcd@gmail.com> wrote:
> >>
> >> Hi all,
> >>
> >> I am writing to kindly ask one question: is there any tracing
> >> mechanism in Linux kernel that can trace all the executed instructions
> >> of a user process? If this user process is run on different
> >> processors, traces of this process on different processors should be
> >> also recorded.
> >
> >
> > You've not explained in detail what is the purpose of the tracing that you want to do. Missing this information I can only provide you a list of links to various tools and methods. Take a look by yourself and try to figure out what is better suited for your needs...
>
> I want to log all the executed instructions of a user process (e.g.,
> poc.c in syzkaller) in the kernel mode and then would like to leverage
> backward analysis to capture the root cause of kernel panic/crash.
>
> Therefore, I need the instruction-level tracing mechanisms or tools.
Then use a userspace debugger like gdb, that is what they are designed
for.
good luck!
greg k-h
_______________________________________________
Kernelnewbies mailing list
Kernelnewbies@kernelnewbies.org
https://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: Any tracing mechanism can track the executed instructions of a user process in the kernel?
2021-10-18 8:41 ` Dongliang Mu
@ 2021-10-18 9:17 ` Dongliang Mu
-1 siblings, 0 replies; 18+ messages in thread
From: Dongliang Mu @ 2021-10-18 9:17 UTC (permalink / raw)
To: FMDF
Cc: kernelnewbies, linux-kernel, Greg KH, Dan Carpenter,
Pavel Skripkin, Brendan Gregg
+Brendan Gregg
On Mon, Oct 18, 2021 at 4:41 PM Dongliang Mu <mudongliangabcd@gmail.com> wrote:
>
> On Mon, Oct 18, 2021 at 4:07 PM FMDF <fmdefrancesco@gmail.com> wrote:
> >
> > On Mon, 18 Oct 2021, 04:46 Dongliang Mu, <mudongliangabcd@gmail.com> wrote:
> >>
> >> Hi all,
> >>
> >> I am writing to kindly ask one question: is there any tracing
> >> mechanism in Linux kernel that can trace all the executed instructions
> >> of a user process? If this user process is run on different
> >> processors, traces of this process on different processors should be
> >> also recorded.
> >
> >
> > You've not explained in detail what is the purpose of the tracing that you want to do. Missing this information I can only provide you a list of links to various tools and methods. Take a look by yourself and try to figure out what is better suited for your needs...
>
> I want to log all the executed instructions of a user process (e.g.,
> poc.c in syzkaller) in the kernel mode and then would like to leverage
> backward analysis to capture the root cause of kernel panic/crash.
>
> Therefore, I need the instruction-level tracing mechanisms or tools.
>
> >
> > https://lwn.net/Kernel/Index/#Development_tools-Kernel_tracing
> > trace-cmd: https://trace-cmd.org
>
> Ftrace is working at the function level. Therefore, ftrace and tracers
> based on ftrace are not suitable for my scenario.
>
> > perf ftrace: <Linux source>/tools/perf/Documentation/perf-trace.txt
> > perf-tools: https://github.com/brendangregg/perf-tools
>
> I am not familiar with the perf tool and cannot verify if it works for
> my scenario.
>
> >
> > In particular, the "function_graph" tracer of ftrace seems to be suited for your purposes:
> >
> > https://man7.org/linux/man-pages/man1/trace-cmd.1.html
> > https://man7.org/linux/man-pages/man1/trace-cmd-record.1.html
> > https://man7.org/linux/man-pages/man1/trace-cmd-report.1.html
> >
> > Obviously, gdb (and friends) is still invaluable (but it may be difficult to use, depending on your knowledge and experience, so use it if you really know that you need it):
> >
> > https://www.kernel.org/doc/html/latest/dev-tools/gdb-kernel-debugging.html
> > https://www.kernel.org/doc/html/latest/dev-tools/kgdb.html
> >
> > Regards,
> >
> > Fabio M. De Francesco
> >
> >>
> >> Any comment is welcome.
> >>
> >> --
> >> My best regards to you.
> >>
> >> No System Is Safe!
> >> Dongliang Mu
> >>
> >> _______________________________________________
> >> Kernelnewbies mailing list
> >> Kernelnewbies@kernelnewbies.org
> >> https://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: Any tracing mechanism can track the executed instructions of a user process in the kernel?
