Mirror of the lttng-dev mailing list
 help / color / mirror / Atom feed
* [lttng-dev] https://lists.lttng.org/pipermail/lttng-dev/2020-May/029631.html
       [not found] <CALx39aqN_YvfcDhhKgURZZU-bWZvuiZptbb-kJDYerEex9Q-7A@mail.gmail.com>
@ 2023-03-26 15:00 ` yashvardhan kukreti via lttng-dev
  2023-03-27 12:27   ` Mathieu Desnoyers via lttng-dev
  0 siblings, 1 reply; 5+ messages in thread
From: yashvardhan kukreti via lttng-dev @ 2023-03-26 15:00 UTC (permalink / raw)
  To: mathieu.desnoyers; +Cc: lttng-dev


[-- Attachment #1.1: Type: text/plain, Size: 1288 bytes --]

>
>
> Hi Mathew,
>
> I have a question about this patch for lttng-modules and the use of
> register_kprobe() to fetch the function ptr.
> The question in this regard is especially from PPC64 ELF_ABI_v1
> perspective.
>
> The functions on PPC64 are accessed via the Function descriptor while what
> register_kprobes returns is the entry point of the function.
> Hence using the return pointer tends to interpret the addr as the address
> of the function descriptor and dereferences the ppc_inst as the function
> entry point and crashes
>
> [ 4145.483594] kernel tried to execute exec-protected page
> (7c0802a6fb81ffe0) - exploit attempt? (uid: 0)
>
> here 7c0802a6 is the mfspr instruction from the code text section of the
> kallsyms_lookup_name()
>
> note for PPC_ELF_ABI_v1 the register_kprobes() searches for the dot
> variant of the symbol and only in case if cannot find the dot variant looks
> for the normal symbol.
> register_kprobe() -> kprobe_addr() -> kprobe_lookup_name() [arch variant
> replaces weak symbol]
> https://elixir.bootlin.com/linux/v5.10.174/C/ident/kprobe_lookup_name
>
> Please let me know if i make sense or that i may have missed something.
>
> I have looked at the code of 2.12.8 as well and 2.12.3 verstion of
> lttng-modules.
>
> Regards,
> Shashank
>
>

[-- Attachment #1.2: Type: text/html, Size: 1819 bytes --]

[-- Attachment #2: Type: text/plain, Size: 156 bytes --]

_______________________________________________
lttng-dev mailing list
lttng-dev@lists.lttng.org
https://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [lttng-dev] https://lists.lttng.org/pipermail/lttng-dev/2020-May/029631.html
  2023-03-26 15:00 ` [lttng-dev] https://lists.lttng.org/pipermail/lttng-dev/2020-May/029631.html yashvardhan kukreti via lttng-dev
@ 2023-03-27 12:27   ` Mathieu Desnoyers via lttng-dev
  2023-03-27 15:52     ` yashvardhan kukreti via lttng-dev
  0 siblings, 1 reply; 5+ messages in thread
From: Mathieu Desnoyers via lttng-dev @ 2023-03-27 12:27 UTC (permalink / raw)
  To: yashvardhan kukreti; +Cc: lttng-dev

On 2023-03-26 11:00, yashvardhan kukreti wrote:
> 
>     Hi Mathew,
> 
>     I have a question about this patch for lttng-modules and the use of
>     register_kprobe() to fetch the function ptr.
>     The question in this regard is especially from PPC64 ELF_ABI_v1
>     perspective.
> 
>     The functions on PPC64 are accessed via the Function descriptor
>     while what register_kprobes returns is the entry point of the function.
>     Hence using the return pointer tends to interpret the addr as the
>     address of the function descriptor and dereferences the ppc_inst as
>     the function entry point and crashes
> 
>     [ 4145.483594] kernel tried to execute exec-protected page
>     (7c0802a6fb81ffe0) - exploit attempt? (uid: 0)
>     here 7c0802a6 is the mfspr instruction from the code text section of
>     the kallsyms_lookup_name()
> 
>     note for PPC_ELF_ABI_v1 the register_kprobes() searches for the dot
>     variant of the symbol and only in case if cannot find the dot
>     variant looks for the normal symbol.
>     register_kprobe() -> kprobe_addr() -> kprobe_lookup_name() [arch
>     variant replaces weak symbol]
>     https://elixir.bootlin.com/linux/v5.10.174/C/ident/kprobe_lookup_name <https://elixir.bootlin.com/linux/v5.10.174/C/ident/kprobe_lookup_name>
> 
>     Please let me know if i make sense or that i may have missed something.
> 
>     I have looked at the code of 2.12.8 as well and 2.12.3 verstion of
>     lttng-modules.

