[Crash-utility] Crash-utility Digest, Vol 178, Issue 21

[lijiang]

在 2020年08月11日 16:23, Mathias Krause 写道:
> Hi Lianbo,
> 
> Am 11.08.20 um 04:08 schrieb lijiang:
>> 在 2020年07月31日 00:00, crash-utility-request at redhat.com 写道:
>>> Message: 1
>>> Date: Thu, 30 Jul 2020 15:34:59 +0200
>>> From: Mathias Krause <minipli at grsecurity.net>
>>> To: crash-utility at redhat.com
>>> Subject: [Crash-utility] [PATCH RESEND] Basic support for PaX's split
>>> 	module	layout
>>> Message-ID: <20200730133459.7868-1-minipli at grsecurity.net>
>>> Content-Type: text/plain; charset=US-ASCII
>>>
>>> PaX and grsecurity kernels split module memory into dedicated r/x and
>>> r/w mappings using '*_rw' and '*_rx' named member variables in 'struct
>>> module'. To add basic support for such kernels detect the split layout
>>> by testing for the corresponding structure members and use these
>>> instead.
>>>
>>> So far we limit ourself to only track module code mappings for such
>>> kernels as adding support for separate data mappings violates lots of
>>> invariants in the rest of our code base, thereby would require a major
>>> rework. However, with that patch applied, module code references can be
>>> resolved in backtraces, memory and code dumps, which makes it already
>>> very useful for analyzing such kernels.
>>>
>>> Signed-off-by: Mathias Krause <minipli at grsecurity.net>
>>> ---
>>> Resend as the original posting got stuck in the mail queue.
>>>
>>>  defs.h    | 13 +++++++++++
>>>  kernel.c  | 68 +++++++++++++++++++++++++++++++++++++++++++++++++++----
>>>  symbols.c | 20 ++++++++--------
>>>  3 files changed, 86 insertions(+), 15 deletions(-)
>>>
>>> diff --git a/defs.h b/defs.h
>>> index d7adb23b86d5..160974ed554a 100644
>>> --- a/defs.h
>>> +++ b/defs.h
>>> @@ -654,12 +654,15 @@ struct new_utsname {
>>>  #define TIMER_BASES                (0x20ULL)
>>>  #define IRQ_DESC_TREE_RADIX        (0x40ULL)
>>>  #define IRQ_DESC_TREE_XARRAY       (0x80ULL)
>>> +#define KMOD_PAX                  (0x100ULL)
>>>  
>>>  #define XEN()       (kt->flags & ARCH_XEN)
>>>  #define OPENVZ()    (kt->flags & ARCH_OPENVZ)
>>>  #define PVOPS()     (kt->flags & ARCH_PVOPS)
>>>  #define PVOPS_XEN() (kt->flags & ARCH_PVOPS_XEN)
>>>  
>>> +#define PAX_MODULE_SPLIT() (kt->flags2 & KMOD_PAX)
>>> +
>>>  #define XEN_MACHINE_TO_MFN(m)    ((ulonglong)(m) >> PAGESHIFT())
>>>  #define XEN_PFN_TO_PSEUDO(p)     ((ulonglong)(p) << PAGESHIFT())
>>>  
>>> @@ -1346,7 +1349,11 @@ struct offset_table {                    /* stash of commonly-used offsets */
>>>  	long module_gpl_syms;
>>>  	long module_num_gpl_syms;
>>>  	long module_module_core;
>>> +	long module_module_core_rw;
>>> +	long module_module_core_rx;
>>>  	long module_core_size;
>>> +	long module_core_size_rw;
>>> +	long module_core_size_rx;
>>>  	long module_core_text_size;
>>>  	long module_num_symtab;
>>>  	long module_symtab;
>>> @@ -1776,6 +1783,8 @@ struct offset_table {                    /* stash of commonly-used offsets */
>>>  	long mm_struct_rss_stat;
>>>  	long mm_rss_stat_count;
>>>  	long module_module_init;
>>> +	long module_module_init_rw;
>>> +	long module_module_init_rx;
>>>  	long module_init_text_size;
>>>  	long cpu_context_save_fp;
>>>  	long cpu_context_save_sp;
>>> @@ -1793,6 +1802,8 @@ struct offset_table {                    /* stash of commonly-used offsets */
>>>  	long unwind_idx_insn;
>>>  	long signal_struct_nr_threads;
>>>  	long module_init_size;
>>> +	long module_init_size_rw;
>>> +	long module_init_size_rx;
>>>  	long module_percpu;
>>>  	long radix_tree_node_slots;
>>>  	long s390_stack_frame_back_chain;
>>> @@ -2313,6 +2324,8 @@ struct array_table {
>>>   *  in the offset table, size table or array_table.
>>>   */
>>>  #define OFFSET(X)	   (OFFSET_verify(offset_table.X, (char *)__FUNCTION__, __FILE__, __LINE__, #X))
>>> +#define MODULE_OFFSET(X,Y) (PAX_MODULE_SPLIT() ? OFFSET(Y) : OFFSET(X))
>>> +#define MODULE_OFFSET2(X,T) MODULE_OFFSET(X, X##_##T)
>> The above definition has a code style issue(space required after that ','), but we could correct
>> it when this patch is applied.
> 
> I just wanted to stick with the coding style nearby, like with
> MEMBER_OFFSET(), MEMBER_EXISTS(), MEMBER_SIZE(), etc. all have no space

That's true.

> between the arguments. But I've no strong opinion about it. Feel free to
> adapt.
> 
No worry. :-)

Thanks.
Lianbo

>>
>> Otherwise, it looks good to me.
>> Acked-by: Lianbo Jiang <lijiang at redhat.com>
> 
> Thanks for the review!
> 
> Mathias
> 
>>
>> Thanks.
>> Lianbo
>>
>>> [...]
>