Is it worth to fix the crashkernel reserved memory blocks the hotplug issue?

Is it worth to fix the crashkernel reserved memory blocks the hotplug issue?

[Pingfan Liu]

Hi,
I found in powerpc code, it is doable to reserve memory region in
movable zone, such as crashkernel does. But in x86 code, it checks the
hotpluggable attribute of memory, hence if manually specifying a
region in hotpluggable region, it will fail.
The x86 code:
/* 0 means: find the address automatically */
if (crash_base <= 0) {
/*
* Set CRASH_ADDR_LOW_MAX upper bound for crash memory,
* as old kexec-tools loads bzImage below that, unless
* "crashkernel=size[KMG],high" is specified.
*/
crash_base = memblock_find_in_range(CRASH_ALIGN,
   high ? CRASH_ADDR_HIGH_MAX
: CRASH_ADDR_LOW_MAX,
   crash_size, CRASH_ALIGN);
if (!crash_base) {
pr_info("crashkernel reservation failed - No suitable area found.\n");
return;
}

} else {
unsigned long long start;

start = memblock_find_in_range(crash_base,  --> this func will check
the hotpluggable attribute of memory and return failure if the
specifying region intersects with it.
      crash_base + crash_size,
      crash_size, 1 << 20);
if (start != crash_base) {
pr_info("crashkernel reservation failed - memory is in use.\n");
return;
}
}

Thanks,
Pingfan

Re: Is it worth to fix the crashkernel reserved memory blocks the hotplug issue?

[Baoquan He]

On 12/10/18 at 12:08pm, Pingfan Liu wrote:
> Hi,
> I found in powerpc code, it is doable to reserve memory region in
> movable zone, such as crashkernel does. But in x86 code, it checks the
> hotpluggable attribute of memory, hence if manually specifying a
> region in hotpluggable region, it will fail.

Yes, it is a problem. In x86, for crashkernel=xx@yy case to specify base
address of crashkernel reservation, it will find the region firstly, if
found region's base is not equal to specified base 'yy', means that
region has been occupied. The reservation will fail. And memblock
finding will only iterate unhotpluggable area, this will avoid reserving
crashkernel memory on hotpluggable region. 

In my opinion, it's worth fixing.

Thanks
Baoquan


> The x86 code:
> /* 0 means: find the address automatically */
> if (crash_base <= 0) {
> /*
> * Set CRASH_ADDR_LOW_MAX upper bound for crash memory,
> * as old kexec-tools loads bzImage below that, unless
> * "crashkernel=size[KMG],high" is specified.
> */
> crash_base = memblock_find_in_range(CRASH_ALIGN,
>    high ? CRASH_ADDR_HIGH_MAX
> : CRASH_ADDR_LOW_MAX,
>    crash_size, CRASH_ALIGN);
> if (!crash_base) {
> pr_info("crashkernel reservation failed - No suitable area found.\n");
> return;
> }
> 
> } else {
> unsigned long long start;
> 
> start = memblock_find_in_range(crash_base,  --> this func will check
> the hotpluggable attribute of memory and return failure if the
> specifying region intersects with it.
>       crash_base + crash_size,
>       crash_size, 1 << 20);
> if (start != crash_base) {
> pr_info("crashkernel reservation failed - memory is in use.\n");
> return;
> }
> }
> 
> Thanks,
> Pingfan