@ 2021-10-18 9:17 ` Dongliang Mu
0 siblings, 0 replies; 18+ messages in thread
From: Dongliang Mu @ 2021-10-18 9:17 UTC (permalink / raw)
To: FMDF
Cc: kernelnewbies, Greg KH, linux-kernel, Pavel Skripkin,
Brendan Gregg, Dan Carpenter
+Brendan Gregg
On Mon, Oct 18, 2021 at 4:41 PM Dongliang Mu <mudongliangabcd@gmail.com> wrote:
>
> On Mon, Oct 18, 2021 at 4:07 PM FMDF <fmdefrancesco@gmail.com> wrote:
> >
> > On Mon, 18 Oct 2021, 04:46 Dongliang Mu, <mudongliangabcd@gmail.com> wrote:
> >>
> >> Hi all,
> >>
> >> I am writing to kindly ask one question: is there any tracing
> >> mechanism in Linux kernel that can trace all the executed instructions
> >> of a user process? If this user process is run on different
> >> processors, traces of this process on different processors should be
> >> also recorded.
> >
> >
> > You've not explained in detail what is the purpose of the tracing that you want to do. Missing this information I can only provide you a list of links to various tools and methods. Take a look by yourself and try to figure out what is better suited for your needs...
>
> I want to log all the executed instructions of a user process (e.g.,
> poc.c in syzkaller) in the kernel mode and then would like to leverage
> backward analysis to capture the root cause of kernel panic/crash.
>
> Therefore, I need the instruction-level tracing mechanisms or tools.
>
> >
> > https://lwn.net/Kernel/Index/#Development_tools-Kernel_tracing
> > trace-cmd: https://trace-cmd.org
>
> Ftrace is working at the function level. Therefore, ftrace and tracers
> based on ftrace are not suitable for my scenario.
>
> > perf ftrace: <Linux source>/tools/perf/Documentation/perf-trace.txt
> > perf-tools: https://github.com/brendangregg/perf-tools
>
> I am not familiar with the perf tool and cannot verify if it works for
> my scenario.
>
> >
> > In particular, the "function_graph" tracer of ftrace seems to be suited for your purposes:
> >
> > https://man7.org/linux/man-pages/man1/trace-cmd.1.html
> > https://man7.org/linux/man-pages/man1/trace-cmd-record.1.html
> > https://man7.org/linux/man-pages/man1/trace-cmd-report.1.html
> >
> > Obviously, gdb (and friends) is still invaluable (but it may be difficult to use, depending on your knowledge and experience, so use it if you really know that you need it):
> >
> > https://www.kernel.org/doc/html/latest/dev-tools/gdb-kernel-debugging.html
> > https://www.kernel.org/doc/html/latest/dev-tools/kgdb.html
> >
> > Regards,
> >
> > Fabio M. De Francesco
> >
> >>
> >> Any comment is welcome.
> >>
> >> --
> >> My best regards to you.
> >>
> >> No System Is Safe!
> >> Dongliang Mu
> >>
> >> _______________________________________________
> >> Kernelnewbies mailing list
> >> Kernelnewbies@kernelnewbies.org
> >> https://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
_______________________________________________
Kernelnewbies mailing list
Kernelnewbies@kernelnewbies.org
https://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: Any tracing mechanism can track the executed instructions of a user process in the kernel?
2021-10-18 9:17 ` Dongliang Mu
(?)
@ 2021-10-18 9:48 ` FMDF
2021-10-18 10:12 ` FMDF
-1 siblings, 1 reply; 18+ messages in thread
From: FMDF @ 2021-10-18 9:48 UTC (permalink / raw)
To: Dongliang Mu
Cc: Greg KH, Brendan Gregg, Pavel Skripkin, Dan Carpenter, kernelnewbies
[-- Attachment #1.1: Type: text/plain, Size: 939 bytes --]
On Mon, 18 Oct 2021, 11:18 Dongliang Mu, <mudongliangabcd@gmail.com> wrote:
> +Brendan Gregg
>
Good catch!
Take a look at his "System Performance, 2nd ed., ISBN 978-0-13-682015-4.
