Skip to contentSkip to conversation

fs/ufs: replace kmap() with kmap_local_page

PATCH Superseded Fabio M. De Francesco 1 participant 6 messages Dec 17, 2022 – Dec 21, 2022

Subsystems (2)

v3 of 5 Superseded Latest is v5 →

lore
Share & tools
lore
#0 lore raw

v3 cover fs/ufs: replace kmap() with kmap_local_page

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

There are two main problems with kmap(): (1) It comes with an overhead as the 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 fs/ufs is safe everywhere, it should be preferred.

Therefore, replace kmap() with kmap_local_page() in fs/ufs. kunmap_local() requires the mapping address, so return that address from ufs_get_page() to be used in ufs_put_page().

This series could have not been ever made because nothing prevented the previous patch from working properly but Al Viro made a long series of very appreciated comments about how many unnecessary and redundant lines of code I could have removed. He could see things I was entirely unable to notice. Furthermore, he also provided solutions and details about how I could decompose a single patch into a small series of three independent units.[1][2][3]

I want to thank him so much for the patience, kindness and the time he decided to spend to provide those analysis and write three messages full of interesting insights.[1][2][3]

Changes from v1:
1/3: No changes.
2/3: Restore the return of "err" that was mistakenly deleted
together with the removal of the "out" label in ufs_add_link(). Thanks to Al Viro.[4]
Return the address of the kmap()'ed page instead of a pointer to a pointer to the mapped page; a page_address() had been overlooked in ufs_get_page(). Thanks to Al Viro.[5]
3/3: Return the kernel virtual address got from the call to
kmap_local_page() after conversion from kmap(). Again thanks to Al Viro.[6]

Changes from v2:
1/3: No changes.
2/3: Rework ufs_get_page() because the previous version had two
errors: (1) It could return an invalid pages with the out argument "page" and (2) it could return "page_address(page)" also in cases where read_mapping_page() returned an error and the page is never kmap()'ed. Thanks to Al Viro.[7]
3/3: Rework ufs_get_page() after conversion to
kmap_local_page(), in accordance to the last changes in 2/3.

[1] https://lore.kernel.org/lkml/Y4E++JERgUMoqfjG@ZenIV/
[2] https://lore.kernel.org/lkml/Y4FG0O7VWTTng5yh@ZenIV/
[3] https://lore.kernel.org/lkml/Y4ONIFJatIGsVNpf@ZenIV/
[4] https://lore.kernel.org/lkml/Y5Zc0qZ3+zsI74OZ@ZenIV/
[5] https://lore.kernel.org/lkml/Y5ZZy23FFAnQDR3C@ZenIV/
[6] https://lore.kernel.org/lkml/Y5ZcMPzPG9h6C9eh@ZenIV/
[7] https://lore.kernel.org/lkml/Y5glgpD7fFifC4Fi@ZenIV/#t

The cover letter of the v1 series is at
https://lore.kernel.org/lkml/[email protected]/
The cover letter of the v2 series is at
https://lore.kernel.org/lkml/[email protected]/

Fabio M. De Francesco (3):
fs/ufs: Use the offset_in_page() helper
fs/ufs: Change the signature of ufs_get_page()
fs/ufs: Replace kmap() with kmap_local_page()

 fs/ufs/dir.c | 134 +++++++++++++++++++++++++++------------------------
 1 file changed, 71 insertions(+), 63 deletions(-)
Signature · 1 line
2.39.0
#1 lore raw

PATCH v3 1/3 fs/ufs: Use the offset_in_page() helper

Use the offset_in_page() helper because it is more suitable than doing explicit subtractions between pointers to directory entries and kernel virtual addresses of mapped pages.

---
 fs/ufs/dir.c | 12 +++++-------
 1 file changed, 5 insertions(+), 7 deletions(-)
fs/ufs/dir.c +5 -7
index 391efaf1d528..69f78583c9c1 100644
@@ -87,8 +87,7 @@ void ufs_set_link(struct inode *dir, struct ufs_dir_entry *de,
 		  struct page *page, struct inode *inode,
 		  bool update_times)
 {
-	loff_t pos = page_offset(page) +
-			(char *) de - (char *) page_address(page);
+	loff_t pos = page_offset(page) + offset_in_page(de);
 	unsigned len = fs16_to_cpu(dir->i_sb, de->d_reclen);
 	int err;
 