Please have a look at commits (from stable-2.12 branch of lttng-modules):

commit 53772db24facd84f1f3ddcf21a1ef5f162608721
Author: He Zhe <zhe.he@windriver.com>
Date:   Tue Sep 27 15:59:42 2022 +0800

     wrapper: powerpc64: fix kernel crash caused by do_get_kallsyms

commit 8fe888d86ccad4226b05a536efb73d71bb091062
Author: Michael Jeanson <mjeanson@efficios.com>
Date:   Thu Nov 24 14:25:33 2022 -0500

     fix: kallsyms wrapper on ppc64el

I suspect you'll also need this change currently in review:

https://review.lttng.org/c/lttng-modules/+/9113

Please let us know if especially this last change fixes things on your side.

Thanks,

Mathieu


> 
>     Regards,
>     Shashank
> 

-- 
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com

_______________________________________________
lttng-dev mailing list
lttng-dev@lists.lttng.org
https://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [lttng-dev] https://lists.lttng.org/pipermail/lttng-dev/2020-May/029631.html
  2023-03-27 12:27   ` Mathieu Desnoyers via lttng-dev
@ 2023-03-27 15:52     ` yashvardhan kukreti via lttng-dev
  2023-03-27 18:29       ` Michael Jeanson via lttng-dev
  0 siblings, 1 reply; 5+ messages in thread
From: yashvardhan kukreti via lttng-dev @ 2023-03-27 15:52 UTC (permalink / raw)
  To: Mathieu Desnoyers; +Cc: lttng-dev


[-- Attachment #1.1: Type: text/plain, Size: 3583 bytes --]

 62 #ifdef LTTNG_CONFIG_PPC64_ELF_ABI_V2
 63         /* Substract 4 bytes to get what we originally want */
 64         addr = (unsigned long)(((char *)probe.addr) - 4);
 65 #elif defined*(LTTNG_*CONFIG_PPC64_ELF_ABI_V1)   << incorrect MACRO to
used in gerrit review. this is the correct one.
 66         /*
 67          * Build a function descriptor from the address of

/lttng-module-2.13.8/src/wrapper/kallsyms.c:72:32: error: invalid use of
undefined type 'struct func_desc'
|    72 |  kallsyms_lookup_name_func_desc.addr = (unsigned long)probe.addr;
/lttng-module-2.13.8/src/wrapper/kallsyms.c:73:32: error: invalid use of
undefined type 'struct func_desc'
|    73 |  kallsyms_lookup_name_func_desc.toc = ((struct func_desc *)
&sprint_symbol)->toc;

you either want to use func_descr_t
https://elixir.bootlin.com/linux/v5.15.104/source/arch/powerpc/include/asm/types.h#L30
or
struct ppc64_opd_entry
https://elixir.bootlin.com/linux/v5.17.15/source/arch/powerpc/include/uapi/asm/elf.h#L293


both of which are discontinued from 5.18



On Mon, Mar 27, 2023 at 5:56 PM Mathieu Desnoyers <
mathieu.desnoyers@efficios.com> wrote:

