lkml.org 
[lkml]   [2023]   [Oct]   [9]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
SubjectRe: [PATCH v6 2/8] hugetlb: restructure pool allocations
From


On 10/9/23 17:15, Mike Kravetz wrote:
> On 10/09/23 08:04, Mike Kravetz wrote:
>> On 10/09/23 12:11, Konrad Dybcio wrote:
>>> On 9.10.2023 05:29, Mike Kravetz wrote:
>>>> On 10/06/23 15:35, Mike Kravetz wrote:
>>>>> On 10/06/23 23:39, Konrad Dybcio wrote:
>>>>>> On 6.10.2023 05:08, Mike Kravetz wrote:
>>>>>>> On 10/02/23 11:57, Konrad Dybcio wrote:
>>>>>>>> On 9/29/23 22:57, Mike Kravetz wrote:
>>>>>>>>> On 09/27/23 13:26, Konrad Dybcio wrote:
>>>>>>>>>> On 26.09.2023 01:48, Mike Kravetz wrote:
>>>
>>> I managed to get a boot log:
>>>
>>> https://pastebin.com/GwurpCw9
>>>
>>> This is using arch/arm64/boot/dts/qcom/sm8550-mtp.dts for reference
>>>
>>
>> Early on in boot log before the panic, I see this in the log:
>>
>> [ 0.000000] efi: UEFI not found.
>> [ 0.000000] [Firmware Bug]: Kernel image misaligned at boot, please fix your bootloader!
>>
>> Isn't that misalignment pretty serious? Or, is is possible to run with that?
>>
>> There are no hugetlb pages allocated at boot time:
>>
>> [ 0.000000] Kernel command line: PMOS_NO_OUTPUT_REDIRECT console=ttyMSM0 earlycon clk_ignore_unused pd_ignore_unused androidboot.bootdevice=1d84000.ufshc androidboot.fstab_suffix=default androidboot.boot_devices=soc/1d84000.ufshc androidboot.serialno=ab855d8d androidboot.baseband=msm
>>
>> So, the routine where we are panic'ing (gather_bootmem_prealloc) should be a
>> noop. The first thing it does is:
>> list_for_each_entry(m, &huge_boot_pages, list) {
>> ...
>> }
>>
>> However, huge_boot_pages should be empty as initialized here:
>> __initdata LIST_HEAD(huge_boot_pages);
>>
>> At the end of the routine, we call prep_and_add_bootmem_folios to
>> process the local list created withing that above loop:
>>
>> LIST_HEAD(folio_list);
>>
>> This should also be empty and a noop.
>>
>> Is it possible that the misaligned kernel image could make these lists
>> appear as non-empty?
>
> Actually, just saw this:
>
> https://lore.kernel.org/linux-mm/20231009145605.2150897-1-usama.arif@bytedance.com/
>
> Will take a look, although as mentioned above prep_and_add_bootmem_folios on
> an empty list should be a noop.
I'll try it out atop the series tomorrow or so.

Konrad

\
 
 \ /
  Last update: 2023-10-09 23:11    [W:0.077 / U:0.136 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site