@@ -371,8 +370,7 @@ int ufs_add_link(struct dentry *dentry, struct inode *inode)
 	return -EINVAL;
 
 got_it:
-	pos = page_offset(page) +
-			(char*)de - (char*)page_address(page);
+	pos = page_offset(page) + offset_in_page(de);
 	err = ufs_prepare_chunk(page, pos, rec_len);
 	if (err)
 		goto out_unlock;
@@ -497,8 +495,8 @@ int ufs_delete_entry(struct inode *inode, struct ufs_dir_entry *dir,
 {
 	struct super_block *sb = inode->i_sb;
 	char *kaddr = page_address(page);
-	unsigned from = ((char*)dir - kaddr) & ~(UFS_SB(sb)->s_uspi->s_dirblksize - 1);
-	unsigned to = ((char*)dir - kaddr) + fs16_to_cpu(sb, dir->d_reclen);
+	unsigned int from = offset_in_page(dir) & ~(UFS_SB(sb)->s_uspi->s_dirblksize - 1);
+	unsigned int to = offset_in_page(dir) + fs16_to_cpu(sb, dir->d_reclen);
 	loff_t pos;
 	struct ufs_dir_entry *pde = NULL;
 	struct ufs_dir_entry *de = (struct ufs_dir_entry *) (kaddr + from);
@@ -522,7 +520,7 @@ int ufs_delete_entry(struct inode *inode, struct ufs_dir_entry *dir,
 		de = ufs_next_entry(sb, de);
 	}
 	if (pde)
-		from = (char*)pde - (char*)page_address(page);
+		from = offset_in_page(pde);
 
 	pos = page_offset(page) + from;
 	lock_page(page);
Signature · 1 line
2.39.0
#2 lore raw

PATCH v3 2/3 fs/ufs: Change the signature of ufs_get_page()

In reply to Fabio M. De Francesco

Change the signature of ufs_get_page() in order to prepare this function to the conversion to the use of kmap_local_page(). Change also those call sites which are required to conform its invocations to the new signature.

---
 fs/ufs/dir.c | 49 +++++++++++++++++++++----------------------------
 1 file changed, 21 insertions(+), 28 deletions(-)
fs/ufs/dir.c +21 -28
index 69f78583c9c1..9fa86614d2d1 100644
@@ -185,7 +185,7 @@ static bool ufs_check_page(struct page *page)
 	return false;
 }
 
-static struct page *ufs_get_page(struct inode *dir, unsigned long n)
+static void *ufs_get_page(struct inode *dir, unsigned long n, struct page **p)
 {
 	struct address_space *mapping = dir->i_mapping;
 	struct page *page = read_mapping_page(mapping, n, NULL);
@@ -195,8 +195,10 @@ static struct page *ufs_get_page(struct inode *dir, unsigned long n)
 			if (!ufs_check_page(page))
 				goto fail;
 		}
+		*p = page;
+		return page_address(page);
 	}
-	return page;
+	return ERR_CAST(page);
 
 fail:
 	ufs_put_page(page);
@@ -227,15 +229,12 @@ ufs_next_entry(struct super_block *sb, struct ufs_dir_entry *p)
 
 struct ufs_dir_entry *ufs_dotdot(struct inode *dir, struct page **p)
 {
-	struct page *page = ufs_get_page(dir, 0);
-	struct ufs_dir_entry *de = NULL;
+	struct ufs_dir_entry *de = ufs_get_page(dir, 0, p);
 
-	if (!IS_ERR(page)) {
-		de = ufs_next_entry(dir->i_sb,
-				    (struct ufs_dir_entry *)page_address(page));
-		*p = page;
-	}
-	return de;
+	if (!IS_ERR(de))
+		return ufs_next_entry(dir->i_sb, de);
+	else
+		return NULL;
 }
 
 /*
@@ -273,11 +272,10 @@ struct ufs_dir_entry *ufs_find_entry(struct inode *dir, const struct qstr *qstr,
 		start = 0;
 	n = start;
 	do {
-		char *kaddr;
-		page = ufs_get_page(dir, n);
-		if (!IS_ERR(page)) {
-			kaddr = page_address(page);
-			de = (struct ufs_dir_entry *) kaddr;
+		char *kaddr = ufs_get_page(dir, n, &page);
+
+		if (!IS_ERR(kaddr)) {
+			de = (struct ufs_dir_entry *)kaddr;
 			kaddr += ufs_last_byte(dir, n) - reclen;
 			while ((char *) de <= kaddr) {
 				if (ufs_match(sb, namelen, name, de))
@@ -328,12 +326,10 @@ int ufs_add_link(struct dentry *dentry, struct inode *inode)
 	for (n = 0; n <= npages; n++) {
 		char *dir_end;
 
