linux-fsdevel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Xie Yongji <xieyongji@bytedance.com>
To: mst@redhat.com, jasowang@redhat.com, stefanha@redhat.com,
	sgarzare@redhat.com, parav@nvidia.com, akpm@linux-foundation.org,
	rdunlap@infradead.org, willy@infradead.org,
	viro@zeniv.linux.org.uk, axboe@kernel.dk, bcrl@kvack.org,
	corbet@lwn.net
Cc: virtualization@lists.linux-foundation.org,
	netdev@vger.kernel.org, kvm@vger.kernel.org, linux-aio@kvack.org,
	linux-fsdevel@vger.kernel.org, linux-mm@kvack.org
Subject: [RFC v2 11/13] vduse/iova_domain: Support reclaiming bounce pages
Date: Tue, 22 Dec 2020 22:52:19 +0800	[thread overview]
Message-ID: <20201222145221.711-12-xieyongji@bytedance.com> (raw)
In-Reply-To: <20201222145221.711-1-xieyongji@bytedance.com>

Introduce vduse_domain_reclaim() to support reclaiming bounce page
when necessary. We will do reclaiming chunk by chunk. And only
reclaim the iova chunk that no one used.

Signed-off-by: Xie Yongji <xieyongji@bytedance.com>
---
 drivers/vdpa/vdpa_user/iova_domain.c | 83 ++++++++++++++++++++++++++++++++++--
 drivers/vdpa/vdpa_user/iova_domain.h | 10 +++++
 2 files changed, 89 insertions(+), 4 deletions(-)

diff --git a/drivers/vdpa/vdpa_user/iova_domain.c b/drivers/vdpa/vdpa_user/iova_domain.c
index 27022157abc6..c438cc85d33d 100644
--- a/drivers/vdpa/vdpa_user/iova_domain.c
+++ b/drivers/vdpa/vdpa_user/iova_domain.c
@@ -29,6 +29,8 @@ struct vduse_mmap_vma {
 	struct list_head list;
 };
 
+struct percpu_counter vduse_total_bounce_pages;
+
 static inline struct page *
 vduse_domain_get_bounce_page(struct vduse_iova_domain *domain,
 				unsigned long iova)
@@ -48,6 +50,13 @@ vduse_domain_set_bounce_page(struct vduse_iova_domain *domain,
 	unsigned long chunkoff = iova & ~IOVA_CHUNK_MASK;
 	unsigned long pgindex = chunkoff >> PAGE_SHIFT;
 
+	if (page) {
+		domain->chunks[index].used_bounce_pages++;
+		percpu_counter_inc(&vduse_total_bounce_pages);
+	} else {
+		domain->chunks[index].used_bounce_pages--;
+		percpu_counter_dec(&vduse_total_bounce_pages);
+	}
 	domain->chunks[index].bounce_pages[pgindex] = page;
 }
 
@@ -175,6 +184,29 @@ void vduse_domain_remove_mapping(struct vduse_iova_domain *domain,
 	}
 }
 
+static bool vduse_domain_try_unmap(struct vduse_iova_domain *domain,
+				unsigned long iova, size_t size)
+{
+	struct vduse_mmap_vma *mmap_vma;
+	unsigned long uaddr;
+	bool unmap = true;
+
+	mutex_lock(&domain->vma_lock);
+	list_for_each_entry(mmap_vma, &domain->vma_list, list) {
+		if (!mmap_read_trylock(mmap_vma->vma->vm_mm)) {
+			unmap = false;
+			break;
+		}
+
+		uaddr = iova + mmap_vma->vma->vm_start;
+		zap_page_range(mmap_vma->vma, uaddr, size);
+		mmap_read_unlock(mmap_vma->vma->vm_mm);
+	}
+	mutex_unlock(&domain->vma_lock);
+
+	return unmap;
+}
+
 void vduse_domain_unmap(struct vduse_iova_domain *domain,
 			unsigned long iova, size_t size)
 {
@@ -302,6 +334,32 @@ bool vduse_domain_is_direct_map(struct vduse_iova_domain *domain,
 	return atomic_read(&chunk->map_type) == TYPE_DIRECT_MAP;
 }
 
+int vduse_domain_reclaim(struct vduse_iova_domain *domain)
+{
+	struct vduse_iova_chunk *chunk;
+	int i, freed = 0;
+
+	for (i = domain->chunk_num - 1; i >= 0; i--) {
+		chunk = &domain->chunks[i];
+		if (!chunk->used_bounce_pages)
+			continue;
+
+		if (atomic_cmpxchg(&chunk->state, 0, INT_MIN) != 0)
+			continue;
+
+		if (!vduse_domain_try_unmap(domain,
+				chunk->start, IOVA_CHUNK_SIZE)) {
+			atomic_sub(INT_MIN, &chunk->state);
+			break;
+		}
+		freed += vduse_domain_free_bounce_pages(domain,
+				chunk->start, IOVA_CHUNK_SIZE);
+		atomic_sub(INT_MIN, &chunk->state);
+	}
+
+	return freed;
+}
+
 unsigned long vduse_domain_alloc_iova(struct vduse_iova_domain *domain,
 					size_t size, enum iova_map_type type)
 {
@@ -319,10 +377,13 @@ unsigned long vduse_domain_alloc_iova(struct vduse_iova_domain *domain,
 		if (atomic_read(&chunk->map_type) != type)
 			continue;
 