> On 2023-03-26 11:00, yashvardhan kukreti wrote:
> >
> >     Hi Mathew,
> >
> >     I have a question about this patch for lttng-modules and the use of
> >     register_kprobe() to fetch the function ptr.
> >     The question in this regard is especially from PPC64 ELF_ABI_v1
> >     perspective.
> >
> >     The functions on PPC64 are accessed via the Function descriptor
> >     while what register_kprobes returns is the entry point of the
> function.
> >     Hence using the return pointer tends to interpret the addr as the
> >     address of the function descriptor and dereferences the ppc_inst as
> >     the function entry point and crashes
> >
> >     [ 4145.483594] kernel tried to execute exec-protected page
> >     (7c0802a6fb81ffe0) - exploit attempt? (uid: 0)
> >     here 7c0802a6 is the mfspr instruction from the code text section of
> >     the kallsyms_lookup_name()
> >
> >     note for PPC_ELF_ABI_v1 the register_kprobes() searches for the dot
> >     variant of the symbol and only in case if cannot find the dot
> >     variant looks for the normal symbol.
> >     register_kprobe() -> kprobe_addr() -> kprobe_lookup_name() [arch
> >     variant replaces weak symbol]
> >
> https://elixir.bootlin.com/linux/v5.10.174/C/ident/kprobe_lookup_name <
> https://elixir.bootlin.com/linux/v5.10.174/C/ident/kprobe_lookup_name>
> >
> >     Please let me know if i make sense or that i may have missed
> something.
> >
> >     I have looked at the code of 2.12.8 as well and 2.12.3 verstion of
> >     lttng-modules.
>
> Please have a look at commits (from stable-2.12 branch of lttng-modules):
>
> commit 53772db24facd84f1f3ddcf21a1ef5f162608721
> Author: He Zhe <zhe.he@windriver.com>
> Date:   Tue Sep 27 15:59:42 2022 +0800
>
>      wrapper: powerpc64: fix kernel crash caused by do_get_kallsyms
>
> commit 8fe888d86ccad4226b05a536efb73d71bb091062
> Author: Michael Jeanson <mjeanson@efficios.com>
> Date:   Thu Nov 24 14:25:33 2022 -0500
>
>      fix: kallsyms wrapper on ppc64el
>
> I suspect you'll also need this change currently in review:
>
> https://review.lttng.org/c/lttng-modules/+/9113
>
> Please let us know if especially this last change fixes things on your
> side.
>
> Thanks,
>
> Mathieu
>
>
> >
> >     Regards,
> >     Shashank
> >
>
> --
> Mathieu Desnoyers
> EfficiOS Inc.
> https://www.efficios.com
>
>

[-- Attachment #1.2: Type: text/html, Size: 5184 bytes --]

[-- Attachment #2: Type: text/plain, Size: 156 bytes --]

_______________________________________________
lttng-dev mailing list
lttng-dev@lists.lttng.org
https://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [lttng-dev] https://lists.lttng.org/pipermail/lttng-dev/2020-May/029631.html
  2023-03-27 15:52     ` yashvardhan kukreti via lttng-dev
@ 2023-03-27 18:29       ` Michael Jeanson via lttng-dev
  2023-03-28 13:34         ` yashvardhan kukreti via lttng-dev
  0 siblings, 1 reply; 5+ messages in thread
From: Michael Jeanson via lttng-dev @ 2023-03-27 18:29 UTC (permalink / raw)
  To: yashvardhan kukreti, Mathieu Desnoyers; +Cc: lttng-dev

On 2023-03-27 11:52, yashvardhan kukreti wrote:
>   62 #ifdef LTTNG_CONFIG_PPC64_ELF_ABI_V2
>   63         /* Substract 4 bytes to get what we originally want */
>   64         addr = (unsigned long)(((char *)probe.addr) - 4);
>   65 #elif defined*(LTTNG_*CONFIG_PPC64_ELF_ABI_V1)   << incorrect MACRO to 
> used in gerrit review. this is the correct one.
>   66         /*
>   67          * Build a function descriptor from the address of
> 
> /lttng-module-2.13.8/src/wrapper/kallsyms.c:72:32: error: invalid use of 
> undefined type 'struct func_desc'
> |    72 |  kallsyms_lookup_name_func_desc.addr = (unsigned long)probe.addr;
> /lttng-module-2.13.8/src/wrapper/kallsyms.c:73:32: error: invalid use of 
> undefined type 'struct func_desc'
> |    73 |  kallsyms_lookup_name_func_desc.toc = ((struct func_desc *) 
> &sprint_symbol)->toc;
> 
> you either want to use func_descr_t
> https://elixir.bootlin.com/linux/v5.15.104/source/arch/powerpc/include/asm/types.h#L30 <https://elixir.bootlin.com/linux/v5.15.104/source/arch/powerpc/include/asm/types.h#L30>
> or
> struct ppc64_opd_entry
> https://elixir.bootlin.com/linux/v5.17.15/source/arch/powerpc/include/uapi/asm/elf.h#L293 <https://elixir.bootlin.com/linux/v5.17.15/source/arch/powerpc/include/uapi/asm/elf.h#L293>
> 
> both of which are discontinued from 5.18


Hi,

I updated the patch at https://review.lttng.org/c/lttng-modules/+/9113 to 
support kernels older than v5.18. It's only build tested on v5.17 and v5.18.

