Re: [PATCH v2 5/7] x86/mm, tracing: Fix CR2 corruption

[Eiichi Tsukata]

On 2019/07/07 7:27, Steven Rostedt wrote:

> 
> We also have to deal with reading vmalloc'd data as that can fault too.
> The perf ring buffer IIUC is vmalloc, so if perf records in one of
> these locations, then the reading of the vmalloc area has a potential
> to fault corrupting the CR2 register as well. Or have we changed
> vmalloc to no longer fault?
> 
> -- Steve
> 

It seems that perf ring buffer does not normally use vmalloc.
It depends on CONFIG_PERF_USE_VMALLOC introduced by the following commit:

commit 906010b2134e14a2e377decbadd357b3d0ab9c6a
Author: Peter Zijlstra <a.p.zijlstra@xxxxxxxxx>
Date:   Mon Sep 21 16:08:49 2009 +0200

    perf_event: Provide vmalloc() based mmap() backing
    
    Some architectures such as Sparc, ARM and MIPS (basically
    everything with flush_dcache_page()) need to deal with dcache
    aliases by carefully placing pages in both kernel and user maps.
    
    These architectures typically have to use vmalloc_user() for this.
    
    However, on other architectures, vmalloc() is not needed and has
    the downsides of being more restricted and slower than regular
    allocations.
    
    Signed-off-by: Peter Zijlstra <a.p.zijlstra@xxxxxxxxx>
    Acked-by: David Miller <davem@xxxxxxxxxxxxx>
    Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
    Cc: Jens Axboe <jens.axboe@xxxxxxxxxx>
    Cc: Paul Mackerras <paulus@xxxxxxxxx>
    LKML-Reference: <1254830228.21044.272.camel@laptop>
    Signed-off-by: Ingo Molnar <mingo@xxxxxxx>