-		page = ufs_get_page(dir, n);
-		err = PTR_ERR(page);
-		if (IS_ERR(page))
-			goto out;
+		kaddr = ufs_get_page(dir, n, &page);
+		if (IS_ERR(kaddr))
+			return PTR_ERR(kaddr);
 		lock_page(page);
-		kaddr = page_address(page);
 		dir_end = kaddr + ufs_last_byte(dir, n);
 		de = (struct ufs_dir_entry *)kaddr;
 		kaddr += PAGE_SIZE - reclen;
@@ -395,7 +391,6 @@ int ufs_add_link(struct dentry *dentry, struct inode *inode)
 	/* OFFSET_CACHE */
 out_put:
 	ufs_put_page(page);
-out:
 	return err;
 out_unlock:
 	unlock_page(page);
@@ -429,6 +424,7 @@ ufs_readdir(struct file *file, struct dir_context *ctx)
 	unsigned chunk_mask = ~(UFS_SB(sb)->s_uspi->s_dirblksize - 1);
 	bool need_revalidate = !inode_eq_iversion(inode, file->f_version);
 	unsigned flags = UFS_SB(sb)->s_flags;
+	struct page *page;
 
 	UFSD("BEGIN\n");
 
@@ -439,16 +435,14 @@ ufs_readdir(struct file *file, struct dir_context *ctx)
 		char *kaddr, *limit;
 		struct ufs_dir_entry *de;
 
-		struct page *page = ufs_get_page(inode, n);
-
-		if (IS_ERR(page)) {
+		kaddr = ufs_get_page(inode, n, &page);
+		if (IS_ERR(kaddr)) {
 			ufs_error(sb, __func__,
 				  "bad page in #%lu",
 				  inode->i_ino);
 			ctx->pos += PAGE_SIZE - offset;
 			return -EIO;
 		}
-		kaddr = page_address(page);
 		if (unlikely(need_revalidate)) {
 			if (offset) {
 				offset = ufs_validate_entry(sb, kaddr, offset, chunk_mask);
@@ -595,12 +589,11 @@ int ufs_empty_dir(struct inode * inode)
 	for (i = 0; i < npages; i++) {
 		char *kaddr;
 		struct ufs_dir_entry *de;
-		page = ufs_get_page(inode, i);
 
-		if (IS_ERR(page))
+		kaddr = ufs_get_page(inode, i, &page);
+		if (IS_ERR(kaddr))
 			continue;
 
-		kaddr = page_address(page);
 		de = (struct ufs_dir_entry *)kaddr;
 		kaddr += ufs_last_byte(inode, i) - UFS_DIR_REC_LEN(1);
 
Signature · 1 line
2.39.0
#3 lore raw

PATCH v3 3/3 fs/ufs: Replace kmap() with kmap_local_page()

In reply to Fabio M. De Francesco

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

There are two main problems with kmap(): (1) It comes with an overhead as the 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 fs/ufs is safe everywhere, it should be preferred.

Therefore, replace kmap() with kmap_local_page() in fs/ufs. kunmap_local() requires the mapping address, so return that address from ufs_get_page() to be used in ufs_put_page().

---
 fs/ufs/dir.c | 75 ++++++++++++++++++++++++++++++++--------------------
 1 file changed, 46 insertions(+), 29 deletions(-)
fs/ufs/dir.c +46 -29
index 9fa86614d2d1..a9dd5023b604 100644
@@ -61,9 +61,9 @@ static int ufs_commit_chunk(struct page *page, loff_t pos, unsigned len)
 	return err;
 }
 
-static inline void ufs_put_page(struct page *page)
+static inline void ufs_put_page(struct page *page, void *page_addr)
 {
-	kunmap(page);
+	kunmap((void *)((unsigned long)page_addr & PAGE_MASK));
 	put_page(page);
 }
 
@@ -76,7 +76,7 @@ ino_t ufs_inode_by_name(struct inode *dir, const struct qstr *qstr)
 	de = ufs_find_entry(dir, qstr, &page);
 	if (de) {
 		res = fs32_to_cpu(dir->i_sb, de->d_ino);
-		ufs_put_page(page);
+		ufs_put_page(page, de);
 	}
 	return res;
 }
@@ -99,18 +99,17 @@ void ufs_set_link(struct inode *dir, struct ufs_dir_entry *de,
 	ufs_set_de_type(dir->i_sb, de, inode->i_mode);
 
 	err = ufs_commit_chunk(page, pos, len);
-	ufs_put_page(page);
+	ufs_put_page(page, de);
 	if (update_times)
 		dir->i_mtime = dir->i_ctime = current_time(dir);
 	mark_inode_dirty(dir);
 }
 
