ext4 extent split/convert refactor and kunit tests

PATCH Superseded Ojaswin Mujoo 3 participants 24 messages Jan 4 – Jan 12 ext4 file system

Share & tools

Ojaswin Mujoo

author reviewer

2026-01-04 12:19 UTC #0 lore raw

cover ext4 extent split/convert refactor and kunit tests

Offlate we've have seen multiple issues and inconsistencies in the our extent splitting and conversion logic causing subtle bugs. Recent patches by Yhang Zi [1] helped address some of the issues however the messy use of EXT4_EXT_DATA_VALID* and EXT4_EXT_MARK_UNWRIT* flags made the implementation confusing and error prone.

This patchset aims to refactor the explent split and convert code paths to make the code simpler and the behavior consistent and easy to understand. It also adds several Kunit tests to stress various permutations of extent splitting and conversion.

I've rebased this over [2] since it seems like it'll go in first.

Another idea I want to try out after this is proactively zeroout before even trying to split, as Jan suggested here [3], but before trying to do that I wanted to refactor and add some tests hence sending these patches out first.

[1] https://lore.kernel.org/linux-ext4/[email protected]/
[2] https://lore.kernel.org/linux-ext4/[email protected]/T/#t
[3] https://lore.kernel.org/linux-ext4/yro4hwpttmy6e2zspvwjfdbpej6qvhlqjvlr5kp3nwffqgcnfd@z6qual55zhfq/

Rest of the details can be found in the commit messages.

Patch 1-2: new kunit tests
Patch 3-4: minor fixes
Patch 5: refactoring zeroout and making sure zeroout handles all
permutations correctly
Patch 6: Refactoring ext4_split_* functions.
Patch 7: Enable zeroout for writ to unwrit case

Testing:
- I've run kvm-xfstests for 4k, 1k and bigalloc_4k with -g quick and I see no failures.
- Some new kunit tests that were failing due to inconsistencies are not passing
- Due to dev servers being down for eoy maintanence I couldn't run more rigourous tests. In the coming week I'll run stress and auto group as well.

Thoughts and comments are welcome.

Ojaswin Mujoo (7):
ext4: kunit tests for extent splitting and conversion
ext4: kunit tests for higher level extent manipulation functions
ext4: propagate flags to convert_initialized_extent()
ext4: propagate flags to ext4_convert_unwritten_extents_endio()
ext4: Refactor zeroout path and handle all cases
ext4: Refactor split and convert extents
ext4: Allow zeroout when doing written to unwritten split

 fs/ext4/extents-test.c   | 871 +++++++++++++++++++++++++++++++++++++++
 fs/ext4/extents.c        | 591 +++++++++++++++-----------
 fs/ext4/extents_status.c |   3 +
 fs/ext4/inode.c          |   4 +
 4 files changed, 1226 insertions(+), 243 deletions(-)
 create mode 100644 fs/ext4/extents-test.c

Signature · 1 line

2.51.0

Ojaswin Mujoo

author reviewer

2026-01-04 12:19 UTC #2 lore raw

PATCH 2/7 ext4: kunit tests for higher level extent manipulation functions

Add more kunit tests to cover all high level callers of ext4_split_convert_extents(). The main functions we cover are:

1. ext4_ext_handle_unwritten_extents()
1.1 - Split/Convert unwritten extent to written in endio convtext. 1.2 - Split/Convert unwritten extent to written in non endio context.
2. convert_initialized_extent() - Convert written extent to unwritten during zero range

Signed-off-by Ojaswin Mujoo <[email protected]>

---
 fs/ext4/extents-test.c | 275 ++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 274 insertions(+), 1 deletion(-)

fs/ext4/extents-test.c +274 -1

index 937810a0f264..4fb94d3c8a1e 100644
@@ -90,6 +90,9 @@ struct kunit_ext_test_param {
 	/* map describing range to split */
 	struct ext4_map_blocks split_map;
 
+	/* disable zeroout */
+	bool disable_zeroout;
+
 	/* no of extents expected after split */
 	int nr_exp_ext;

@@ -131,6 +134,9 @@ static struct file_system_type ext_fs_type = {
 
 static void extents_kunit_exit(struct kunit *test)
 {
+	struct ext4_sb_info *sbi = k_ctx.k_ei->vfs_inode.i_sb->s_fs_info;
+
+	kfree(sbi);
 	kfree(k_ctx.k_ei);
 	kfree(k_ctx.k_data);
 }

@@ -220,6 +226,7 @@ static int extents_kunit_init(struct kunit *test)
 	struct ext4_inode_info *ei;
 	struct inode *inode;
 	struct super_block *sb;
+	struct ext4_sb_info *sbi = NULL;
 	struct kunit_ext_test_param *param =
 		(struct kunit_ext_test_param *)(test->param_value);

@@ -237,7 +244,18 @@ static int extents_kunit_init(struct kunit *test)
 	sb->s_blocksize = 4096;
 	sb->s_blocksize_bits = 12;
 
-	ei->i_disksize = (EX_DATA_LBLK + EX_DATA_LEN + 10) << sb->s_blocksize_bits;
+	sbi = kzalloc(sizeof(struct ext4_sb_info), GFP_KERNEL);
+	if (sbi == NULL)
+		return -ENOMEM;
+
+	sbi->s_sb = sb;
+	sb->s_fs_info = sbi;
+
+	if (!param || !param->disable_zeroout)
+		sbi->s_extent_max_zeroout_kb = 32;
+
+	ei->i_disksize = (EX_DATA_LBLK + EX_DATA_LEN + 10)
+			 << sb->s_blocksize_bits;
 	inode->i_sb = sb;
 
 	k_ctx.k_data = kzalloc(EX_DATA_LEN * 4096, GFP_KERNEL);

@@ -279,6 +297,8 @@ static int extents_kunit_init(struct kunit *test)
 				   ext4_es_remove_extent_stub);
 	kunit_activate_static_stub(test, ext4_zeroout_es, ext4_zeroout_es_stub);
 	kunit_activate_static_stub(test, ext4_ext_zeroout, ext4_ext_zeroout_stub);
+	kunit_activate_static_stub(test, ext4_issue_zeroout,
+				   ext4_issue_zeroout_stub);
 	kunit_activate_static_stub(test, ext4_issue_zeroout,
 				   ext4_issue_zeroout_stub);
 	return 0;

@@ -372,6 +392,150 @@ static void test_split_convert(struct kunit *test)
 	return;
 }
 