Michael

_______________________________________________
lttng-dev mailing list
lttng-dev@lists.lttng.org
https://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [lttng-dev] https://lists.lttng.org/pipermail/lttng-dev/2020-May/029631.html
  2023-03-27 18:29       ` Michael Jeanson via lttng-dev
@ 2023-03-28 13:34         ` yashvardhan kukreti via lttng-dev
  0 siblings, 0 replies; 5+ messages in thread
From: yashvardhan kukreti via lttng-dev @ 2023-03-28 13:34 UTC (permalink / raw)
  To: Michael Jeanson; +Cc: lttng-dev


[-- Attachment #1.1: Type: text/plain, Size: 1986 bytes --]

Thanks Michael, the patch is working

[  157.606699]  SK DEBUG look up symbol .kallsyms_lookup_name
[  157.613527]  SK DEBUG look up symbol get_pfnblock_flags_mask
[  157.644191] LTTng: Loaded modules v2.13.8 (Nordicité) -
devtool-base-1-g1011df74-dirty

Regards,
Shahank

On Mon, Mar 27, 2023 at 11:59 PM Michael Jeanson <mjeanson@efficios.com>
wrote:

> On 2023-03-27 11:52, yashvardhan kukreti wrote:
> >   62 #ifdef LTTNG_CONFIG_PPC64_ELF_ABI_V2
> >   63         /* Substract 4 bytes to get what we originally want */
> >   64         addr = (unsigned long)(((char *)probe.addr) - 4);
> >   65 #elif defined*(LTTNG_*CONFIG_PPC64_ELF_ABI_V1)   << incorrect MACRO
> to
> > used in gerrit review. this is the correct one.
> >   66         /*
> >   67          * Build a function descriptor from the address of
> >
> > /lttng-module-2.13.8/src/wrapper/kallsyms.c:72:32: error: invalid use of
> > undefined type 'struct func_desc'
> > |    72 |  kallsyms_lookup_name_func_desc.addr = (unsigned
> long)probe.addr;
> > /lttng-module-2.13.8/src/wrapper/kallsyms.c:73:32: error: invalid use of
> > undefined type 'struct func_desc'
> > |    73 |  kallsyms_lookup_name_func_desc.toc = ((struct func_desc *)
> > &sprint_symbol)->toc;
> >
> > you either want to use func_descr_t
> >
> https://elixir.bootlin.com/linux/v5.15.104/source/arch/powerpc/include/asm/types.h#L30
> <
> https://elixir.bootlin.com/linux/v5.15.104/source/arch/powerpc/include/asm/types.h#L30
> >
> > or
> > struct ppc64_opd_entry
> >
> https://elixir.bootlin.com/linux/v5.17.15/source/arch/powerpc/include/uapi/asm/elf.h#L293
> <
> https://elixir.bootlin.com/linux/v5.17.15/source/arch/powerpc/include/uapi/asm/elf.h#L293
> >
> >
> > both of which are discontinued from 5.18
>
>
> Hi,
>
> I updated the patch at https://review.lttng.org/c/lttng-modules/+/9113 to
> support kernels older than v5.18. It's only build tested on v5.17 and
> v5.18.
>
> Michael
>
>

[-- Attachment #1.2: Type: text/html, Size: 3158 bytes --]

[-- Attachment #2: Type: text/plain, Size: 156 bytes --]

_______________________________________________
lttng-dev mailing list
lttng-dev@lists.lttng.org
https://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev

^ permalink raw reply	[flat|nested] 5+ messages in thread

end of thread, other threads:[~2023-03-28 13:35 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
     [not found] <CALx39aqN_YvfcDhhKgURZZU-bWZvuiZptbb-kJDYerEex9Q-7A@mail.gmail.com>
2023-03-26 15:00 ` [lttng-dev] https://lists.lttng.org/pipermail/lttng-dev/2020-May/029631.html yashvardhan kukreti via lttng-dev
2023-03-27 12:27   ` Mathieu Desnoyers via lttng-dev
2023-03-27 15:52     ` yashvardhan kukreti via lttng-dev
2023-03-27 18:29       ` Michael Jeanson via lttng-dev
2023-03-28 13:34         ` yashvardhan kukreti via lttng-dev

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox