linux-kernel - [RESEND PATCH 2/3] hfs: Replace kmap() with kmap_local

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite for Android: free password hash cracker in your pocket
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20220821180400.8198-3-fmdefrancesco@gmail.com>
Date:   Sun, 21 Aug 2022 20:03:59 +0200
From:   "Fabio M. De Francesco" <fmdefrancesco@...il.com>
To:     "Matthew Wilcox (Oracle)" <willy@...radead.org>,
        Ira Weiny <ira.weiny@...el.com>,
        Jeff Layton <jlayton@...nel.org>,
        Damien Le Moal <damien.lemoal@...nsource.wdc.com>,
        Andrew Morton <akpm@...ux-foundation.org>,
        Christian Brauner <brauner@...nel.org>,
        Arnd Bergmann <arnd@...db.de>, Jens Axboe <axboe@...nel.dk>,
        "Martin K. Petersen" <martin.petersen@...cle.com>,
        Chaitanya Kulkarni <kch@...dia.com>,
        Kees Cook <keescook@...omium.org>,
        Muchun Song <songmuchun@...edance.com>,
        Roman Gushchin <roman.gushchin@...ux.dev>,
        "Theodore Ts'o" <tytso@....edu>, linux-fsdevel@...r.kernel.org,
        linux-kernel@...r.kernel.org
Cc:     "Fabio M. De Francesco" <fmdefrancesco@...il.com>,
        Viacheslav Dubeyko <slava@...eyko.com>
Subject: [RESEND PATCH 2/3] hfs: Replace kmap() with kmap_local_page() in bnode.c

kmap() is being deprecated in favor of kmap_local_page().

Two main problems with kmap(): (1) It comes with an overhead as mapping
space is restricted and protected by a global lock for synchronization and
(2) it also requires global TLB invalidation when the kmap’s pool wraps
and it might block when the mapping space is fully utilized until a slot
becomes available.

With kmap_local_page() the mappings are per thread, CPU local, can take
page faults, and can be called from any context (including interrupts).
It is faster than kmap() in kernels with HIGHMEM enabled. Furthermore,
the tasks can be preempted and, when they are scheduled to run again, the
kernel virtual addresses are restored and still valid.

Since its use in bnode.c is safe everywhere, it should be preferred.

Therefore, replace kmap() with kmap_local_page() in bnode.c. Where
possible, use the suited standard helpers (memzero_page(), memcpy_page())
instead of open coding kmap_local_page() plus memset() or memcpy().

Tested in a QEMU/KVM x86_32 VM, 6GB RAM, booting a kernel with
HIGHMEM64GB enabled.

Suggested-by: Ira Weiny <ira.weiny@...el.com>
Reviewed-by: Viacheslav Dubeyko <slava@...eyko.com>
Signed-off-by: Fabio M. De Francesco <fmdefrancesco@...il.com>
---
 fs/hfs/bnode.c | 32 ++++++++++++--------------------
 1 file changed, 12 insertions(+), 20 deletions(-)

diff --git a/fs/hfs/bnode.c b/fs/hfs/bnode.c
index c83fd0e8404d..2015e42e752a 100644
--- a/fs/hfs/bnode.c
+++ b/fs/hfs/bnode.c
@@ -21,7 +21,6 @@ void hfs_bnode_read(struct hfs_bnode *node, void *buf, int off, int len)
 	int pagenum;
 	int bytes_read;
 	int bytes_to_read;
-	void *vaddr;
 
 	off += node->page_offset;
 	pagenum = off >> PAGE_SHIFT;
@@ -33,9 +32,7 @@ void hfs_bnode_read(struct hfs_bnode *node, void *buf, int off, int len)
 		page = node->page[pagenum];
 		bytes_to_read = min_t(int, len - bytes_read, PAGE_SIZE - off);
 
-		vaddr = kmap_atomic(page);
-		memcpy(buf + bytes_read, vaddr + off, bytes_to_read);
-		kunmap_atomic(vaddr);
+		memcpy_from_page(buf + bytes_read, page, off, bytes_to_read);
 
 		pagenum++;
 		off = 0; /* page offset only applies to the first page */
@@ -80,8 +77,7 @@ void hfs_bnode_write(struct hfs_bnode *node, void *buf, int off, int len)
 	off += node->page_offset;
 	page = node->page[0];
 
-	memcpy(kmap(page) + off, buf, len);
-	kunmap(page);
+	memcpy_to_page(page, off, buf, len);
 	set_page_dirty(page);
 }
 
@@ -105,8 +101,7 @@ void hfs_bnode_clear(struct hfs_bnode *node, int off, int len)
 	off += node->page_offset;
 	page = node->page[0];
 
-	memset(kmap(page) + off, 0, len);
-	kunmap(page);
+	memzero_page(page, off, len);
 	set_page_dirty(page);
 }
 
@@ -123,9 +118,7 @@ void hfs_bnode_copy(struct hfs_bnode *dst_node, int dst,
 	src_page = src_node->page[0];
 	dst_page = dst_node->page[0];
 
-	memcpy(kmap(dst_page) + dst, kmap(src_page) + src, len);
-	kunmap(src_page);
-	kunmap(dst_page);
+	memcpy_page(dst_page, dst, src_page, src, len);
 	set_page_dirty(dst_page);
 }
 
@@ -140,9 +133,9 @@ void hfs_bnode_move(struct hfs_bnode *node, int dst, int src, int len)
 	src += node->page_offset;
 	dst += node->page_offset;
 	page = node->page[0];
-	ptr = kmap(page);
+	ptr = kmap_local_page(page);
 	memmove(ptr + dst, ptr + src, len);
-	kunmap(page);
+	kunmap_local(ptr);
 	set_page_dirty(page);
 }
 
@@ -346,13 +339,14 @@ struct hfs_bnode *hfs_bnode_find(struct hfs_btree *tree, u32 num)
 	if (!test_bit(HFS_BNODE_NEW, &node->flags))
 		return node;
 
-	desc = (struct hfs_bnode_desc *)(kmap(node->page[0]) + node->page_offset);
+	desc = (struct hfs_bnode_desc *)(kmap_local_page(node->page[0]) +
+					 node->page_offset);
 	node->prev = be32_to_cpu(desc->prev);
 	node->next = be32_to_cpu(desc->next);
 	node->num_recs = be16_to_cpu(desc->num_recs);
 	node->type = desc->type;
 	node->height = desc->height;
-	kunmap(node->page[0]);
+	kunmap_local(desc);
 
 	switch (node->type) {
 	case HFS_NODE_HEADER:
@@ -436,14 +430,12 @@ struct hfs_bnode *hfs_bnode_create(struct hfs_btree *tree, u32 num)
 	}
 
 	pagep = node->page;
-	memset(kmap(*pagep) + node->page_offset, 0,
-	       min((int)PAGE_SIZE, (int)tree->node_size));
+	memzero_page(*pagep, node->page_offset,
+		     min((int)PAGE_SIZE, (int)tree->node_size));
 	set_page_dirty(*pagep);
-	kunmap(*pagep);
 	for (i = 1; i < tree->pages_per_bnode; i++) {
-		memset(kmap(*++pagep), 0, PAGE_SIZE);
+		memzero_page(*++pagep, 0, PAGE_SIZE);
 		set_page_dirty(*pagep);
-		kunmap(*pagep);
 	}
 	clear_bit(HFS_BNODE_NEW, &node->flags);
 	wake_up(&node->lock_wq);
-- 
2.37.1