lkml.org 
[lkml]   [2022]   [Mar]   [31]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
Date
SubjectRe: [PATCH] hugetlb: Fix hugepages_setup when deal with pernode
From
On 3/31/22 04:23, liupeng (DM) wrote:
> On 2022/3/30 1:43, Mike Kravetz wrote:
>> On 3/28/22 20:59, liupeng (DM) wrote:
>>> On 2022/3/29 10:46, Mike Kravetz wrote:
>>>> Yes, I agree that the change is needed and the current behavior is
>>>> unacceptable.
>>>>
>>>> One remaining question is the change from returning '0' to '1' in the case
>>>> of error.  I do understand this is to prevent the invalid parameter string
>>>> from being passed to init.  It may not be correct/right, but in every other
>>>> case where an invalid parameter in encountered in hugetlb command line
>>>> processing we return "0".  Should we perhaps change all these other places
>>>> to be consistent?  I honestly do not know what is the appropriate behavior
>>>> in these situations.
>>> Thank you for your carefulness and question.
>>>
>>> I have checked default_hugepagesz_setup and hugepages_setup will both print
>>> some information before return '0', so there is no need to print again in
>>> "Unknown kernel command line parameters".
>>>
>>> Should I send another patch to repair the rest "return 0" in hugetlb?
>> I would suggest two patches:
>>
>> 1) Fix the issue with invalid nodes specified.  However, leave the "return 0"
>>     behavior in hugepages_setup to be consistent with the rest of the code.
>>     This patch can be sent to stable with "Fixes: b5389086ad7b" tag as it
>>     addresses an existing issue.
>> 2) Clean up the places where we return 0 and it would be better to return 1.
>>     No cc stable here and just let the changes target future releases.
> I have tried to write a patch as your suggestion, but the best way I can carry it
> out is the original patch. To meet "Fix invalid nodes issue and leave thereturn
> 0 behavior", I have to add the following redundant code:
>
>  invalid:
>         pr_warn("HugeTLB: Invalid hugepages parameter %s\n", p);
> +
> +       /* Allocate gigantic hstates for successfully parsed parameters*/
> +       if (hugetlb_max_hstate && hstate_is_gigantic(parsed_hstate))
> + hugetlb_hstate_alloc_pages(parsed_hstate);
> +       last_mhp = mhp;
> return 0;
>

I was thinking something like the attached (untested). It is very similar to
your original code.

--
Mike Kravetzdiff --git a/mm/hugetlb.c b/mm/hugetlb.c
index f294db835f4b..4deea62dbbac 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -4126,6 +4126,7 @@ static int __init hugepages_setup(char *s)
int count;
unsigned long tmp;
char *p = s;
+ int ret = 1;

if (!parsed_valid_hugepagesz) {
pr_warn("HugeTLB: hugepages=%s does not follow a valid hugepagesz, ignoring\n", s);
@@ -4184,6 +4185,7 @@ static int __init hugepages_setup(char *s)
}
}

+out:
/*
* Global state is always initialized later in hugetlb_init.
* But we need to allocate gigantic hstates here early to still
@@ -4194,11 +4196,12 @@ static int __init hugepages_setup(char *s)

last_mhp = mhp;

- return 1;
+ return ret;

invalid:
pr_warn("HugeTLB: Invalid hugepages parameter %s\n", p);
- return 0;
+ ret = 0;
+ goto out;
}
__setup("hugepages=", hugepages_setup);
\
 
 \ /
  Last update: 2022-03-31 23:12    [W:0.058 / U:1.388 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site