He has an interesting blog at https://www.brendangregg.com/blog/index.html
As Greg, Pavel, and I said, use GDB if you need it. However I'd still
consider using the "func_graph" tracer of ftrace in order to know where to
attach GDB in the calls chain.
With ftrace you'll also get info about which CPU is running (you wrote that
you need to know this information).
Now I recall that, by using Perf, somehow I was able to go interactively
down to the assembly code and see where it was stuck in an endless loop.
Unfortunately at this moment I cannot remember how I did that and I cannot
try to reproduce it for you because these days I'm away from my PC (I'm
writing with a smartphone). Please try to figure it out by your own.
Regards,
Fabio
>
>
[-- Attachment #1.2: Type: text/html, Size: 1975 bytes --]
[-- Attachment #2: Type: text/plain, Size: 170 bytes --]
_______________________________________________
Kernelnewbies mailing list
Kernelnewbies@kernelnewbies.org
https://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: Any tracing mechanism can track the executed instructions of a user process in the kernel?
2021-10-18 9:48 ` FMDF
@ 2021-10-18 10:12 ` FMDF
0 siblings, 0 replies; 18+ messages in thread
From: FMDF @ 2021-10-18 10:12 UTC (permalink / raw)
To: Dongliang Mu
Cc: Greg KH, Brendan Gregg, Pavel Skripkin, Dan Carpenter, kernelnewbies
[-- Attachment #1.1: Type: text/plain, Size: 439 bytes --]
On Mon, 18 Oct 2021, 11:48 FMDF, <fmdefrancesco@gmail.com> wrote:
Now I recall that, by using Perf, somehow I was able to go interactively
> down to the assembly code and see where it was stuck in an endless loop.
>
Ah, yes. The loop was somewhere into glibc. So I'm not sure if it is
possible to see the kernel assembly with it (it seems highly not probable).
I guess that you still need GDB and friends for that.
Regards,
Fabio
>
[-- Attachment #1.2: Type: text/html, Size: 1161 bytes --]
[-- Attachment #2: Type: text/plain, Size: 170 bytes --]
_______________________________________________
Kernelnewbies mailing list
Kernelnewbies@kernelnewbies.org
https://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: Any tracing mechanism can track the executed instructions of a user process in the kernel?
2021-10-18 8:41 ` Dongliang Mu
@ 2021-10-18 15:10 ` Valdis Klētnieks
-1 siblings, 0 replies; 18+ messages in thread
From: Valdis Klētnieks @ 2021-10-18 15:10 UTC (permalink / raw)
To: Dongliang Mu
Cc: FMDF, kernelnewbies, linux-kernel, Greg KH, Dan Carpenter,
Pavel Skripkin
[-- Attachment #1: Type: text/plain, Size: 2986 bytes --]
On Mon, 18 Oct 2021 16:41:14 +0800, Dongliang Mu said:
> I want to log all the executed instructions of a user process (e.g.,
> poc.c in syzkaller) in the kernel mode and then would like to leverage
> backward analysis to capture the root cause of kernel panic/crash.
> Therefore, I need the instruction-level tracing mechanisms or tools.
Tracing just the instructions won't get you where you want to be if
you're going through this approach.
You *also* need to track all the data - the instruction path inside two
different runs of syzkaller may be essentially identical, but pass 2 different
values as the 3rd parameter of a syscall.
You may also have to deal with insane amounts of data - the actual error could
have been minutes or even hours before, or the interaction between two
different processes.
You probably want to take a *really* close look at how prof and friends avoid
infinite regress when code execution drops inside the prof code, because you're
going to hit the same issues.
Or....
You can work smarter rather than harder, and ask yourself what's the minimum
amount and type of additional information to make a significant improvement in
the debugging of system crashes.
For example, 95% of the time, you can figure out what the bug is by merely
looking at the stack traceback. For most of the rest of the cases, simply
capturing the parameter values from the syscall and the basic info for page
faults and other interrupts is probably sufficient, and you can probably
leverage the audit subsystem for most of that. It can already record syscall
parameters, while logging page faults and other interrupts can probably be done
with prof.
At that point, you don't actually *need* every instruction - only tracing
branch and call instructions is sufficient, because you already know that each
instruction between the target of a branch/call and the next branch/call will
be executed.
