[PATCH v17 03/10] x86: kdump: use macro CRASH_ADDR_LOW_MAX in functions reserve_crashkernel()

Leizhen (ThunderTown) thunder.leizhen at huawei.com
Thu Dec 16 18:51:04 PST 2021



On 2021/12/16 22:48, Borislav Petkov wrote:
> On Thu, Dec 16, 2021 at 08:08:30PM +0800, Leizhen (ThunderTown) wrote:
>> If the memory of 'crash_base' is successfully allocated at (1), because the last
>> parameter CRASH_ADDR_LOW_MAX is the upper bound, so we can sure that
>> "crash_base < CRASH_ADDR_LOW_MAX". So that, reserve_crashkernel_low() will not be
>> invoked at (3). That's why I said (1ULL << 32) is inaccurate and enlarge the CRASH_ADDR_LOW
>> upper limit.
> 
> No, this is actually wrong - that check *must* be 4G. See:
> 
>   eb6db83d1059 ("x86/setup: Do not reserve crashkernel high memory if low reservation failed")
> 
> It is even documented:
> 
>         crashkernel=size[KMG],low
>                         [KNL, X86-64] range under 4G. When crashkernel=X,high

[KNL, X86-64], This doc is for X86-64, not for X86-32

>                         is passed, kernel could allocate physical memory region
>                         above 4G, that cause second kernel crash on system
>                         that require some amount of low memory, e.g. swiotlb
>                         requires at least 64M+32K low memory, also enough extra
>                         low memory is needed to make sure DMA buffers for 32-bit
>                         devices won't run out.

vi arch/x86/kernel/setup.c +398

/*
 * Keep the crash kernel below this limit.
 *
 * Earlier 32-bits kernels would limit the kernel to the low 512 MB range
 * due to mapping restrictions.

If there is no such restriction, we can make CRASH_ADDR_LOW_MAX equal to (1ULL << 32) minus 1 on X86_32.

> 
> so you need to do a low allocation for DMA *when* the reserved memory is
> above 4G. *NOT* above 512M. But that works due to the obscure situation,
> as Baoquan stated, that reserve_crashkernel_low() returns 0 on 32-bit.
> 
> So all this is telling us is that that function needs serious cleanup.
> 



More information about the kexec mailing list