 
-static bool ufs_check_page(struct page *page)
+static bool ufs_check_page(struct page *page, char *kaddr)
 {
 	struct inode *dir = page->mapping->host;
 	struct super_block *sb = dir->i_sb;
-	char *kaddr = page_address(page);
 	unsigned offs, rec_len;
 	unsigned limit = PAGE_SIZE;
 	const unsigned chunk_mask = UFS_SB(sb)->s_uspi->s_dirblksize - 1;
@@ -185,23 +184,32 @@ static bool ufs_check_page(struct page *page)
 	return false;
 }
 
+/*
+ * Calls to ufs_get_page()/ufs_put_page() must be nested according to the
+ * rules documented in kmap_local_page()/kunmap_local().
+ *
+ * NOTE: ufs_find_entry() and ufs_dotdot() act as calls to ufs_get_page()
+ * and must be treated accordingly for nesting purposes.
+ */
 static void *ufs_get_page(struct inode *dir, unsigned long n, struct page **p)
 {
+	char *kaddr;
+
 	struct address_space *mapping = dir->i_mapping;
 	struct page *page = read_mapping_page(mapping, n, NULL);
 	if (!IS_ERR(page)) {
-		kmap(page);
+		kaddr = kmap_local_page(page);
 		if (unlikely(!PageChecked(page))) {
-			if (!ufs_check_page(page))
+			if (!ufs_check_page(page, kaddr))
 				goto fail;
 		}
 		*p = page;
-		return page_address(page);
+		return kaddr;
 	}
 	return ERR_CAST(page);
 
 fail:
-	ufs_put_page(page);
+	ufs_put_page(page, kaddr);
 	return ERR_PTR(-EIO);
 }
 
@@ -227,6 +235,13 @@ ufs_next_entry(struct super_block *sb, struct ufs_dir_entry *p)
 					fs16_to_cpu(sb, p->d_reclen));
 }
 
+/*
+ * Calls to ufs_get_page()/ufs_put_page() must be nested according to the
+ * rules documented in kmap_local_page()/kunmap_local().
+ *
+ * ufs_dotdot() acts as a call to ufs_get_page() and must be treated
+ * accordingly for nesting purposes.
+ */
 struct ufs_dir_entry *ufs_dotdot(struct inode *dir, struct page **p)
 {
 	struct ufs_dir_entry *de = ufs_get_page(dir, 0, p);
@@ -238,12 +253,15 @@ struct ufs_dir_entry *ufs_dotdot(struct inode *dir, struct page **p)
 }
 
 /*
- *	ufs_find_entry()
+ * Finds an entry in the specified directory with the wanted name. It returns a
+ * pointer to the directory's entry. The page in which the entry was found is
+ * in the res_page out parameter. The page is returned mapped and unlocked.
+ * The entry is guaranteed to be valid.
  *
- * finds an entry in the specified directory with the wanted name. It
- * returns the page in which the entry was found, and the entry itself
- * (as a parameter - res_dir). Page is returned mapped and unlocked.
- * Entry is guaranteed to be valid.
+ * On Success ufs_put_page() should be called on *res_page.
+ *
+ * ufs_find_entry() acts as a call to ufs_get_page() and must be treated
+ * accordingly for nesting purposes.
  */
 struct ufs_dir_entry *ufs_find_entry(struct inode *dir, const struct qstr *qstr,
 				     struct page **res_page)
@@ -282,7 +300,7 @@ struct ufs_dir_entry *ufs_find_entry(struct inode *dir, const struct qstr *qstr,
 					goto found;
 				de = ufs_next_entry(sb, de);
 			}
-			ufs_put_page(page);
+			ufs_put_page(page, kaddr);
 		}
 		if (++n >= npages)
 			n = 0;
@@ -360,7 +378,7 @@ int ufs_add_link(struct dentry *dentry, struct inode *inode)
 			de = (struct ufs_dir_entry *) ((char *) de + rec_len);
 		}
 		unlock_page(page);
-		ufs_put_page(page);
+		ufs_put_page(page, kaddr);
 	}
 	BUG();
 	return -EINVAL;
@@ -390,7 +408,7 @@ int ufs_add_link(struct dentry *dentry, struct inode *inode)
 	mark_inode_dirty(dir);
 	/* OFFSET_CACHE */
 out_put:
-	ufs_put_page(page);
+	ufs_put_page(page, kaddr);
 	return err;
 out_unlock:
 	unlock_page(page);
@@ -468,13 +486,13 @@ ufs_readdir(struct file *file, struct dir_context *ctx)
 					       ufs_get_de_namlen(sb, de),
 					       fs32_to_cpu(sb, de->d_ino),
 					       d_type)) {
-					ufs_put_page(page);
+					ufs_put_page(page, kaddr);
 					return 0;
 				}
 			}
 			ctx->pos += fs16_to_cpu(sb, de->d_reclen);
 		}
-		ufs_put_page(page);
+		ufs_put_page(page, kaddr);
 	}
 	return 0;
 }
@@ -485,10 +503,10 @@ ufs_readdir(struct file *file, struct dir_context *ctx)
  * previous entry.
  */
 int ufs_delete_entry(struct inode *inode, struct ufs_dir_entry *dir,
-		     struct page * page)
+		     struct page *page)
 {
 	struct super_block *sb = inode->i_sb;
-	char *kaddr = page_address(page);
+	char *kaddr = (char *)((unsigned long)dir & PAGE_MASK);
 	unsigned int from = offset_in_page(dir) & ~(UFS_SB(sb)->s_uspi->s_dirblksize - 1);
 	unsigned int to = offset_in_page(dir) + fs16_to_cpu(sb, dir->d_reclen);
 	loff_t pos;
@@ -527,7 +545,7 @@ int ufs_delete_entry(struct inode *inode, struct ufs_dir_entry *dir,
 	inode->i_ctime = inode->i_mtime = current_time(inode);
 	mark_inode_dirty(inode);
 out:
-	ufs_put_page(page);
+	ufs_put_page(page, kaddr);
 	UFSD("EXIT\n");
 	return err;
 }
@@ -551,8 +569,7 @@ int ufs_make_empty(struct inode * inode, struct inode *dir)
 		goto fail;
 	}
 
