netdev - Re: [PATCH net] xdp: xdp_mem_allocator can be NULL in trace_mem

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: <d6621a7b-12ad-1e3d-848f-fff576be2dfd@redhat.com>
Date:   Thu, 3 Mar 2022 18:31:11 +0100
From:   Jesper Dangaard Brouer <jbrouer@...hat.com>
To:     Sebastian Andrzej Siewior <bigeasy@...utronix.de>,
        Toke Høiland-Jørgensen <toke@...hat.com>
Cc:     brouer@...hat.com, netdev@...r.kernel.org, bpf@...r.kernel.org,
        Alexei Starovoitov <ast@...nel.org>,
        Andrii Nakryiko <andrii@...nel.org>,
        Daniel Borkmann <daniel@...earbox.net>,
        "David S. Miller" <davem@...emloft.net>,
        Ingo Molnar <mingo@...hat.com>,
        Jakub Kicinski <kuba@...nel.org>,
        Jesper Dangaard Brouer <hawk@...nel.org>,
        John Fastabend <john.fastabend@...il.com>,
        KP Singh <kpsingh@...nel.org>, Martin KaFai Lau <kafai@...com>,
        Song Liu <songliubraving@...com>,
        Steven Rostedt <rostedt@...dmis.org>,
        Thomas Gleixner <tglx@...utronix.de>,
        Yonghong Song <yhs@...com>
Subject: Re: [PATCH net] xdp: xdp_mem_allocator can be NULL in
 trace_mem_connect().



On 03/03/2022 15.12, Sebastian Andrzej Siewior wrote:
> On 2022-03-03 14:59:47 [+0100], Toke Høiland-Jørgensen wrote:
>> Sebastian Andrzej Siewior <bigeasy@...utronix.de> writes:
>>
>>> Since the commit mentioned below __xdp_reg_mem_model() can return a NULL
>>> pointer. This pointer is dereferenced in trace_mem_connect() which leads
>>> to segfault. It can be reproduced with enabled trace events during ifup.
>>>
>>> Only assign the arguments in the trace-event macro if `xa' is set.
>>> Otherwise set the parameters to 0.
>>>
>>> Fixes: 4a48ef70b93b8 ("xdp: Allow registering memory model without rxq reference")
>>> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@...utronix.de>
>>
>> Hmm, so before the commit you mention, the tracepoint wasn't triggered
>> at all in the code path that now sets xdp_alloc is NULL. So I'm
>> wondering if we should just do the same here? Is the trace event useful
>> in all cases?
> 
> Correct. It says:
> |              ip-1230    [003] .....     3.053473: mem_connect: mem_id=0 mem_type=PAGE_SHARED allocator=0000000000000000 ifindex=2
> 
>> Alternatively, if we keep it, I think the mem.id and mem.type should be
>> available from rxq->mem, right?
> 
> Yes, if these are the same things. In my case they are also 0:
> 
> |              ip-1245    [000] .....     3.045684: mem_connect: mem_id=0 mem_type=PAGE_SHARED allocator=0000000000000000 ifindex=2
> |        ifconfig-1332    [003] .....    21.030879: mem_connect: mem_id=0 mem_type=PAGE_SHARED allocator=0000000000000000 ifindex=3
> 
> So depends on what makes sense that tp can be skipped for xa == NULL or
> remain with
>                 __entry->mem_id         = rxq->mem.id;
>                 __entry->mem_type       = rxq->mem.type;
> 	       __entry->allocator      = xa ? xa->allocator : NULL;
> 
> if it makes sense.
> 

I have two bpftrace scripts [1] and [2] that use this tracepoint.
It is scripts to help driver developers detect memory leaks when using 
page_pool from their drivers.

I'm a little pressured on time, so I've not evaluated if your change 
makes sense.
In the scripts I do use both mem.id and mem.type.


[1] 
https://github.com/xdp-project/xdp-project/blob/master/areas/mem/bpftrace/xdp_mem_track01.bt

[2] 
https://github.com/xdp-project/xdp-project/blob/master/areas/mem/bpftrace/xdp_mem_track02.bt

--Jesper