lkml.org 
[lkml]   [2015]   [May]   [18]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [v2 0/5] arm64: add kdump support
Mark,

On 05/12/2015 02:43 PM, Dave Young wrote:
> On 05/11/15 at 03:16pm, AKASHI Takahiro wrote:
>> Hi
>>
>> Sorry for late response. I was on vacation.
>>
>> On 04/24/2015 06:53 PM, Mark Rutland wrote:
>>> Hi,
>>>
>>> On Fri, Apr 24, 2015 at 08:53:03AM +0100, AKASHI Takahiro wrote:
>>>> This patch set enables kdump (crash dump kernel) support on arm64 on top of
>>>> Geoff's kexec patchset.
>>>>
>>>> In this version, there are some arm64-specific usage/constraints:
>>>> 1) "mem=" boot parameter must be specified on crash dump kernel
>>>> if the system starts on uefi.
>>>
>>> This sounds very painful. Why is this the case, and how do x86 and/or
>>> ia64 get around that?
>>
>> As Dave (Young) said, x86 uses "memmap=XX" kernel commandline parameters
>> to specify usable memory for crash dump kernel.
>
> Originally x86 use memmap=exactmap memmap=XX to specify each section of
> memories for 2nd kernel. But later because a lot of reserved type ranges
> need to be passed ie. for pci mmconfig, and kernel cmdline buffer is
> limited so kexec-tools later switch to passing these in x86 boot params as
> E820 memory ranges directly.
>
>> On my arm64 implementation, "linux,usable-memory" property is added
>> to device tree blob by kexec-tools for this purpose.
>> This is because, when I first implemented kdump on arm64, ppc is the only
>> architecture that supports kdump AND utilizes device trees.
>> Since kexec-tools as well as the kernel already has this framework,
>> I believed that device-tree approach was smarter than a commandline
>> parameter.
>>
>> However, uefi-based kernel ignores all the memory-related properties
>> in a device tree and so this "mem=" workaround
> Kdump kernel reuses the memmap info getting from firmware during 1st kernel
> boot, I do not think the memmap info can be cooked for crash kernel usable
> memory. But it might be a better way to use a special fdt node for crash
> kernel memory even for UEFI..

Do you still prefer "memmap=" approach?

Just FYI,

x86 arm arm64 powerpc{,64}

(a) cmdline(crashkernel=) cmdline cmdline cmdline
(b) iomem iomem iomem dev tree
(c) cmdline(memmap=) cmdline(mem=) dev tree dev tree
& cmdline(mem=)

(a) how we specify/reserve crash dump kernel's memory on 1st kernel
(b) how we inform userspace tool of crash dump kernel's memory on 1st kernel
- iomem: add "Crash kernel" entry in /proc/iomem
- dev tree: add /chosen/linux,crashkernel-base (and crashkernel-size)
(c) how we specify usable memory on crash dump kernel
- dev tree: add /chosen/memory/linux,usable-memory


As you see, arm64, and arm as well, does things in a *hybrid* way of x86 and ppc.
I think we should go for a uniform way with device tree as ppc does.
(Even for (a), we can use a device tree.)

> Another way is introducing a similar memmap=, but maybe consider only
> system_ram type ranges. For other memory areas still use UEFI memmap.

Yeah, I don't know whether we need additional formats of memmap=, like
memmap=XX$YY or memmap=XX#YY, for ACPI stuffs.

Thanks,
-Takahiro AKASHI

> Thanks
> Dave
>


\
 
 \ /
  Last update: 2015-05-18 11:01    [W:0.131 / U:0.564 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site