-	kmap(page);
-	base = (char*)page_address(page);
+	base = kmap_local_page(page);
 	memset(base, 0, PAGE_SIZE);
 
 	de = (struct ufs_dir_entry *) base;
@@ -569,7 +586,7 @@ int ufs_make_empty(struct inode * inode, struct inode *dir)
 	de->d_reclen = cpu_to_fs16(sb, chunk_size - UFS_DIR_REC_LEN(1));
 	ufs_set_de_namlen(sb, de, 2);
 	strcpy (de->d_name, "..");
-	kunmap(page);
+	kunmap_local(base);
 
 	err = ufs_commit_chunk(page, 0, chunk_size);
 fail:
@@ -585,9 +602,9 @@ int ufs_empty_dir(struct inode * inode)
 	struct super_block *sb = inode->i_sb;
 	struct page *page = NULL;
 	unsigned long i, npages = dir_pages(inode);
+	char *kaddr;
 
 	for (i = 0; i < npages; i++) {
-		char *kaddr;
 		struct ufs_dir_entry *de;
 
 		kaddr = ufs_get_page(inode, i, &page);
@@ -620,12 +637,12 @@ int ufs_empty_dir(struct inode * inode)
 			}
 			de = ufs_next_entry(sb, de);
 		}
-		ufs_put_page(page);
+		ufs_put_page(page, kaddr);
 	}
 	return 1;
 
 not_empty:
-	ufs_put_page(page);
+	ufs_put_page(page, kaddr);
 	return 0;
 }
 
Signature · 1 line
2.39.0
#4 lore raw

Re: [PATCH v3 3/3] fs/ufs: Replace kmap() with kmap_local_page()

Quoted 32 lines · On sabato 17 dicembre 2022 19:47:49 CET Fabio M. De Francesco wrote:

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

There are two main problems with kmap(): (1) It comes with an overhead as the 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 fs/ufs is safe everywhere, it should be preferred.

Therefore, replace kmap() with kmap_local_page() in fs/ufs. kunmap_local() requires the mapping address, so return that address from ufs_get_page() to be used in ufs_put_page().

---
 fs/ufs/dir.c | 75 ++++++++++++++++++++++++++++++++--------------------
 1 file changed, 46 insertions(+), 29 deletions(-)
fs/ufs/dir.c +0 -0
index 9fa86614d2d1..a9dd5023b604 100644

fs/ufs/dir.c static int ufs_commit_chunk(struct page *page, loff_t pos, · line 66

Quoted 8 lines
fs/ufs/dir.c +2 -2
@@ -61,9 +61,9 @@ static int ufs_commit_chunk(struct page *page, loff_t pos,
 unsigned len) return err;
 }
 
-static inline void ufs_put_page(struct page *page)
+static inline void ufs_put_page(struct page *page, void *page_addr)
 {
-	kunmap(page);
+	kunmap((void *)((unsigned long)page_addr & PAGE_MASK));

While working on patch 3/3 of a similar series for fs/sysv a few minutes ago, I noticed this call to kunmap() instead of kunmap_local().