-		iova = gen_pool_alloc_algo(chunk->pool, size,
+		if (atomic_fetch_inc(&chunk->state) >= 0) {
+			iova = gen_pool_alloc_algo(chunk->pool, size,
 					gen_pool_first_fit_align, &data);
-		if (iova)
-			break;
+			if (iova)
+				break;
+		}
+		atomic_dec(&chunk->state);
 	}
 
 	return iova;
@@ -335,6 +396,7 @@ void vduse_domain_free_iova(struct vduse_iova_domain *domain,
 	struct vduse_iova_chunk *chunk = &domain->chunks[index];
 
 	gen_pool_free(chunk->pool, iova, size);
+	atomic_dec(&chunk->state);
 }
 
 static void vduse_iova_chunk_cleanup(struct vduse_iova_chunk *chunk)
@@ -351,7 +413,8 @@ void vduse_iova_domain_destroy(struct vduse_iova_domain *domain)
 
 	for (i = 0; i < domain->chunk_num; i++) {
 		chunk = &domain->chunks[i];
-		vduse_domain_free_bounce_pages(domain,
+		if (chunk->used_bounce_pages)
+			vduse_domain_free_bounce_pages(domain,
 					chunk->start, IOVA_CHUNK_SIZE);
 		vduse_iova_chunk_cleanup(chunk);
 	}
@@ -390,8 +453,10 @@ static int vduse_iova_chunk_init(struct vduse_iova_chunk *chunk,
 	if (!chunk->iova_map)
 		goto err;
 
+	chunk->used_bounce_pages = 0;
 	chunk->start = addr;
 	atomic_set(&chunk->map_type, TYPE_NONE);
+	atomic_set(&chunk->state, 0);
 
 	return 0;
 err:
@@ -440,3 +505,13 @@ struct vduse_iova_domain *vduse_iova_domain_create(size_t size)
 
 	return NULL;
 }
+
+int vduse_domain_init(void)
+{
+	return percpu_counter_init(&vduse_total_bounce_pages, 0, GFP_KERNEL);
+}
+
+void vduse_domain_exit(void)
+{
+	percpu_counter_destroy(&vduse_total_bounce_pages);
+}
diff --git a/drivers/vdpa/vdpa_user/iova_domain.h b/drivers/vdpa/vdpa_user/iova_domain.h
index fe1816287f5f..6815b00629d2 100644
--- a/drivers/vdpa/vdpa_user/iova_domain.h
+++ b/drivers/vdpa/vdpa_user/iova_domain.h
@@ -31,8 +31,10 @@ struct vduse_iova_chunk {
 	struct gen_pool *pool;
 	struct page **bounce_pages;
 	struct vduse_iova_map **iova_map;
+	int used_bounce_pages;
 	unsigned long start;
 	atomic_t map_type;
+	atomic_t state;
 };
 
 struct vduse_iova_domain {
@@ -44,6 +46,8 @@ struct vduse_iova_domain {
 	struct list_head vma_list;
 };
 
+extern struct percpu_counter vduse_total_bounce_pages;
+
 int vduse_domain_add_vma(struct vduse_iova_domain *domain,
 				struct vm_area_struct *vma);
 
@@ -77,6 +81,8 @@ int vduse_domain_bounce_map(struct vduse_iova_domain *domain,
 bool vduse_domain_is_direct_map(struct vduse_iova_domain *domain,
 				unsigned long iova);
 
+int vduse_domain_reclaim(struct vduse_iova_domain *domain);
+
 unsigned long vduse_domain_alloc_iova(struct vduse_iova_domain *domain,
 					size_t size, enum iova_map_type type);
 
@@ -90,4 +96,8 @@ void vduse_iova_domain_destroy(struct vduse_iova_domain *domain);
 
 struct vduse_iova_domain *vduse_iova_domain_create(size_t size);
 
+int vduse_domain_init(void);
+
+void vduse_domain_exit(void);
+
 #endif /* _VDUSE_IOVA_DOMAIN_H */
-- 
2.11.0


  parent reply	other threads:[~2020-12-22 14:55 UTC|newest]