+static void test_convert_initialized(struct kunit *test)
+{
+	struct ext4_ext_path *path;
+	struct inode *inode = &k_ctx.k_ei->vfs_inode;
+	struct ext4_extent *ex;
+	struct ext4_map_blocks map;
+	const struct kunit_ext_test_param *param =
+		(const struct kunit_ext_test_param *)(test->param_value);
+	int blkbits = inode->i_sb->s_blocksize_bits;
+	int allocated = 0;
+
+	if (param->is_zeroout_test)
+		/*
+		 * Force zeroout by making ext4_ext_insert_extent return ENOSPC
+		 */
+		kunit_activate_static_stub(test, ext4_ext_insert_extent,
+					   ext4_ext_insert_extent_stub);
+
+	path = ext4_find_extent(inode, EX_DATA_LBLK, NULL, 0);
+	ex = path->p_ext;
+	KUNIT_EXPECT_EQ(test, 10, ex->ee_block);
+	KUNIT_EXPECT_EQ(test, 3, ext4_ext_get_actual_len(ex));
+	KUNIT_EXPECT_EQ(test, param->is_unwrit_at_start, ext4_ext_is_unwritten(ex));
+	if (param->is_zeroout_test)
+		KUNIT_EXPECT_EQ(test, 0,
+				check_buffer(k_ctx.k_data, 'X',
+					     EX_DATA_LEN << blkbits));
+
+	map.m_lblk = param->split_map.m_lblk;
+	map.m_len = param->split_map.m_len;
+	convert_initialized_extent(NULL, inode, &map, path, &allocated);
+
+	path = ext4_find_extent(inode, EX_DATA_LBLK, NULL, 0);
+	ex = path->p_ext;
+
+	for (int i = 0; i < param->nr_exp_ext; i++) {
+		struct kunit_ext_state exp_ext = param->exp_ext_state[i];
+
+		KUNIT_EXPECT_EQ(test, exp_ext.ex_lblk, ex->ee_block);
+		KUNIT_EXPECT_EQ(test, exp_ext.ex_len, ext4_ext_get_actual_len(ex));
+		KUNIT_EXPECT_EQ_MSG(
+			test, exp_ext.is_unwrit, ext4_ext_is_unwritten(ex),
+			"# exp: lblk:%d len:%d unwrit:%d, got: lblk:%d len:%d unwrit:%d\n",
+			exp_ext.ex_lblk, exp_ext.ex_len, exp_ext.is_unwrit,
+			ex->ee_block, ext4_ext_get_actual_len(ex), ext4_ext_is_unwritten(ex));
+
+		ex = ex + 1;
+	}
+
+	if (!param->is_zeroout_test)
+		return;
+
+	/*
+	 * Check that then data area has been zeroed out correctly
+	 */
+	for (int i = 0; i < param->nr_exp_data_segs; i++) {
+		loff_t off, len;
+		struct kunit_ext_data_state exp_data_seg = param->exp_data_state[i];
+
+		off = exp_data_seg.off_blk << blkbits;
+		len = exp_data_seg.len_blk << blkbits;
+		KUNIT_EXPECT_EQ_MSG(test, 0,
+				    check_buffer(k_ctx.k_data + off,
+						 exp_data_seg.exp_char, len),
+				    "# corruption in byte range [%lld, %lld)",
+				    off, len);
+	}
+
+	return;
+}
+
+static void test_handle_unwritten(struct kunit *test)
+{
+	struct ext4_ext_path *path;
+	struct inode *inode = &k_ctx.k_ei->vfs_inode;
+	struct ext4_extent *ex;
+	struct ext4_map_blocks map;
+	const struct kunit_ext_test_param *param =
+		(const struct kunit_ext_test_param *)(test->param_value);
+	int blkbits = inode->i_sb->s_blocksize_bits;
+	int allocated = 0;
+	ext4_fsblk_t dummy_pblk = 999;
+
+	if (param->is_zeroout_test)
+		/*
+		 * Force zeroout by making ext4_ext_insert_extent return ENOSPC
+		 */
+		kunit_activate_static_stub(test, ext4_ext_insert_extent,
+					   ext4_ext_insert_extent_stub);
+
+	path = ext4_find_extent(inode, EX_DATA_LBLK, NULL, 0);
+	ex = path->p_ext;
+	KUNIT_EXPECT_EQ(test, 10, ex->ee_block);
+	KUNIT_EXPECT_EQ(test, 3, ext4_ext_get_actual_len(ex));
+	KUNIT_EXPECT_EQ(test, param->is_unwrit_at_start, ext4_ext_is_unwritten(ex));
+	if (param->is_zeroout_test)
+		KUNIT_EXPECT_EQ(test, 0,
+				check_buffer(k_ctx.k_data, 'X',
+					     EX_DATA_LEN << blkbits));
+
+	map.m_lblk = param->split_map.m_lblk;
+	map.m_len = param->split_map.m_len;
+	ext4_ext_handle_unwritten_extents(NULL, inode, &map, path, param->split_flags,
+					  &allocated, dummy_pblk);
+
+	path = ext4_find_extent(inode, EX_DATA_LBLK, NULL, 0);
+	ex = path->p_ext;
+
+	for (int i = 0; i < param->nr_exp_ext; i++) {
+		struct kunit_ext_state exp_ext = param->exp_ext_state[i];
+
+		KUNIT_EXPECT_EQ(test, exp_ext.ex_lblk, ex->ee_block);
+		KUNIT_EXPECT_EQ(test, exp_ext.ex_len, ext4_ext_get_actual_len(ex));
+		KUNIT_EXPECT_EQ_MSG(
+			test, exp_ext.is_unwrit, ext4_ext_is_unwritten(ex),
+			"# exp: lblk:%d len:%d unwrit:%d, got: lblk:%d len:%d unwrit:%d\n",
+			exp_ext.ex_lblk, exp_ext.ex_len, exp_ext.is_unwrit,
+			ex->ee_block, ext4_ext_get_actual_len(ex), ext4_ext_is_unwritten(ex));
+
+		ex = ex + 1;
+	}
+
+	if (!param->is_zeroout_test)
+		return;
+
+	/*
+	 * Check that then data area has been zeroed out correctly
+	 */
+	for (int i = 0; i < param->nr_exp_data_segs; i++) {
+		loff_t off, len;
+		struct kunit_ext_data_state exp_data_seg = param->exp_data_state[i];
+
+		off = exp_data_seg.off_blk << blkbits;
+		len = exp_data_seg.len_blk << blkbits;
+		KUNIT_EXPECT_EQ_MSG(test, 0,
+				    check_buffer(k_ctx.k_data + off,
+						 exp_data_seg.exp_char, len),
+				    "# corruption in byte range [%lld, %lld)",
+				    off, len);
+	}
+
+	return;
+}
+
 static const struct kunit_ext_test_param test_split_convert_params[] = {
 	/* unwrit to writ splits */
 	{ .desc = "split unwrit extent to 2 extents and convert 1st half writ",

@@ -523,6 +687,93 @@ static const struct kunit_ext_test_param test_split_convert_params[] = {
 	  .exp_data_state = { { .exp_char = 0, .off_blk = 0, .len_blk = 3 } } },
 };
 