I will wait a day or two before sending the next version, to leave additional time for those who want to comment / review.

Fabio

fs/ufs/dir.c ino_t ufs_inode_by_name(struct inode *dir, const struct

Quoted 4 lines
fs/ufs/dir.c +0 -0
 	put_page(page);
 }
 
@@ -76,7 +76,7 @@ ino_t ufs_inode_by_name(struct inode *dir, const struct 

qstr

fs/ufs/dir.c void ufs_set_link(struct inode *dir, struct

Quoted 9 lines
fs/ufs/dir.c +1 -1
 *qstr) de = ufs_find_entry(dir, qstr, &page);
 	if (de) {
 		res = fs32_to_cpu(dir->i_sb, de->d_ino);
-		ufs_put_page(page);
+		ufs_put_page(page, de);
 	}
 	return res;
 }
@@ -99,18 +99,17 @@ void ufs_set_link(struct inode *dir, struct 

ufs_dir_entry

Quoted 72 lines
fs/ufs/dir.c +22 -7
 *de, ufs_set_de_type(dir->i_sb, de, inode->i_mode);
 
 	err = ufs_commit_chunk(page, pos, len);
-	ufs_put_page(page);
+	ufs_put_page(page, de);
 	if (update_times)
 		dir->i_mtime = dir->i_ctime = current_time(dir);
 	mark_inode_dirty(dir);
 }
 
 
-static bool ufs_check_page(struct page *page)
+static bool ufs_check_page(struct page *page, char *kaddr)
 {
 	struct inode *dir = page->mapping->host;
 	struct super_block *sb = dir->i_sb;
-	char *kaddr = page_address(page);
 	unsigned offs, rec_len;
 	unsigned limit = PAGE_SIZE;
 	const unsigned chunk_mask = UFS_SB(sb)->s_uspi->s_dirblksize - 1;
@@ -185,23 +184,32 @@ static bool ufs_check_page(struct page *page)
 	return false;
 }
 
+/*
+ * Calls to ufs_get_page()/ufs_put_page() must be nested according to the
+ * rules documented in kmap_local_page()/kunmap_local().
+ *
+ * NOTE: ufs_find_entry() and ufs_dotdot() act as calls to ufs_get_page()
+ * and must be treated accordingly for nesting purposes.
+ */
 static void *ufs_get_page(struct inode *dir, unsigned long n, struct page
 **p) {
+	char *kaddr;
+
 	struct address_space *mapping = dir->i_mapping;
 	struct page *page = read_mapping_page(mapping, n, NULL);
 	if (!IS_ERR(page)) {
-		kmap(page);
+		kaddr = kmap_local_page(page);
 		if (unlikely(!PageChecked(page))) {
-			if (!ufs_check_page(page))
+			if (!ufs_check_page(page, kaddr))
 				goto fail;
 		}
 		*p = page;
-		return page_address(page);
+		return kaddr;
 	}
 	return ERR_CAST(page);
 
 fail:
-	ufs_put_page(page);
+	ufs_put_page(page, kaddr);
 	return ERR_PTR(-EIO);
 }
 
@@ -227,6 +235,13 @@ ufs_next_entry(struct super_block *sb, struct
 ufs_dir_entry *p) fs16_to_cpu(sb, p->d_reclen));
 }
 
+/*
+ * Calls to ufs_get_page()/ufs_put_page() must be nested according to the
+ * rules documented in kmap_local_page()/kunmap_local().
+ *
+ * ufs_dotdot() acts as a call to ufs_get_page() and must be treated
+ * accordingly for nesting purposes.
+ */
 struct ufs_dir_entry *ufs_dotdot(struct inode *dir, struct page **p)
 {
 	struct ufs_dir_entry *de = ufs_get_page(dir, 0, p);

fs/ufs/dir.c struct ufs_dir_entry *ufs_dotdot(struct inode *dir, · line 256

Quoted 5 lines
fs/ufs/dir.c +1 -1
@@ -238,12 +253,15 @@ struct ufs_dir_entry *ufs_dotdot(struct inode *dir,
 struct page **p) }
 
