linux-kernel - Re: [PATCH 1/2] x86/stacktrace: do not fail when regs on stack for ORC

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <16f714bb-1a32-9203-39d5-aab927fba6c9@suse.cz>
Date:   Fri, 11 May 2018 09:01:30 +0200
From:   Jiri Slaby <jslaby@...e.cz>
To:     Josh Poimboeuf <jpoimboe@...hat.com>
Cc:     mingo@...hat.com, linux-kernel@...r.kernel.org,
        Thomas Gleixner <tglx@...utronix.de>,
        "H. Peter Anvin" <hpa@...or.com>, x86@...nel.org
Subject: Re: [PATCH 1/2] x86/stacktrace: do not fail when regs on stack for
 ORC

On 05/10/2018, 02:49 PM, Josh Poimboeuf wrote:
> On Thu, May 10, 2018 at 02:33:03PM +0200, Jiri Slaby wrote:
>> On 04/19/2018, 03:42 PM, Josh Poimboeuf wrote:
>>> On Mon, Apr 16, 2018 at 05:16:53PM -0500, Josh Poimboeuf wrote:
>>>> On Wed, Dec 20, 2017 at 08:07:17PM +0100, Jiri Slaby wrote:
>>>>> On 12/20/2017, 06:45 PM, Josh Poimboeuf wrote:
>>>>>> It might not be until 2018 though.  But in the meantime you can go ahead
>>>>>> and update your patches accordingly and then we can combine them for
>>>>>> testing next year.
>>>>>
>>>>> I already did ;). So when you have that ready, I will send it on top
>>>>> right after.
>>>>
>>>> Sorry for the delay...  Here's a (lightly tested) patch.  Can you test
>>>> it with the latest version of your patches?
>>>
>>> This one actually compiles:
>>>
>>> From: Josh Poimboeuf <jpoimboe@...hat.com>
>>> Subject: [PATCH] x86/unwind/orc: Detect the end of the stack
>>
>> With this patch applied, livepatching never completes. Kthreads are in
>> the unpatched state forever. I have no details yet.
> 
> Hm, I thought I had tested that, but now I'm not sure.  Let me try it
> again.

We need to propagate end from hints to orcs. This works for me™:

commit ad05e939b5809db104528731ed0147ad59466db5
Author: Jiri Slaby <jslaby@...e.cz>
Date:   Fri May 11 09:00:12 2018 +0200

    objtool: take end hint into account

    Signed-off-by: Jiri Slaby <jslaby@...e.cz>

diff --git a/tools/objtool/check.c b/tools/objtool/check.c
index 5409f6f..fef15ce 100644
--- a/tools/objtool/check.c
+++ b/tools/objtool/check.c
@@ -1109,6 +1109,7 @@ static int read_unwind_hints(struct objtool_file
*file)

                cfa->offset = hint->sp_offset;
                insn->state.type = hint->type;
+               insn->state.end = hint->end;
        }

        return 0;
diff --git a/tools/objtool/check.h b/tools/objtool/check.h
index c6b68fc..000ecf3 100644
--- a/tools/objtool/check.h
+++ b/tools/objtool/check.h
@@ -30,6 +30,7 @@ struct insn_state {
        struct cfi_reg regs[CFI_NUM_REGS];
        int stack_size;
        unsigned char type;
+       unsigned char end;
        bool bp_scratch;
        bool drap;
        int drap_reg, drap_offset;
diff --git a/tools/objtool/orc_dump.c b/tools/objtool/orc_dump.c
index c334382..a9bf1861 100644
--- a/tools/objtool/orc_dump.c
+++ b/tools/objtool/orc_dump.c
@@ -203,7 +203,9 @@ int orc_dump(const char *_objname)

                print_reg(orc[i].bp_reg, orc[i].bp_offset);

-               printf(" type:%s\n", orc_type_name(orc[i].type));
+               printf(" type:%s", orc_type_name(orc[i].type));
+
+               printf(" end:%u\n", orc[i].end);
        }

        elf_end(elf);
diff --git a/tools/objtool/orc_gen.c b/tools/objtool/orc_gen.c
index 18384d9..f413cc2 100644
--- a/tools/objtool/orc_gen.c
+++ b/tools/objtool/orc_gen.c
@@ -31,6 +31,9 @@ int create_orc(struct objtool_file *file)
                struct cfi_reg *cfa = &insn->state.cfa;
                struct cfi_reg *bp = &insn->state.regs[CFI_BP];

+               //if (insn->hint)
+                       orc->end = insn->state.end;
+
                if (cfa->base == CFI_UNDEFINED) {
                        orc->sp_reg = ORC_REG_UNDEFINED;
                        continue;

thanks,
-- 
js
suse labs