Thread overview: 55+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-12-22 14:52 [RFC v2 00/13] Introduce VDUSE - vDPA Device in Userspace Xie Yongji
2020-12-22 14:52 ` [RFC v2 01/13] mm: export zap_page_range() for driver use Xie Yongji
2020-12-22 15:44   ` Christoph Hellwig
2020-12-22 14:52 ` [RFC v2 02/13] eventfd: track eventfd_signal() recursion depth separately in different cases Xie Yongji
2020-12-22 14:52 ` [RFC v2 03/13] eventfd: Increase the recursion depth of eventfd_signal() Xie Yongji
2020-12-22 14:52 ` [RFC v2 04/13] vdpa: Remove the restriction that only supports virtio-net devices Xie Yongji
2020-12-22 14:52 ` [RFC v2 05/13] vdpa: Pass the netlink attributes to ops.dev_add() Xie Yongji
2020-12-22 14:52 ` [RFC v2 06/13] vduse: Introduce VDUSE - vDPA Device in Userspace Xie Yongji
2020-12-23  8:08   ` Jason Wang
2020-12-23 14:17     ` Yongji Xie
2020-12-24  3:01       ` Jason Wang
2020-12-24  8:34         ` Yongji Xie
2020-12-25  6:59           ` Jason Wang
2021-01-08 13:32   ` Bob Liu
2021-01-10 10:03     ` Yongji Xie
2020-12-22 14:52 ` [RFC v2 07/13] vduse: support get/set virtqueue state Xie Yongji
2020-12-22 14:52 ` [RFC v2 08/13] vdpa: Introduce process_iotlb_msg() in vdpa_config_ops Xie Yongji
2020-12-23  8:36   ` Jason Wang
2020-12-23 11:06     ` Yongji Xie
2020-12-24  2:36       ` Jason Wang
2020-12-24  7:24         ` Yongji Xie
2020-12-22 14:52 ` [RFC v2 09/13] vduse: Add support for processing vhost iotlb message Xie Yongji
2020-12-23  9:05   ` Jason Wang
2020-12-23 12:14     ` [External] " Yongji Xie
2020-12-24  2:41       ` Jason Wang
2020-12-24  7:37         ` Yongji Xie
2020-12-25  2:37           ` Yongji Xie
2020-12-25  7:02             ` Jason Wang
2020-12-25 11:36               ` Yongji Xie
2020-12-25  6:57           ` Jason Wang
2020-12-25 10:31             ` Yongji Xie
2020-12-28  7:43               ` Jason Wang
2020-12-28  8:14                 ` Yongji Xie
2020-12-28  8:43                   ` Jason Wang
2020-12-28  9:12                     ` Yongji Xie
2020-12-29  9:11                       ` Jason Wang
2020-12-29 10:26                         ` Yongji Xie
2020-12-30  6:10                           ` Jason Wang
2020-12-30  7:09                             ` Yongji Xie
2020-12-30  8:41                               ` Jason Wang
2020-12-30 10:12                                 ` Yongji Xie
2020-12-31  2:49                                   ` Jason Wang
2020-12-31  5:15                                     ` Yongji Xie
2020-12-31  5:49                                       ` Jason Wang
2020-12-31  6:52                                         ` Yongji Xie
2020-12-31  7:11                                           ` Jason Wang
2020-12-31  8:00                                             ` Yongji Xie
2020-12-22 14:52 ` [RFC v2 10/13] vduse: grab the module's references until there is no vduse device Xie Yongji
2020-12-22 14:52 ` Xie Yongji [this message]
2020-12-22 14:52 ` [RFC v2 12/13] vduse: Add memory shrinker to reclaim bounce pages Xie Yongji
2020-12-22 14:52 ` [RFC v2 13/13] vduse: Introduce a workqueue for irq injection Xie Yongji
2020-12-23  6:38 ` [RFC v2 00/13] Introduce VDUSE - vDPA Device in Userspace Jason Wang
2020-12-23  8:14   ` Jason Wang
2020-12-23 10:59   ` Yongji Xie
2020-12-24  2:24     ` Jason Wang

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20201222145221.711-12-xieyongji@bytedance.com \
    --to=xieyongji@bytedance.com \
    --cc=akpm@linux-foundation.org \
    --cc=axboe@kernel.dk \
    --cc=bcrl@kvack.org \
    --cc=corbet@lwn.net \
    --cc=jasowang@redhat.com \
    --cc=kvm@vger.kernel.org \
    --cc=linux-aio@kvack.org \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=mst@redhat.com \
    --cc=netdev@vger.kernel.org \
    --cc=parav@nvidia.com \
    --cc=rdunlap@infradead.org \
    --cc=sgarzare@redhat.com \
    --cc=stefanha@redhat.com \
    --cc=viro@zeniv.linux.org.uk \
    --cc=virtualization@lists.linux-foundation.org \
    --cc=willy@infradead.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).