 /*
- *	ufs_find_entry()
+ * Finds an entry in the specified directory with the wanted name. It 

returns

Quoted 1 line

a + * pointer to the directory's entry. The page in which the entry was

found

fs/ufs/dir.c ufs_readdir(struct file *file, struct dir_context

Quoted 42 lines

is + * in the res_page out parameter. The page is returned mapped and unlocked. + * The entry is guaranteed to be valid.
*
- * finds an entry in the specified directory with the wanted name. It
- * returns the page in which the entry was found, and the entry itself
- * (as a parameter - res_dir). Page is returned mapped and unlocked.
- * Entry is guaranteed to be valid.
+ * On Success ufs_put_page() should be called on *res_page.
+ *
+ * ufs_find_entry() acts as a call to ufs_get_page() and must be treated
+ * accordingly for nesting purposes.
*/
struct ufs_dir_entry *ufs_find_entry(struct inode *dir, const struct qstr
*qstr, struct page **res_page)

fs/ufs/dir.c +3 -3
@@ -282,7 +300,7 @@ struct ufs_dir_entry *ufs_find_entry(struct inode *dir,
 const struct qstr *qstr, goto found;
 				de = ufs_next_entry(sb, de);
 			}
-			ufs_put_page(page);
+			ufs_put_page(page, kaddr);
 		}
 		if (++n >= npages)
 			n = 0;
@@ -360,7 +378,7 @@ int ufs_add_link(struct dentry *dentry, struct inode
 *inode) de = (struct ufs_dir_entry *) ((char *) de + rec_len);
 		}
 		unlock_page(page);
-		ufs_put_page(page);
+		ufs_put_page(page, kaddr);
 	}
 	BUG();
 	return -EINVAL;
@@ -390,7 +408,7 @@ int ufs_add_link(struct dentry *dentry, struct inode
 *inode) mark_inode_dirty(dir);
 	/* OFFSET_CACHE */
 out_put:
-	ufs_put_page(page);
+	ufs_put_page(page, kaddr);
 	return err;
 out_unlock:
 	unlock_page(page);
@@ -468,13 +486,13 @@ ufs_readdir(struct file *file, struct dir_context 

*ctx)

fs/ufs/dir.c ufs_readdir(struct file *file, struct dir_context · line 486

Quoted 1 line
fs/ufs/dir.c +0 -0
 					       ufs_get_de_namlen(sb, 

de),

fs/ufs/dir.c ufs_readdir(struct file *file, struct dir_context

Quoted 16 lines
fs/ufs/dir.c +2 -2
 					       fs32_to_cpu(sb, de-
 d_ino),
 					       d_type)) {
-					ufs_put_page(page);
+					ufs_put_page(page, kaddr);
 					return 0;
 				}
 			}
 			ctx->pos += fs16_to_cpu(sb, de->d_reclen);
 		}
-		ufs_put_page(page);
+		ufs_put_page(page, kaddr);
 	}
 	return 0;

}

fs/ufs/dir.c +0 -0
@@ -485,10 +503,10 @@ ufs_readdir(struct file *file, struct dir_context 

*ctx)

Quoted 70 lines
fs/ufs/dir.c +2 -2
  * previous entry.
  */
 int ufs_delete_entry(struct inode *inode, struct ufs_dir_entry *dir,
-		     struct page * page)
+		     struct page *page)
 {
 	struct super_block *sb = inode->i_sb;
-	char *kaddr = page_address(page);
+	char *kaddr = (char *)((unsigned long)dir & PAGE_MASK);
 	unsigned int from = offset_in_page(dir) & ~(UFS_SB(sb)->s_uspi-

s_dirblksize
- 1); unsigned int to = offset_in_page(dir) + fs16_to_cpu(sb, dir- d_reclen);
loff_t pos;

fs/ufs/dir.c +6 -8
@@ -527,7 +545,7 @@ int ufs_delete_entry(struct inode *inode, struct
 ufs_dir_entry *dir, inode->i_ctime = inode->i_mtime = current_time(inode);
 	mark_inode_dirty(inode);
 out:
-	ufs_put_page(page);
+	ufs_put_page(page, kaddr);
 	UFSD("EXIT\n");
 	return err;
 }
@@ -551,8 +569,7 @@ int ufs_make_empty(struct inode * inode, struct inode
 *dir) goto fail;
 	}
 