Similarly, the lockdep code will catch most locking issues. But it won't flag
issues with data that should be protected with a lock, but are bereft of any
locking. So ask yourself: What ways are there to analyze the code and detect
critical sections prone to race conditions? Is there a sparse-on-steroids
approach that wil do the heavy lifting for those? (Note that this isn't an easy
task for the general case, but identifying two or three specific common
patterns and finding a way to detect them may be worthwhile)
And many of the rest of crashes are timing related, and "let's trace every single
instruction" is almost guaranteed to make things slow enough to change/bypass
the timing issue.
So... What's left that would be the most helpful with the least amount of data?
Go look at some threads on linux-kernel. Look at the kernel bugs that were the
result of a Homer Simpson "D'oh!" moment. What can we do to make those
bugs less likely to make it into the code in the first place? For the more subtle
bugs, what data finally made the debugging come together?
[-- Attachment #2: Type: application/pgp-signature, Size: 494 bytes --]
^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: Any tracing mechanism can track the executed instructions of a user process in the kernel?
@ 2021-10-18 15:10 ` Valdis Klētnieks
0 siblings, 0 replies; 18+ messages in thread
From: Valdis Klētnieks @ 2021-10-18 15:10 UTC (permalink / raw)
To: Dongliang Mu
Cc: kernelnewbies, Greg KH, linux-kernel, Pavel Skripkin, FMDF,
Dan Carpenter
[-- Attachment #1.1: Type: text/plain, Size: 2986 bytes --]
On Mon, 18 Oct 2021 16:41:14 +0800, Dongliang Mu said:
> I want to log all the executed instructions of a user process (e.g.,
> poc.c in syzkaller) in the kernel mode and then would like to leverage
> backward analysis to capture the root cause of kernel panic/crash.
> Therefore, I need the instruction-level tracing mechanisms or tools.
Tracing just the instructions won't get you where you want to be if
you're going through this approach.
You *also* need to track all the data - the instruction path inside two
different runs of syzkaller may be essentially identical, but pass 2 different
values as the 3rd parameter of a syscall.
You may also have to deal with insane amounts of data - the actual error could
have been minutes or even hours before, or the interaction between two
different processes.
You probably want to take a *really* close look at how prof and friends avoid
infinite regress when code execution drops inside the prof code, because you're
going to hit the same issues.
Or....
You can work smarter rather than harder, and ask yourself what's the minimum
amount and type of additional information to make a significant improvement in
the debugging of system crashes.
For example, 95% of the time, you can figure out what the bug is by merely
looking at the stack traceback. For most of the rest of the cases, simply
capturing the parameter values from the syscall and the basic info for page
faults and other interrupts is probably sufficient, and you can probably
leverage the audit subsystem for most of that. It can already record syscall
parameters, while logging page faults and other interrupts can probably be done
with prof.
At that point, you don't actually *need* every instruction - only tracing
branch and call instructions is sufficient, because you already know that each
instruction between the target of a branch/call and the next branch/call will
be executed.
Similarly, the lockdep code will catch most locking issues. But it won't flag
issues with data that should be protected with a lock, but are bereft of any
locking. So ask yourself: What ways are there to analyze the code and detect
critical sections prone to race conditions? Is there a sparse-on-steroids
approach that wil do the heavy lifting for those? (Note that this isn't an easy
task for the general case, but identifying two or three specific common
patterns and finding a way to detect them may be worthwhile)
And many of the rest of crashes are timing related, and "let's trace every single
instruction" is almost guaranteed to make things slow enough to change/bypass
the timing issue.
So... What's left that would be the most helpful with the least amount of data?
Go look at some threads on linux-kernel. Look at the kernel bugs that were the
result of a Homer Simpson "D'oh!" moment. What can we do to make those
bugs less likely to make it into the code in the first place? For the more subtle
bugs, what data finally made the debugging come together?
[-- Attachment #1.2: Type: application/pgp-signature, Size: 494 bytes --]
[-- Attachment #2: Type: text/plain, Size: 170 bytes --]
_______________________________________________
Kernelnewbies mailing list
Kernelnewbies@kernelnewbies.org
https://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
^ permalink raw reply [flat|nested] 18+ messages in thread