+static const struct kunit_ext_test_param
+test_convert_initialized_params[] = {
+	/* writ to unwrit splits */
+	{ .desc = "split writ extent to 2 extents and convert 1st half unwrit",
+	  .is_unwrit_at_start = 0,
+	  .split_map = { .m_lblk = 10, .m_len = 1 },
+	  .nr_exp_ext = 2,
+	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 1, .is_unwrit = 1 },
+			     { .ex_lblk = 11, .ex_len = 2, .is_unwrit = 0 } },
+	  .is_zeroout_test = 0 },
+	{ .desc = "split writ extent to 2 extents and convert 2nd half unwrit",
+	  .is_unwrit_at_start = 0,
+	  .split_map = { .m_lblk = 11, .m_len = 2 },
+	  .nr_exp_ext = 2,
+	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 1, .is_unwrit = 0 },
+			     { .ex_lblk = 11, .ex_len = 2, .is_unwrit = 1 } },
+	  .is_zeroout_test = 0 },
+	{ .desc = "split writ extent to 3 extents and convert 2nd half to unwrit",
+	  .is_unwrit_at_start = 0,
+	  .split_map = { .m_lblk = 11, .m_len = 1 },
+	  .nr_exp_ext = 3,
+	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 1, .is_unwrit = 0 },
+			     { .ex_lblk = 11, .ex_len = 1, .is_unwrit = 1 },
+			     { .ex_lblk = 12, .ex_len = 1, .is_unwrit = 0 } },
+	  .is_zeroout_test = 0 },
+};
+
+static const struct kunit_ext_test_param test_handle_unwritten_params[] = {
+	/* unwrit to writ splits via endio path */
+	{ .desc = "split unwrit extent to 2 extents and convert 1st half writ (endio)",
+	  .is_unwrit_at_start = 1,
+	  .split_flags = EXT4_GET_BLOCKS_CONVERT,
+	  .split_map = { .m_lblk = 10, .m_len = 1 },
+	  .nr_exp_ext = 2,
+	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 1, .is_unwrit = 0 },
+			     { .ex_lblk = 11, .ex_len = 2, .is_unwrit = 1 } },
+	  .is_zeroout_test = 0 },
+	{ .desc = "split unwrit extent to 2 extents and convert 2nd half writ (endio)",
+	  .is_unwrit_at_start = 1,
+	  .split_flags = EXT4_GET_BLOCKS_CONVERT,
+	  .split_map = { .m_lblk = 11, .m_len = 2 },
+	  .nr_exp_ext = 2,
+	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 1, .is_unwrit = 1 },
+			     { .ex_lblk = 11, .ex_len = 2, .is_unwrit = 0 } },
+	  .is_zeroout_test = 0 },
+	{ .desc = "split unwrit extent to 3 extents and convert 2nd half to writ (endio)",
+	  .is_unwrit_at_start = 1,
+	  .split_flags = EXT4_GET_BLOCKS_CONVERT,
+	  .split_map = { .m_lblk = 11, .m_len = 1 },
+	  .nr_exp_ext = 3,
+	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 1, .is_unwrit = 1 },
+			     { .ex_lblk = 11, .ex_len = 1, .is_unwrit = 0 },
+			     { .ex_lblk = 12, .ex_len = 1, .is_unwrit = 1 } },
+	  .is_zeroout_test = 0 },
+
+	/* unwrit to writ splits via non-endio path */
+	{ .desc = "split unwrit extent to 2 extents and convert 1st half writ (non endio)",
+	  .is_unwrit_at_start = 1,
+	  .split_flags = EXT4_GET_BLOCKS_CREATE,
+	  .split_map = { .m_lblk = 10, .m_len = 1 },
+	  .nr_exp_ext = 2,
+	  .disable_zeroout = true,
+	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 1, .is_unwrit = 0 },
+			     { .ex_lblk = 11, .ex_len = 2, .is_unwrit = 1 } },
+	  .is_zeroout_test = 0 },
+	{ .desc = "split unwrit extent to 2 extents and convert 2nd half writ (non endio)",
+	  .is_unwrit_at_start = 1,
+	  .split_flags = EXT4_GET_BLOCKS_CREATE,
+	  .split_map = { .m_lblk = 11, .m_len = 2 },
+	  .nr_exp_ext = 2,
+	  .disable_zeroout = true,
+	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 1, .is_unwrit = 1 },
+			     { .ex_lblk = 11, .ex_len = 2, .is_unwrit = 0 } },
+	  .is_zeroout_test = 0 },
+	{ .desc = "split unwrit extent to 3 extents and convert 2nd half to writ (non endio)",
+	  .is_unwrit_at_start = 1,
+	  .split_flags = EXT4_GET_BLOCKS_CREATE,
+	  .split_map = { .m_lblk = 11, .m_len = 1 },
+	  .nr_exp_ext = 3,
+	  .disable_zeroout = true,
+	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 1, .is_unwrit = 1 },
+			     { .ex_lblk = 11, .ex_len = 1, .is_unwrit = 0 },
+			     { .ex_lblk = 12, .ex_len = 1, .is_unwrit = 1 } },
+	  .is_zeroout_test = 0 },
+
+};
+
 static void ext_get_desc(struct kunit *test, const void *p, char *desc)
 
 {

@@ -540,6 +791,24 @@ static int test_split_convert_param_init(struct kunit *test)
 	return 0;
 }
 
+static int test_convert_initialized_param_init(struct kunit *test)
+{
+	size_t arr_size = ARRAY_SIZE(test_convert_initialized_params);
+
+	kunit_register_params_array(test, test_convert_initialized_params,
+				    arr_size, ext_get_desc);
+	return 0;
+}
+
+static int test_handle_unwritten_init(struct kunit *test)
+{
+	size_t arr_size = ARRAY_SIZE(test_handle_unwritten_params);
+
+	kunit_register_params_array(test, test_handle_unwritten_params,
+				    arr_size, ext_get_desc);
+	return 0;
+}
+
 /*
  * Note that we use KUNIT_CASE_PARAM_WITH_INIT() instead of the more compact
  * KUNIT_ARRAY_PARAM() because the later currently has a limitation causing the

@@ -550,6 +819,10 @@ static int test_split_convert_param_init(struct kunit *test)
 static struct kunit_case extents_test_cases[] = {
 	KUNIT_CASE_PARAM_WITH_INIT(test_split_convert, kunit_array_gen_params,
 				   test_split_convert_param_init, NULL),
+	KUNIT_CASE_PARAM_WITH_INIT(test_convert_initialized, kunit_array_gen_params,
+				   test_convert_initialized_param_init, NULL),
+	KUNIT_CASE_PARAM_WITH_INIT(test_handle_unwritten, kunit_array_gen_params,
+				   test_handle_unwritten_init, NULL),
 	{}
 };

Signature · 1 line

2.51.0

Ojaswin Mujoo

author reviewer

2026-01-04 12:19 UTC #1 lore raw

PATCH 1/7 ext4: kunit tests for extent splitting and conversion

In reply to Ojaswin Mujoo

Add multiple KUnit tests to test various permutations of extent splitting and conversion.

We test the following cases:

1. Split of unwritten extent into 2 parts and convert 1 part to written
2. Split of unwritten extent into 3 parts and convert 1 part to written
3. Split of unwritten extent into 2 unwritten extents
4. Split of unwritten extent into 3 unwritten extents
5. Split of written extent into 2 parts and convert 1 part to unwritten
6. Split of written extent into 3 parts and convert 1 part to unwritten
7. Zeroout fallback for all the above cases except 5-6 because zeroout is not supported for written to unwritten splits

The main function we test here is ext4_split_convert_extents(). Currently some of the tests are failing due to issues in implementation. All failures are mitigated at other layers in ext4 [1] but still point out the mismatch in expectation of what the caller wants vs what the function does.

The aim is to eventually fix all the failures we see here. More detailed implementation notes can be found in the topmost commit in the test file.

[1] for example, EXT4_GET_BLOCKS_CONVERT doesn't
really convert the split extent to written, but rather the callers end up doing the conversion.

Signed-off-by Ojaswin Mujoo <[email protected]>

---
 fs/ext4/extents-test.c   | 565 +++++++++++++++++++++++++++++++++++++++
 fs/ext4/extents.c        |  22 +-
 fs/ext4/extents_status.c |   3 +
 fs/ext4/inode.c          |   4 +
 4 files changed, 592 insertions(+), 2 deletions(-)
 create mode 100644 fs/ext4/extents-test.c

fs/ext4/extents-test.c new file +565 -0

new file mode 100644
index 000000000000..937810a0f264
@@ -0,0 +1,565 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Written by Ojaswin Mujoo <[email protected]> (IBM)
+ *
+ * These Kunit tests are designed to test the functionality of
+ * extent split and conversion in ext4.
+ *
+ * Currently, ext4 can split extents in 2 ways:
+ * 1. By splitting the extents in the extent tree and optionally converting them
+ *    to written or unwritten based on flags passed.
+ * 2. In case 1 encounters an error, ext4 instead zerooes out the unwritten
+ *    areas of the extent and marks the complete extent written.
+ *
+ * The primary function that handles this is ext4_split_convert_extents().
+ *
+ * We test both of the methods of split. The behavior we try to enforce is:
+ * 1. When passing EXT4_GET_BLOCKS_CONVERT flag to ext4_split_convert_extents(),
+ *    the split extent should be converted to initialized.
+ * 2. When passing EXT4_GET_BLOCKS_CONVERT_UNWRITTEN flag to
+ *    ext4_split_convert_extents(), the split extent should be converted to
+ *    uninitialized.
+ * 3. In case we use the zeroout method, then we should correctly write zeroes
+ *    to the unwritten areas of the extent and we should not corrupt/leak any
+ *    data.
+ *
+ * Enforcing 1 and 2 is straight forward, we just setup a minimal inode with
+ * extent tree, call ext4_split_convert_extents() and check the final state of
+ * the extent tree.
+ *
+ * For zeroout testing, we maintain a separate buffer which represents the disk
+ * data corresponding to the extents. We then override ext4's zeroout functions
+ * to instead write zeroes to our buffer. Then, we override
+ * ext4_ext_insert_extent() to return -ENOSPC, which triggers the zeroout.
+ * Finally, we check the state of the extent tree and zeroout buffer to confirm
+ * everything went well.
+ */
+
+#include <kunit/test.h>
+#include <kunit/static_stub.h>
+#include <linux/gfp_types.h>
+#include <linux/stddef.h>
+
+#include "ext4.h"
+#include "ext4_extents.h"
+
+#define EX_DATA_PBLK 100
+#define EX_DATA_LBLK 10
+#define EX_DATA_LEN 3
+
+struct kunit_ctx {
+	/*
+	 * Ext4 inode which has only 1 unwrit extent
+	 */
+	struct ext4_inode_info *k_ei;
+	/*
+	 * Represents the underlying data area (used for zeroout testing)
+	 */
+	char *k_data;
+} k_ctx;
+
+/*
+ * describes the state of an expected extent in extent tree.
+ */
+struct kunit_ext_state {
+	ext4_lblk_t ex_lblk;
+	ext4_lblk_t ex_len;
+	bool is_unwrit;
+};
+
+/*
+ * describes the state of the data area of a writ extent. Used for testing
+ * correctness of zeroout.
+ */
+struct kunit_ext_data_state {
+	char exp_char;
+	ext4_lblk_t off_blk;
+	ext4_lblk_t len_blk;
+};
+
+struct kunit_ext_test_param {
+	/* description of test */
+	char *desc;
+
+	/* is extent unwrit at beginning of test */
+	bool is_unwrit_at_start;
+
+	/* flags to pass while splitting */
+	int split_flags;
+
+	/* map describing range to split */
+	struct ext4_map_blocks split_map;
+
+	/* no of extents expected after split */
+	int nr_exp_ext;
+
+	/*
+	 * expected state of extents after split. We will never split into more
+	 * than 3 extents
+	 */
+	struct kunit_ext_state exp_ext_state[3];
+
+	/* Below fields used for zeroout tests */
+
+	bool is_zeroout_test;
+	/*
+	 * no of expected data segments (zeroout tests). Example, if we expect
+	 * data to be 4kb 0s, followed by 8kb non-zero, then nr_exp_data_segs==2
+	 */
+	int nr_exp_data_segs;
+
+	/*
+	 * expected state of data area after zeroout.
+	 */
+	struct kunit_ext_data_state exp_data_state[3];
+};
+
+static void ext_kill_sb(struct super_block *sb)
+{
+	generic_shutdown_super(sb);
+}
+
+static int ext_set(struct super_block *sb, void *data)
+{
+	return 0;
+}
+
+static struct file_system_type ext_fs_type = {
+	.name = "extents test",
+	.kill_sb = ext_kill_sb,
+};
+
+static void extents_kunit_exit(struct kunit *test)
+{
+	kfree(k_ctx.k_ei);
+	kfree(k_ctx.k_data);
+}
+
+static void ext4_cache_extents_stub(struct inode *inode,
+				    struct ext4_extent_header *eh)
+{
+	return;
+}
+
+static int __ext4_ext_dirty_stub(const char *where, unsigned int line,
+				 handle_t *handle, struct inode *inode,
+				 struct ext4_ext_path *path)
+{
+	return 0;
+}
+
+static struct ext4_ext_path *
+ext4_ext_insert_extent_stub(handle_t *handle, struct inode *inode,
+			    struct ext4_ext_path *path,
+			    struct ext4_extent *newext, int gb_flags)
+{
+	return ERR_PTR(-ENOSPC);
+}
+
+static void ext4_es_remove_extent_stub(struct inode *inode, ext4_lblk_t lblk,
+				       ext4_lblk_t len)
+{
+	return;
+}
+
+static void ext4_zeroout_es_stub(struct inode *inode, struct ext4_extent *ex)
+{
+	return;
+}
+
+/*
+ * We will zeroout the equivalent range in the data area
+ */
+static int ext4_ext_zeroout_stub(struct inode *inode, struct ext4_extent *ex)
+{
+	ext4_lblk_t ee_block, off_blk;
+	loff_t ee_len;
+	loff_t off_bytes;
+	struct kunit *test = kunit_get_current_test();
+
+	ee_block = le32_to_cpu(ex->ee_block);
+	ee_len = ext4_ext_get_actual_len(ex);
+
+	KUNIT_EXPECT_EQ_MSG(test, 1, ee_block >= EX_DATA_LBLK, "ee_block=%d",
+			    ee_block);
+	KUNIT_EXPECT_EQ(test, 1,
+			ee_block + ee_len <= EX_DATA_LBLK + EX_DATA_LEN);
+
+	off_blk = ee_block - EX_DATA_LBLK;
+	off_bytes = off_blk << inode->i_sb->s_blocksize_bits;
+	memset(k_ctx.k_data + off_bytes, 0,
+	       ee_len << inode->i_sb->s_blocksize_bits);
+
+	return 0;
+}
+
+static int ext4_issue_zeroout_stub(struct inode *inode, ext4_lblk_t lblk,
+				   ext4_fsblk_t pblk, ext4_lblk_t len)
+{
+	ext4_lblk_t off_blk;
+	loff_t off_bytes;
+	struct kunit *test = kunit_get_current_test();
+
+	kunit_log(KERN_ALERT, test,
+		  "%s: lblk=%u pblk=%llu len=%u", __func__, lblk, pblk, len);
+	KUNIT_EXPECT_EQ(test, 1, lblk >= EX_DATA_LBLK);
+	KUNIT_EXPECT_EQ(test, 1, lblk + len <= EX_DATA_LBLK + EX_DATA_LEN);
+	KUNIT_EXPECT_EQ(test, 1, lblk - EX_DATA_LBLK == pblk - EX_DATA_PBLK);
+
+	off_blk = lblk - EX_DATA_LBLK;
+	off_bytes = off_blk << inode->i_sb->s_blocksize_bits;
+	memset(k_ctx.k_data + off_bytes, 0,
+	       len << inode->i_sb->s_blocksize_bits);
+
+	return 0;
+}
+
+static int extents_kunit_init(struct kunit *test)
+{
+	struct ext4_extent_header *eh = NULL;
+	struct ext4_inode_info *ei;
+	struct inode *inode;
+	struct super_block *sb;
+	struct kunit_ext_test_param *param =
+		(struct kunit_ext_test_param *)(test->param_value);
+
+	/* setup the mock inode */
+	k_ctx.k_ei = kzalloc(sizeof(struct ext4_inode_info), GFP_KERNEL);
+	if (k_ctx.k_ei == NULL)
+		return -ENOMEM;
+	ei = k_ctx.k_ei;
+	inode = &ei->vfs_inode;
+
+	sb = sget(&ext_fs_type, NULL, ext_set, 0, NULL);
+	if (IS_ERR(sb))
+		return PTR_ERR(sb);
+
+	sb->s_blocksize = 4096;
+	sb->s_blocksize_bits = 12;
+
+	ei->i_disksize = (EX_DATA_LBLK + EX_DATA_LEN + 10) << sb->s_blocksize_bits;
+	inode->i_sb = sb;
+
+	k_ctx.k_data = kzalloc(EX_DATA_LEN * 4096, GFP_KERNEL);
+	if (k_ctx.k_data == NULL)
+		return -ENOMEM;
+
+	/*
+	 * set the data area to a junk value
+	 */
+	memset(k_ctx.k_data, 'X', EX_DATA_LEN * 4096);
+
+	/* create a tree with depth 0 */
+	eh = (struct ext4_extent_header *)k_ctx.k_ei->i_data;
+
+	/* Fill extent header */
+	eh = ext_inode_hdr(&k_ctx.k_ei->vfs_inode);
+	eh->eh_depth = 0;
+	eh->eh_entries = cpu_to_le16(1);
+	eh->eh_magic = EXT4_EXT_MAGIC;
+	eh->eh_max =
+		cpu_to_le16(ext4_ext_space_root_idx(&k_ctx.k_ei->vfs_inode, 0));
+	eh->eh_generation = 0;
+
+	/*
+	 * add 1 extent in leaf node covering lblks [10,13) and pblk [100,103)
+	 */
+	EXT_FIRST_EXTENT(eh)->ee_block = cpu_to_le32(EX_DATA_LBLK);
+	EXT_FIRST_EXTENT(eh)->ee_len = cpu_to_le16(EX_DATA_LEN);
+	ext4_ext_store_pblock(EXT_FIRST_EXTENT(eh), EX_DATA_PBLK);
+	if (!param || param->is_unwrit_at_start)
+		ext4_ext_mark_unwritten(EXT_FIRST_EXTENT(eh));
+
+	/* Add stubs */
+	kunit_activate_static_stub(test, ext4_cache_extents,
+				   ext4_cache_extents_stub);
+	kunit_activate_static_stub(test, __ext4_ext_dirty,
+				   __ext4_ext_dirty_stub);
+	kunit_activate_static_stub(test, ext4_es_remove_extent,
+				   ext4_es_remove_extent_stub);
+	kunit_activate_static_stub(test, ext4_zeroout_es, ext4_zeroout_es_stub);
+	kunit_activate_static_stub(test, ext4_ext_zeroout, ext4_ext_zeroout_stub);
+	kunit_activate_static_stub(test, ext4_issue_zeroout,
+				   ext4_issue_zeroout_stub);
+	return 0;
+}
+
+/*
+ * Return 1 if all bytes in the buf equal to c, else return the offset of first mismatch
+ */
+static int check_buffer(char *buf, int c, int size)
+{
+	void *ret = NULL;
+
+	ret = memchr_inv(buf, c, size);
+	if (ret  == NULL)
+		return 0;
+
+	kunit_log(KERN_ALERT, kunit_get_current_test(),
+		  "# %s: wrong char found at offset %ld (expected:%d got:%d)", __func__,
+		  ((char *)ret - buf), c, *((char *)ret));
+	return 1;
+}
+
+static void test_split_convert(struct kunit *test)
+{
+	struct ext4_ext_path *path;
+	struct inode *inode = &k_ctx.k_ei->vfs_inode;
+	struct ext4_extent *ex;
+	struct ext4_map_blocks map;
+	const struct kunit_ext_test_param *param =
+		(const struct kunit_ext_test_param *)(test->param_value);
+	int blkbits = inode->i_sb->s_blocksize_bits;
+
+	if (param->is_zeroout_test)
+		/*
+		 * Force zeroout by making ext4_ext_insert_extent return ENOSPC
+		 */
+		kunit_activate_static_stub(test, ext4_ext_insert_extent,
+					   ext4_ext_insert_extent_stub);
+
+	path = ext4_find_extent(inode, EX_DATA_LBLK, NULL, 0);
+	ex = path->p_ext;
+	KUNIT_EXPECT_EQ(test, 10, ex->ee_block);
+	KUNIT_EXPECT_EQ(test, 3, ext4_ext_get_actual_len(ex));
+	KUNIT_EXPECT_EQ(test, param->is_unwrit_at_start, ext4_ext_is_unwritten(ex));
+	if (param->is_zeroout_test)
+		KUNIT_EXPECT_EQ(test, 0,
+				check_buffer(k_ctx.k_data, 'X',
+					     EX_DATA_LEN << blkbits));
+
+	map.m_lblk = param->split_map.m_lblk;
+	map.m_len = param->split_map.m_len;
+	ext4_split_convert_extents(NULL, inode, &map, path,
+				   param->split_flags, NULL);
+
+	path = ext4_find_extent(inode, EX_DATA_LBLK, NULL, 0);
+	ex = path->p_ext;
+
+	for (int i = 0; i < param->nr_exp_ext; i++) {
+		struct kunit_ext_state exp_ext = param->exp_ext_state[i];
+
+		KUNIT_EXPECT_EQ(test, exp_ext.ex_lblk, ex->ee_block);
+		KUNIT_EXPECT_EQ(test, exp_ext.ex_len, ext4_ext_get_actual_len(ex));
+		KUNIT_EXPECT_EQ_MSG(
+			test, exp_ext.is_unwrit, ext4_ext_is_unwritten(ex),
+			"# exp: lblk:%d len:%d unwrit:%d, got: lblk:%d len:%d unwrit:%d\n",
+			exp_ext.ex_lblk, exp_ext.ex_len, exp_ext.is_unwrit,
+			ex->ee_block, ext4_ext_get_actual_len(ex), ext4_ext_is_unwritten(ex));
+
+		ex = ex + 1;
+	}
+
+	if (!param->is_zeroout_test)
+		return;
+
+	/*
+	 * Check that then data area has been zeroed out correctly
+	 */
+	for (int i = 0; i < param->nr_exp_data_segs; i++) {
+		loff_t off, len;
+		struct kunit_ext_data_state exp_data_seg = param->exp_data_state[i];
+
+		off = exp_data_seg.off_blk << blkbits;
+		len = exp_data_seg.len_blk << blkbits;
+		KUNIT_EXPECT_EQ_MSG(test, 0,
+				    check_buffer(k_ctx.k_data + off,
+						 exp_data_seg.exp_char, len),
+				    "# corruption in byte range [%lld, %lld)",
+				    off, len);
+	}
+
+	return;
+}
+
+static const struct kunit_ext_test_param test_split_convert_params[] = {
+	/* unwrit to writ splits */
+	{ .desc = "split unwrit extent to 2 extents and convert 1st half writ",
+	  .is_unwrit_at_start = 1,
+	  .split_flags = EXT4_GET_BLOCKS_CONVERT,
+	  .split_map = { .m_lblk = 10, .m_len = 1 },
+	  .nr_exp_ext = 2,
+	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 1, .is_unwrit = 0 },
+			     { .ex_lblk = 11, .ex_len = 2, .is_unwrit = 1 } },
+	  .is_zeroout_test = 0 },
+	{ .desc = "split unwrit extent to 2 extents and convert 2nd half writ",
+	  .is_unwrit_at_start = 1,
+	  .split_flags = EXT4_GET_BLOCKS_CONVERT,
+	  .split_map = { .m_lblk = 11, .m_len = 2 },
+	  .nr_exp_ext = 2,
+	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 1, .is_unwrit = 1 },
+			     { .ex_lblk = 11, .ex_len = 2, .is_unwrit = 0 } },
+	  .is_zeroout_test = 0 },
+	{ .desc = "split unwrit extent to 3 extents and convert 2nd half to writ",
+	  .is_unwrit_at_start = 1,
+	  .split_flags = EXT4_GET_BLOCKS_CONVERT,
+	  .split_map = { .m_lblk = 11, .m_len = 1 },
+	  .nr_exp_ext = 3,
+	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 1, .is_unwrit = 1 },
+			     { .ex_lblk = 11, .ex_len = 1, .is_unwrit = 0 },
+			     { .ex_lblk = 12, .ex_len = 1, .is_unwrit = 1 } },
+	  .is_zeroout_test = 0 },
+
+	/* unwrit to unwrit splits */
+	{ .desc = "split unwrit extent to 2 unwrit extents",
+	  .is_unwrit_at_start = 1,
+	  .split_flags = EXT4_GET_BLOCKS_UNWRIT_EXT,
+	  .split_map = { .m_lblk = 10, .m_len = 1 },
+	  .nr_exp_ext = 2,
+	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 1, .is_unwrit = 1 },
+			     { .ex_lblk = 11, .ex_len = 2, .is_unwrit = 1 } },
+	  .is_zeroout_test = 0 },
+	{ .desc = "split unwrit extent to 2 extents (2)",
+	  .is_unwrit_at_start = 1,
+	  .split_flags = EXT4_GET_BLOCKS_UNWRIT_EXT,
+	  .split_map = { .m_lblk = 11, .m_len = 2 },
+	  .nr_exp_ext = 2,
+	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 1, .is_unwrit = 1 },
+			     { .ex_lblk = 11, .ex_len = 2, .is_unwrit = 1 } },
+	  .is_zeroout_test = 0 },
+	{ .desc = "split unwrit extent to 3 unwrit extents",
+	  .is_unwrit_at_start = 1,
+	  .split_flags = EXT4_GET_BLOCKS_UNWRIT_EXT,
+	  .split_map = { .m_lblk = 11, .m_len = 1 },
+	  .nr_exp_ext = 3,
+	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 1, .is_unwrit = 1 },
+			     { .ex_lblk = 11, .ex_len = 1, .is_unwrit = 1 },
+			     { .ex_lblk = 12, .ex_len = 1, .is_unwrit = 1 } },
+	  .is_zeroout_test = 0 },
+
+	/* writ to unwrit splits */
+	{ .desc = "split writ extent to 2 extents and convert 1st half unwrit",
+	  .is_unwrit_at_start = 0,
+	  .split_flags = EXT4_GET_BLOCKS_CONVERT_UNWRITTEN,
+	  .split_map = { .m_lblk = 10, .m_len = 1 },
+	  .nr_exp_ext = 2,
+	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 1, .is_unwrit = 1 },
+			     { .ex_lblk = 11, .ex_len = 2, .is_unwrit = 0 } },
+	  .is_zeroout_test = 0 },
+	{ .desc = "split writ extent to 2 extents and convert 2nd half unwrit",
+	  .is_unwrit_at_start = 0,
+	  .split_flags = EXT4_GET_BLOCKS_CONVERT_UNWRITTEN,
+	  .split_map = { .m_lblk = 11, .m_len = 2 },
+	  .nr_exp_ext = 2,
+	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 1, .is_unwrit = 0 },
+			     { .ex_lblk = 11, .ex_len = 2, .is_unwrit = 1 } },
+	  .is_zeroout_test = 0 },
+	{ .desc = "split writ extent to 3 extents and convert 2nd half to unwrit",
+	  .is_unwrit_at_start = 0,
+	  .split_flags = EXT4_GET_BLOCKS_CONVERT_UNWRITTEN,
+	  .split_map = { .m_lblk = 11, .m_len = 1 },
+	  .nr_exp_ext = 3,
+	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 1, .is_unwrit = 0 },
+			     { .ex_lblk = 11, .ex_len = 1, .is_unwrit = 1 },
+			     { .ex_lblk = 12, .ex_len = 1, .is_unwrit = 0 } },
+	  .is_zeroout_test = 0 },
+
+	/*
+	 * ***** zeroout tests *****
+	 */
+	/* unwrit to writ splits */
+	{ .desc = "split unwrit extent to 2 extents and convert 1st half writ (zeroout)",
+	  .is_unwrit_at_start = 1,
+	  .split_flags = EXT4_GET_BLOCKS_CONVERT,
+	  .split_map = { .m_lblk = 10, .m_len = 1 },
+	  .nr_exp_ext = 1,
+	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 3, .is_unwrit = 0 } },
+	  .is_zeroout_test = 1,
+	  .nr_exp_data_segs = 2,
+	  /* 1 block of data followed by 2 blocks of zeroes */
+	  .exp_data_state = { { .exp_char = 'X', .off_blk = 0, .len_blk = 1 },
+			      { .exp_char = 0, .off_blk = 1, .len_blk = 2 } } },
+	{ .desc = "split unwrit extent to 2 extents and convert 2nd half writ (zeroout)",
+	  .is_unwrit_at_start = 1,
+	  .split_flags = EXT4_GET_BLOCKS_CONVERT,
+	  .split_map = { .m_lblk = 11, .m_len = 2 },
+	  .nr_exp_ext = 1,
+	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 3, .is_unwrit = 0 } },
+	  .is_zeroout_test = 1,
+	  .nr_exp_data_segs = 2,
+	  /* 1 block of zeroes followed by 2 blocks of data */
+	  .exp_data_state = { { .exp_char = 0, .off_blk = 0, .len_blk = 1 },
+			      { .exp_char = 'X', .off_blk = 1, .len_blk = 2 } } },
+	{ .desc = "split unwrit extent to 3 extents and convert 2nd half writ (zeroout)",
+	  .is_unwrit_at_start = 1,
+	  .split_flags = EXT4_GET_BLOCKS_CONVERT,
+	  .split_map = { .m_lblk = 11, .m_len = 1 },
+	  .nr_exp_ext = 1,
+	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 3, .is_unwrit = 0 } },
+	  .is_zeroout_test = 1,
+	  .nr_exp_data_segs = 3,
+	  /* [zeroes] [data] [zeroes] */
+	  .exp_data_state = { { .exp_char = 0, .off_blk = 0, .len_blk = 1 },
+			      { .exp_char = 'X', .off_blk = 1, .len_blk = 1 },
+			      { .exp_char = 0, .off_blk = 2, .len_blk = 1 } } },
+
+	/* unwrit to unwrit splits */
+	{ .desc = "split unwrit extent to 2 unwrit extents (zeroout)",
+	  .is_unwrit_at_start = 1,
+	  .split_flags = EXT4_GET_BLOCKS_UNWRIT_EXT,
+	  .split_map = { .m_lblk = 10, .m_len = 1 },
+	  .nr_exp_ext = 1,
+	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 3, .is_unwrit = 0 } },
+	  .is_zeroout_test = 1,
+	  .nr_exp_data_segs = 1,
+	  .exp_data_state = { { .exp_char = 0, .off_blk = 0, .len_blk = 3 } } },
+	{ .desc = "split unwrit extent to 2 unwrit extents (2) (zeroout)",
+	  .is_unwrit_at_start = 1,
+	  .split_flags = EXT4_GET_BLOCKS_UNWRIT_EXT,
+	  .split_map = { .m_lblk = 11, .m_len = 2 },
+	  .nr_exp_ext = 1,
+	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 3, .is_unwrit = 0 } },
+	  .is_zeroout_test = 1,
+	  .nr_exp_data_segs = 1,
+	  .exp_data_state = { { .exp_char = 0, .off_blk = 0, .len_blk = 3 } } },
+	{ .desc = "split unwrit extent to 3 unwrit extents (zeroout)",
+	  .is_unwrit_at_start = 1,
+	  .split_flags = EXT4_GET_BLOCKS_UNWRIT_EXT,
+	  .split_map = { .m_lblk = 11, .m_len = 1 },
+	  .nr_exp_ext = 1,
+	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 3, .is_unwrit = 0 } },
+	  .is_zeroout_test = 1,
+	  .nr_exp_data_segs = 1,
+	  .exp_data_state = { { .exp_char = 0, .off_blk = 0, .len_blk = 3 } } },
+};
+
+static void ext_get_desc(struct kunit *test, const void *p, char *desc)
+
+{
+	struct kunit_ext_test_param *param = (struct kunit_ext_test_param *)p;
+
+	snprintf(desc, KUNIT_PARAM_DESC_SIZE, "%s\n", param->desc);
+}
+
+static int test_split_convert_param_init(struct kunit *test)
+{
+	size_t arr_size = ARRAY_SIZE(test_split_convert_params);
+
+	kunit_register_params_array(test, test_split_convert_params, arr_size,
+				    ext_get_desc);
+	return 0;
+}
+
+/*
+ * Note that we use KUNIT_CASE_PARAM_WITH_INIT() instead of the more compact
+ * KUNIT_ARRAY_PARAM() because the later currently has a limitation causing the
+ * output parsing to be prone to error. For more context:
+ *
+ * https://lore.kernel.org/linux-kselftest/[email protected]/
+ */
+static struct kunit_case extents_test_cases[] = {
+	KUNIT_CASE_PARAM_WITH_INIT(test_split_convert, kunit_array_gen_params,
+				   test_split_convert_param_init, NULL),
+	{}
+};
+
+static struct kunit_suite extents_test_suite = {
+	.name = "ext4_extents_test",
+	.init = extents_kunit_init,
+	.exit = extents_kunit_exit,
+	.test_cases = extents_test_cases,
+};
+
+kunit_test_suites(&extents_test_suite);
+
+MODULE_LICENSE("GPL");