-	kmap(page);
-	base = (char*)page_address(page);
+	base = kmap_local_page(page);
 	memset(base, 0, PAGE_SIZE);
 
 	de = (struct ufs_dir_entry *) base;
@@ -569,7 +586,7 @@ int ufs_make_empty(struct inode * inode, struct inode
 *dir) de->d_reclen = cpu_to_fs16(sb, chunk_size - UFS_DIR_REC_LEN(1));
 	ufs_set_de_namlen(sb, de, 2);
 	strcpy (de->d_name, "..");
-	kunmap(page);
+	kunmap_local(base);
 
 	err = ufs_commit_chunk(page, 0, chunk_size);
 fail:
@@ -585,9 +602,9 @@ int ufs_empty_dir(struct inode * inode)
 	struct super_block *sb = inode->i_sb;
 	struct page *page = NULL;
 	unsigned long i, npages = dir_pages(inode);
+	char *kaddr;
 
 	for (i = 0; i < npages; i++) {
-		char *kaddr;
 		struct ufs_dir_entry *de;
 
 		kaddr = ufs_get_page(inode, i, &page);
@@ -620,12 +637,12 @@ int ufs_empty_dir(struct inode * inode)
 			}
 			de = ufs_next_entry(sb, de);
 		}
-		ufs_put_page(page);
+		ufs_put_page(page, kaddr);
 	}
 	return 1;
 
 not_empty:
-	ufs_put_page(page);
+	ufs_put_page(page, kaddr);
 	return 0;
 }
 
--

2.39.0

#5 lore raw
In reply to Fabio M. De Francesco
Quoted 73 lines · On sabato 17 dicembre 2022 19:47:46 CET Fabio M. De Francesco wrote:

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

There are two main problems with kmap(): (1) It comes with an overhead as the 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 fs/ufs is safe everywhere, it should be preferred.

Therefore, replace kmap() with kmap_local_page() in fs/ufs. kunmap_local() requires the mapping address, so return that address from ufs_get_page() to be used in ufs_put_page().

This series could have not been ever made because nothing prevented the previous patch from working properly but Al Viro made a long series of very appreciated comments about how many unnecessary and redundant lines of code I could have removed. He could see things I was entirely unable to notice. Furthermore, he also provided solutions and details about how I could decompose a single patch into a small series of three independent units.[1][2][3]

I want to thank him so much for the patience, kindness and the time he decided to spend to provide those analysis and write three messages full of interesting insights.[1][2][3]

Changes from v1:
1/3: No changes.
2/3: Restore the return of "err" that was mistakenly deleted
together with the removal of the "out" label in ufs_add_link(). Thanks to Al Viro.[4]
Return the address of the kmap()'ed page instead of a pointer to a pointer to the mapped page; a page_address() had been overlooked in ufs_get_page(). Thanks to Al Viro.[5]
3/3: Return the kernel virtual address got from the call to
kmap_local_page() after conversion from kmap(). Again thanks to Al Viro.[6]

Changes from v2:
1/3: No changes.
2/3: Rework ufs_get_page() because the previous version had two
errors: (1) It could return an invalid pages with the out argument "page" and (2) it could return "page_address(page)" also in cases where read_mapping_page() returned an error and the page is never kmap()'ed. Thanks to Al Viro.[7]
3/3: Rework ufs_get_page() after conversion to
kmap_local_page(), in accordance to the last changes in 2/3.

[1] https://lore.kernel.org/lkml/Y4E++JERgUMoqfjG@ZenIV/
[2] https://lore.kernel.org/lkml/Y4FG0O7VWTTng5yh@ZenIV/
[3] https://lore.kernel.org/lkml/Y4ONIFJatIGsVNpf@ZenIV/
[4] https://lore.kernel.org/lkml/Y5Zc0qZ3+zsI74OZ@ZenIV/
[5] https://lore.kernel.org/lkml/Y5ZZy23FFAnQDR3C@ZenIV/
[6] https://lore.kernel.org/lkml/Y5ZcMPzPG9h6C9eh@ZenIV/
[7] https://lore.kernel.org/lkml/Y5glgpD7fFifC4Fi@ZenIV/#t

The cover letter of the v1 series is at
https://lore.kernel.org/lkml/[email protected]/
The cover letter of the v2 series is at
https://lore.kernel.org/lkml/[email protected]/

Fabio M. De Francesco (3):
fs/ufs: Use the offset_in_page() helper
fs/ufs: Change the signature of ufs_get_page()
fs/ufs: Replace kmap() with kmap_local_page()

Quoted 5 lines
 fs/ufs/dir.c | 134 +++++++++++++++++++++++++++------------------------
 1 file changed, 71 insertions(+), 63 deletions(-)

--
2.39.0

Please drop this v3 because of a mistake in patch 3/3. The updated v4 series is at
https://lore.kernel.org/lkml/[email protected]/T/

Thanks,

Fabio