Re: [PATCH 2/3] mm: drop VMA lock before waiting for migration

From: Alistair Popple <apopple@nvidia.com>
To: Suren Baghdasaryan <surenb@google.com>
Cc: akpm@linux-foundation.org, willy@infradead.org,
	hannes@cmpxchg.org, mhocko@suse.com, josef@toxicpanda.com,
	jack@suse.cz, ldufour@linux.ibm.com, laurent.dufour@fr.ibm.com,
	michel@lespinasse.org, liam.howlett@oracle.com,
	jglisse@google.com, vbabka@suse.cz, minchan@google.com,
	dave@stgolabs.net, punit.agrawal@bytedance.com,
	lstoakes@gmail.com, hdanton@sina.com, linux-mm@kvack.org,
	linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org,
	kernel-team@android.com
Subject: Re: [PATCH 2/3] mm: drop VMA lock before waiting for migration
Date: Tue, 02 May 2023 23:21:08 +1000	[thread overview]
Message-ID: <875y9aj23u.fsf@nvidia.com> (raw)
In-Reply-To: <20230501175025.36233-2-surenb@google.com>

Suren Baghdasaryan <surenb@google.com> writes:

[...]

> diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
> index 306a3d1a0fa6..b3b57c6da0e1 100644
> --- a/include/linux/mm_types.h
> +++ b/include/linux/mm_types.h
> @@ -1030,6 +1030,7 @@ typedef __bitwise unsigned int vm_fault_t;
>   *				fsync() to complete (for synchronous page faults
>   *				in DAX)
>   * @VM_FAULT_COMPLETED:		->fault completed, meanwhile mmap lock released
> + * @VM_FAULT_VMA_UNLOCKED:	VMA lock was released

A note here saying vmf->vma should no longer be accessed would be nice.

>   * @VM_FAULT_HINDEX_MASK:	mask HINDEX value
>   *
>   */
> @@ -1047,6 +1048,7 @@ enum vm_fault_reason {
>  	VM_FAULT_DONE_COW       = (__force vm_fault_t)0x001000,
>  	VM_FAULT_NEEDDSYNC      = (__force vm_fault_t)0x002000,
>  	VM_FAULT_COMPLETED      = (__force vm_fault_t)0x004000,
> +	VM_FAULT_VMA_UNLOCKED   = (__force vm_fault_t)0x008000,
>  	VM_FAULT_HINDEX_MASK    = (__force vm_fault_t)0x0f0000,
>  };
>  
> @@ -1070,7 +1072,9 @@ enum vm_fault_reason {
>  	{ VM_FAULT_RETRY,               "RETRY" },	\
>  	{ VM_FAULT_FALLBACK,            "FALLBACK" },	\
>  	{ VM_FAULT_DONE_COW,            "DONE_COW" },	\
> -	{ VM_FAULT_NEEDDSYNC,           "NEEDDSYNC" }
> +	{ VM_FAULT_NEEDDSYNC,           "NEEDDSYNC" },	\
> +	{ VM_FAULT_COMPLETED,           "COMPLETED" },	\

VM_FAULT_COMPLETED isn't used in this patch, guessing that's snuck in
from one of the other patches in the series?

> +	{ VM_FAULT_VMA_UNLOCKED,        "VMA_UNLOCKED" }
>  
>  struct vm_special_mapping {
>  	const char *name;	/* The name, e.g. "[vdso]". */
> diff --git a/mm/memory.c b/mm/memory.c
> index 41f45819a923..8222acf74fd3 100644
> --- a/mm/memory.c
> +++ b/mm/memory.c
> @@ -3714,8 +3714,16 @@ vm_fault_t do_swap_page(struct vm_fault *vmf)
>  	entry = pte_to_swp_entry(vmf->orig_pte);
>  	if (unlikely(non_swap_entry(entry))) {
>  		if (is_migration_entry(entry)) {
> -			migration_entry_wait(vma->vm_mm, vmf->pmd,
> -					     vmf->address);
> +			/* Save mm in case VMA lock is dropped */
> +			struct mm_struct *mm = vma->vm_mm;
> +
> +			if (vmf->flags & FAULT_FLAG_VMA_LOCK) {
> +				/* No need to hold VMA lock for migration */
> +				vma_end_read(vma);
> +				/* CAUTION! VMA can't be used after this */
> +				ret |= VM_FAULT_VMA_UNLOCKED;
> +			}
> +			migration_entry_wait(mm, vmf->pmd, vmf->address);
>  		} else if (is_device_exclusive_entry(entry)) {
>  			vmf->page = pfn_swap_entry_to_page(entry);
>  			ret = remove_device_exclusive_entry(vmf);