fs/ext4/extents.c +20 -2

index c7c66ab825e7..0ad0a9f2e3d4 100644
@@ -32,6 +32,7 @@
 #include "ext4_jbd2.h"
 #include "ext4_extents.h"
 #include "xattr.h"
+#include <kunit/static_stub.h>
 
 #include <trace/events/ext4.h>

@@ -197,6 +198,9 @@ static int __ext4_ext_dirty(const char *where, unsigned int line,
 {
 	int err;
 
+	KUNIT_STATIC_STUB_REDIRECT(__ext4_ext_dirty, where, line, handle, inode,
+				   path);
+
 	WARN_ON(!rwsem_is_locked(&EXT4_I(inode)->i_data_sem));
 	if (path->p_bh) {
 		ext4_extent_block_csum_set(inode, ext_block_hdr(path->p_bh));

@@ -535,6 +539,8 @@ static void ext4_cache_extents(struct inode *inode,
 	ext4_lblk_t prev = 0;
 	int i;
 
+	KUNIT_STATIC_STUB_REDIRECT(ext4_cache_extents, inode, eh);
+
 	for (i = le16_to_cpu(eh->eh_entries); i > 0; i--, ex++) {
 		unsigned int status = EXTENT_STATUS_WRITTEN;
 		ext4_lblk_t lblk = le32_to_cpu(ex->ee_block);

@@ -898,6 +904,8 @@ ext4_find_extent(struct inode *inode, ext4_lblk_t block,
 	int ret;
 	gfp_t gfp_flags = GFP_NOFS;
 
+	KUNIT_STATIC_STUB_REDIRECT(ext4_find_extent, inode, block, path, flags);
+
 	if (flags & EXT4_EX_NOFAIL)
 		gfp_flags |= __GFP_NOFAIL;

@@ -1990,6 +1998,8 @@ ext4_ext_insert_extent(handle_t *handle, struct inode *inode,
 	ext4_lblk_t next;
 	int mb_flags = 0, unwritten;
 
+	KUNIT_STATIC_STUB_REDIRECT(ext4_ext_insert_extent, handle, inode, path, newext, gb_flags);
+
 	if (gb_flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE)
 		mb_flags |= EXT4_MB_DELALLOC_RESERVED;
 	if (unlikely(ext4_ext_get_actual_len(newext) == 0)) {

@@ -3138,8 +3148,10 @@ static void ext4_zeroout_es(struct inode *inode, struct ext4_extent *ex)
 	ext4_fsblk_t ee_pblock;
 	unsigned int ee_len;
 
-	ee_block  = le32_to_cpu(ex->ee_block);
-	ee_len    = ext4_ext_get_actual_len(ex);
+	KUNIT_STATIC_STUB_REDIRECT(ext4_zeroout_es, inode, ex);
+
+	ee_block = le32_to_cpu(ex->ee_block);
+	ee_len = ext4_ext_get_actual_len(ex);
 	ee_pblock = ext4_ext_pblock(ex);
 
 	if (ee_len == 0)

@@ -3155,6 +3167,8 @@ static int ext4_ext_zeroout(struct inode *inode, struct ext4_extent *ex)
 	ext4_fsblk_t ee_pblock;
 	unsigned int ee_len;
 
+	KUNIT_STATIC_STUB_REDIRECT(ext4_ext_zeroout, inode, ex);
+
 	ee_len    = ext4_ext_get_actual_len(ex);
 	ee_pblock = ext4_ext_pblock(ex);
 	return ext4_issue_zeroout(inode, le32_to_cpu(ex->ee_block), ee_pblock,

@@ -6180,3 +6194,7 @@ int ext4_ext_clear_bb(struct inode *inode)
 	ext4_free_ext_path(path);
 	return 0;
 }
+
+#ifdef CONFIG_EXT4_KUNIT_TESTS
+#include "extents-test.c"
+#endif

fs/ext4/extents_status.c +3 -0

index fc83e7e2ca9e..6c1faf7c9f2a 100644
@@ -16,6 +16,7 @@
 #include "ext4.h"
 
 #include <trace/events/ext4.h>
+#include <kunit/static_stub.h>
 
 /*
  * According to previous discussion in Ext4 Developer Workshop, we

@@ -1627,6 +1628,8 @@ void ext4_es_remove_extent(struct inode *inode, ext4_lblk_t lblk,
 	int reserved = 0;
 	struct extent_status *es = NULL;
 
+	KUNIT_STATIC_STUB_REDIRECT(ext4_es_remove_extent, inode, lblk, len);
+
 	if (EXT4_SB(inode->i_sb)->s_mount_state & EXT4_FC_REPLAY)
 		return;

fs/ext4/inode.c +4 -0

index 2e79b09fe2f0..c60813260f9a 100644
@@ -48,6 +48,8 @@
 #include "acl.h"
 #include "truncate.h"
 
+#include <kunit/static_stub.h>
+
 #include <trace/events/ext4.h>
 
 static void ext4_journalled_zero_new_buffers(handle_t *handle,

@@ -401,6 +403,8 @@ int ext4_issue_zeroout(struct inode *inode, ext4_lblk_t lblk, ext4_fsblk_t pblk,
 {
 	int ret;
 
+	KUNIT_STATIC_STUB_REDIRECT(ext4_issue_zeroout, inode, lblk, pblk, len);
+
 	if (IS_ENCRYPTED(inode) && S_ISREG(inode->i_mode))
 		return fscrypt_zeroout_range(inode, lblk, pblk, len);

Signature · 1 line

2.51.0

Ojaswin Mujoo

author reviewer

2026-01-04 12:19 UTC #3 lore raw

PATCH 3/7 ext4: propagate flags to convert_initialized_extent()

In reply to Ojaswin Mujoo

Currently, ext4_zero_range passes EXT4_EX_NOCACHE flag to avoid caching extents however this is not respected by convert_initialized_extent(). Hence, modify it to accept flags from the caller and to pass the flags on to other extent manipulation functions it calls. This makes sure the NOCACHE flag is respected throughout the code path.

