[PATCH] mm/vunmap: Add cond_resched() in vunmap_pmd_range

[PATCH] mm/vunmap: Add cond_resched() in vunmap_pmd_range

[Aneesh Kumar K.V]

Like zap_pte_range add cond_resched so that we can avoid softlockups as reported
below. On non-preemptible kernel with large I/O map region (like the one we get
when using persistent memory with sector mode), an unmap of the namespace can report
below softlockups.

22724.027334] watchdog: BUG: soft lockup - CPU#49 stuck for 23s! [ndctl:50777]
 NIP [c0000000000dc224] plpar_hcall+0x38/0x58
 LR [c0000000000d8898] pSeries_lpar_hpte_invalidate+0x68/0xb0
 Call Trace:
 [c0000004e87a7780] [c0000004fb197c00] 0xc0000004fb197c00 (unreliable)
 [c0000004e87a7810] [c00000000007f4e4] flush_hash_page+0x114/0x200
 [c0000004e87a7890] [c0000000000833cc] hpte_need_flush+0x2dc/0x540
 [c0000004e87a7950] [c0000000003f5798] vunmap_page_range+0x538/0x6f0
 [c0000004e87a7a70] [c0000000003f76d0] free_unmap_vmap_area+0x30/0x70
 [c0000004e87a7aa0] [c0000000003f7a6c] remove_vm_area+0xfc/0x140
 [c0000004e87a7ad0] [c0000000003f7dd8] __vunmap+0x68/0x270
 [c0000004e87a7b50] [c000000000079de4] __iounmap.part.0+0x34/0x60
 [c0000004e87a7bb0] [c000000000376394] memunmap+0x54/0x70
 [c0000004e87a7bd0] [c000000000881d7c] release_nodes+0x28c/0x300
 [c0000004e87a7c40] [c00000000087a65c] device_release_driver_internal+0x16c/0x280
 [c0000004e87a7c80] [c000000000876fc4] unbind_store+0x124/0x170
 [c0000004e87a7cd0] [c000000000875be4] drv_attr_store+0x44/0x60
 [c0000004e87a7cf0] [c00000000057c734] sysfs_kf_write+0x64/0x90
 [c0000004e87a7d10] [c00000000057bc10] kernfs_fop_write+0x1b0/0x290
 [c0000004e87a7d60] [c000000000488e6c] __vfs_write+0x3c/0x70
 [c0000004e87a7d80] [c00000000048c868] vfs_write+0xd8/0x260
 [c0000004e87a7dd0] [c00000000048ccac] ksys_write+0xdc/0x130
 [c0000004e87a7e20] [c00000000000b588] system_call+0x5c/0x70

Reported-by: Harish Sriram <harish@linux.ibm.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
---
 mm/vmalloc.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/mm/vmalloc.c b/mm/vmalloc.c
index 5a2b55c8dd9a..128d20d2d6cb 100644
--- a/mm/vmalloc.c
+++ b/mm/vmalloc.c
@@ -102,6 +102,8 @@ static void vunmap_pmd_range(pud_t *pud, unsigned long addr, unsigned long end,
 		if (pmd_none_or_clear_bad(pmd))
 			continue;
 		vunmap_pte_range(pmd, addr, next, mask);
+
+		cond_resched();
 	} while (pmd++, addr = next, addr != end);
 }
 
-- 
2.26.2

Re: [PATCH] mm/vunmap: Add cond_resched() in vunmap_pmd_range

[Andrew Morton]

On Fri,  7 Aug 2020 13:29:33 +0530 "Aneesh Kumar K.V" <aneesh.kumar@linux.ibm.com> wrote:

> Like zap_pte_range add cond_resched so that we can avoid softlockups as reported
> below. On non-preemptible kernel with large I/O map region (like the one we get
> when using persistent memory with sector mode), an unmap of the namespace can report
> below softlockups.
> 
> ...
>
> --- a/mm/vmalloc.c
> +++ b/mm/vmalloc.c
> @@ -102,6 +102,8 @@ static void vunmap_pmd_range(pud_t *pud, unsigned long addr, unsigned long end,
>  		if (pmd_none_or_clear_bad(pmd))
>  			continue;
>  		vunmap_pte_range(pmd, addr, next, mask);
> +
> +		cond_resched();
>  	} while (pmd++, addr = next, addr != end);
>  }

That looks sane.  I'll aim for merging it in 5.9-rc2ish, with a
cc:stable, OK?

Re: [PATCH] mm/vunmap: Add cond_resched() in vunmap_pmd_range

[Aneesh Kumar K.V]

On 8/11/20 10:16 AM, Andrew Morton wrote:
> On Fri,  7 Aug 2020 13:29:33 +0530 "Aneesh Kumar K.V" <aneesh.kumar@linux.ibm.com> wrote:
> 
>> Like zap_pte_range add cond_resched so that we can avoid softlockups as reported
>> below. On non-preemptible kernel with large I/O map region (like the one we get
>> when using persistent memory with sector mode), an unmap of the namespace can report
>> below softlockups.
>>
>> ...
>>
>> --- a/mm/vmalloc.c
>> +++ b/mm/vmalloc.c
>> @@ -102,6 +102,8 @@ static void vunmap_pmd_range(pud_t *pud, unsigned long addr, unsigned long end,
>>   		if (pmd_none_or_clear_bad(pmd))
>>   			continue;
>>   		vunmap_pte_range(pmd, addr, next, mask);
>> +
>> +		cond_resched();
>>   	} while (pmd++, addr = next, addr != end);
>>   }
> 
> That looks sane.  I'll aim for merging it in 5.9-rc2ish, with a
> cc:stable, OK?
> 

Sure. I was not so particular about stable because it was only a 
softlockup warning.

-aneesh