Re: [PATCH v5 04/18] sparc32: mm: Reduce allocation size for PMD and PTE tables

[Guenter Roeck]

On 5/26/20 7:01 AM, Will Deacon wrote:
> On Tue, May 26, 2020 at 02:26:35PM +0100, Will Deacon wrote:
>> On Sun, May 24, 2020 at 03:32:56PM +0300, Mike Rapoport wrote:
>>> On Thu, May 21, 2020 at 04:02:11PM -0700, Guenter Roeck wrote:
>>>> On 5/20/20 12:51 PM, Mike Rapoport wrote:
>>>>> On Wed, May 20, 2020 at 12:03:31PM -0700, Guenter Roeck wrote:
>>>>>> With above patch applied on top of Ira's patch, I get:
>>>>>>
>>>>>> BUG: spinlock recursion on CPU#0, S01syslogd/139
>>>>>>  lock: 0xf5448350, .magic: dead4ead, .owner: S01syslogd/139, .owner_cpu: 0
>>>>>> CPU: 0 PID: 139 Comm: S01syslogd Not tainted 5.7.0-rc6-next-20200518-00002-gb178d2d56f29-dirty #1
>>>>>> [f0067a64 :
>>>>>> do_raw_spin_lock+0xa8/0xd8 ]
>>>>>> [f00d5034 :
>>>>>> copy_page_range+0x328/0x804 ]
>>>>>> [f0025be4 :
>>>>>> dup_mm+0x334/0x434 ]
>>>>>> [f0027124 :
>>>>>> copy_process+0x1224/0x12b0 ]
>>>>>> [f0027344 :
>>>>>> _do_fork+0x54/0x30c ]
>>>>>> [f0027670 :
>>>>>> do_fork+0x5c/0x6c ]
>>>>>> [f000de44 :
>>>>>> sparc_do_fork+0x18/0x38 ]
>>>>>> [f000b7f4 :
>>>>>> do_syscall+0x34/0x40 ]
>>>>>> [5010cd4c :
>>>>>> 0x5010cd4c ]
>>>>>>
>>>>>> Looks like yet another problem.
>>>>>
>>>>> I've checked the patch above on top of the mmots which already has Ira's
>>>>> patches and it booted fine. I've used sparc32_defconfig to build the
>>>>> kernel and qemu-system-sparc with default machine and CPU. 
>>>>>
>>>>
>>>> Try sparc32_defconfig+SMP.
>>>  
>>> I see a differernt problem, but this could be related:
>>>
>>> INIT: version 2.86 booting
>>> rcu: INFO: rcu_sched detected stalls on CPUs/tasks:
>>> 	(detected by 0, t=5252 jiffies, g=-935, q=3)
>>> rcu: All QSes seen, last rcu_sched kthread activity 5252 (-68674--73926), jiffies_till_next_fqs=1, root ->qsmask 0x0
>>> rcu: rcu_sched kthread starved for 5252 jiffies! g-935 f0x2 RCU_GP_WAIT_FQS(5) ->state=0x0 ->cpu=0
>>> rcu: 	Unless rcu_sched kthread gets sufficient CPU time, OOM is now expected behavior.
>>> rcu: RCU grace-period kthread stack dump:
>>> rcu_sched       R  running task        0    10      2 0x00000000
>>>
>>> I'm running a bit old debian [1] with qemu-img-sparc.
>>>
>>> My bisect pointed at commit 8c8f3156dd40 ("sparc32: mm: Reduce
>>> allocation size for PMD and PTE tables"). The commit ID is valid for
>>> next-20200522.
>>
>> Can you try the diff below please?
> 
> Actually, that's racy. New version below!
> 

Applied on top of next-20200526, with defconfig+SMP, I still get:

BUG: Bad page state in process swapper/0  pfn:0069f

many times. Did I have to revert something else ? Sorry, I lost track.


Note that "-smp 2" on SS-10 works for me (with the same page state
messages).

Guenter


> Will
> 
> --->8
> 
> diff --git a/arch/sparc/mm/srmmu.c b/arch/sparc/mm/srmmu.c
> index c861c0f0df73..068029471aa4 100644
> --- a/arch/sparc/mm/srmmu.c
> +++ b/arch/sparc/mm/srmmu.c
> @@ -363,11 +363,16 @@ pgtable_t pte_alloc_one(struct mm_struct *mm)
>  
>  	if ((ptep = pte_alloc_one_kernel(mm)) == 0)
>  		return NULL;
> +
>  	page = pfn_to_page(__nocache_pa((unsigned long)ptep) >> PAGE_SHIFT);
> -	if (!pgtable_pte_page_ctor(page)) {
> -		__free_page(page);
> -		return NULL;
> +
> +	spin_lock(&mm->page_table_lock);
> +	if (page_ref_inc_return(page) == 2 && !pgtable_pte_page_ctor(page)) {
> +		page_ref_dec(page);
> +		ptep = NULL;
>  	}
> +	spin_unlock(&mm->page_table_lock);
> +
>  	return ptep;
>  }
>  
> @@ -376,7 +381,12 @@ void pte_free(struct mm_struct *mm, pgtable_t ptep)
>  	struct page *page;
>  
>  	page = pfn_to_page(__nocache_pa((unsigned long)ptep) >> PAGE_SHIFT);
> -	pgtable_pte_page_dtor(page);
> +
> +	spin_lock(&mm->page_table_lock);
> +	if (page_ref_dec_return(page) == 1)
> +		pgtable_pte_page_dtor(page);
> +	spin_unlock(&mm->page_table_lock);
> +
>  	srmmu_free_nocache(ptep, SRMMU_PTE_TABLE_SIZE);
>  }
>  
> diff --git a/mm/Kconfig b/mm/Kconfig
> index c1acc34c1c35..97458119cce8 100644
> --- a/mm/Kconfig
> +++ b/mm/Kconfig
> @@ -192,6 +192,9 @@ config MEMORY_HOTREMOVE
>  # Default to 4 for wider testing, though 8 might be more appropriate.
>  # ARM's adjust_pte (unused if VIPT) depends on mm-wide page_table_lock.
>  # PA-RISC 7xxx's spinlock_t would enlarge struct page from 32 to 44 bytes.
> +# SPARC32 allocates multiple pte tables within a single page, and therefore
> +# a per-page lock leads to problems when multiple tables need to be locked
> +# at the same time (e.g. copy_page_range()).
>  # DEBUG_SPINLOCK and DEBUG_LOCK_ALLOC spinlock_t also enlarge struct page.
>  #
>  config SPLIT_PTLOCK_CPUS
> @@ -199,6 +202,7 @@ config SPLIT_PTLOCK_CPUS
>  	default "999999" if !MMU
>  	default "999999" if ARM && !CPU_CACHE_VIPT
>  	default "999999" if PARISC && !PA20
> +	default "999999" if SPARC32
>  	default "4"
>  
>  config ARCH_ENABLE_SPLIT_PMD_PTLOCK
>