Signed-off-by Ojaswin Mujoo <[email protected]>

---
 fs/ext4/extents-test.c | 2 +-
 fs/ext4/extents.c      | 5 +++--
 2 files changed, 4 insertions(+), 3 deletions(-)

fs/ext4/extents-test.c +1 -1

index 4fb94d3c8a1e..54aed3eabfe2 100644
@@ -422,7 +422,7 @@ static void test_convert_initialized(struct kunit *test)
 
 	map.m_lblk = param->split_map.m_lblk;
 	map.m_len = param->split_map.m_len;
-	convert_initialized_extent(NULL, inode, &map, path, &allocated);
+	convert_initialized_extent(NULL, inode, &map, path, 0, &allocated);
 
 	path = ext4_find_extent(inode, EX_DATA_LBLK, NULL, 0);
 	ex = path->p_ext;

fs/ext4/extents.c +3 -2

index 0ad0a9f2e3d4..5228196f5ad4 100644
@@ -3845,6 +3845,7 @@ static struct ext4_ext_path *
 convert_initialized_extent(handle_t *handle, struct inode *inode,
 			   struct ext4_map_blocks *map,
 			   struct ext4_ext_path *path,
+			   int flags,
 			   unsigned int *allocated)
 {
 	struct ext4_extent *ex;

@@ -3870,7 +3871,7 @@ convert_initialized_extent(handle_t *handle, struct inode *inode,
 
 	if (ee_block != map->m_lblk || ee_len > map->m_len) {
 		path = ext4_split_convert_extents(handle, inode, map, path,
-				EXT4_GET_BLOCKS_CONVERT_UNWRITTEN, NULL);
+				flags | EXT4_GET_BLOCKS_CONVERT_UNWRITTEN, NULL);
 		if (IS_ERR(path))
 			return path;

@@ -4264,7 +4265,7 @@ int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
 			if ((!ext4_ext_is_unwritten(ex)) &&
 			    (flags & EXT4_GET_BLOCKS_CONVERT_UNWRITTEN)) {
 				path = convert_initialized_extent(handle,
-					inode, map, path, &allocated);
+					inode, map, path, flags, &allocated);
 				if (IS_ERR(path))
 					err = PTR_ERR(path);
 				goto out;

Signature · 1 line

2.51.0

Ojaswin Mujoo

author reviewer

2026-01-04 12:19 UTC #6 lore raw

PATCH 4/7 ext4: propagate flags to ext4_convert_unwritten_extents_endio()

In reply to Ojaswin Mujoo

Currently, callers like ext4_convert_unwritten_extents() pass EXT4_EX_NOCACHE flag to avoid caching extents however this is not respected by ext4_convert_unwritten_extents_endio(). Hence, modify it to accept flags from the caller and to pass the flags on to other extent manipulation functions it calls. This makes sure the NOCACHE flag is respected throughout the code path.

Also, since the caller already passes METADATA_NOFAIL and CONVERT flags we don't need to explicitly pass it anymore.

Signed-off-by Ojaswin Mujoo <[email protected]>

---
 fs/ext4/extents.c | 7 ++-----
 1 file changed, 2 insertions(+), 5 deletions(-)

fs/ext4/extents.c +2 -5

index 5228196f5ad4..460a70e6dae0 100644
@@ -3785,7 +3785,7 @@ static struct ext4_ext_path *ext4_split_convert_extents(handle_t *handle,
 static struct ext4_ext_path *
 ext4_convert_unwritten_extents_endio(handle_t *handle, struct inode *inode,
 				     struct ext4_map_blocks *map,
-				     struct ext4_ext_path *path)
+				     struct ext4_ext_path *path, int flags)
 {
 	struct ext4_extent *ex;
 	ext4_lblk_t ee_block;

@@ -3802,9 +3802,6 @@ ext4_convert_unwritten_extents_endio(handle_t *handle, struct inode *inode,
 		  (unsigned long long)ee_block, ee_len);
 
 	if (ee_block != map->m_lblk || ee_len > map->m_len) {
-		int flags = EXT4_GET_BLOCKS_CONVERT |
-			    EXT4_GET_BLOCKS_METADATA_NOFAIL;
-
 		path = ext4_split_convert_extents(handle, inode, map, path,
 						  flags, NULL);
 		if (IS_ERR(path))

@@ -3943,7 +3940,7 @@ ext4_ext_handle_unwritten_extents(handle_t *handle, struct inode *inode,
 	/* IO end_io complete, convert the filled extent to written */
 	if (flags & EXT4_GET_BLOCKS_CONVERT) {
 		path = ext4_convert_unwritten_extents_endio(handle, inode,
-							    map, path);
+							    map, path, flags);
 		if (IS_ERR(path))
 			return path;
 		ext4_update_inode_fsync_trans(handle, inode, 1);

Signature · 1 line

2.51.0

Ojaswin Mujoo

author reviewer

2026-01-04 12:19 UTC #21 lore raw

PATCH 7/7 ext4: Allow zeroout when doing written to unwritten split

In reply to Ojaswin Mujoo

Currently, when we are doing an extent split and convert operation of written to unwritten extent (example, as done by ZERO_RANGE), we don't allow the zeroout fallback in case the extent tree manipulation fails. This is mostly because zeroout might take unsually long and the fact that this code path is more tolerant to failures than endio.

Since we have zeroout machinery in place, we might as well use it hence lift this restriction. To mitigate zeroout taking too long respect the max zeroout limit here so that the operation finishes relatively fast.

Also, add kunit tests for this case.

Signed-off-by Ojaswin Mujoo <[email protected]>

---
 fs/ext4/extents-test.c | 33 +++++++++++++++++++++++++++++++++
 fs/ext4/extents.c      | 23 +++++++++++++++--------
 2 files changed, 48 insertions(+), 8 deletions(-)

fs/ext4/extents-test.c +33 -0

index 725d5e79be96..3b5274297fe9 100644
@@ -685,6 +685,39 @@ static const struct kunit_ext_test_param test_split_convert_params[] = {
 	  .is_zeroout_test = 1,
 	  .nr_exp_data_segs = 1,
 	  .exp_data_state = { { .exp_char = 0, .off_blk = 0, .len_blk = 3 } } },
+
+	/* writ to unwrit splits */
+	{ .desc = "split writ extent to 2 extents and convert 1st half unwrit (zeroout)",
+	  .is_unwrit_at_start = 0,
+	  .split_flags = EXT4_GET_BLOCKS_CONVERT_UNWRITTEN,
+	  .split_map = { .m_lblk = 10, .m_len = 1 },
+	  .nr_exp_ext = 1,
+	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 3, .is_unwrit = 0 } },
+	  .is_zeroout_test = 1,
+	  .nr_exp_data_segs = 2,
+	  .exp_data_state = { { .exp_char = 0, .off_blk = 0, .len_blk = 1 },
+			      { .exp_char = 'X', .off_blk = 1, .len_blk = 2 }}},
+	{ .desc = "split writ extent to 2 extents and convert 2nd half unwrit (zeroout)",
+	  .is_unwrit_at_start = 0,
+	  .split_flags = EXT4_GET_BLOCKS_CONVERT_UNWRITTEN,
+	  .split_map = { .m_lblk = 11, .m_len = 2 },
+	  .nr_exp_ext = 1,
+	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 3, .is_unwrit = 0 } },
+	  .is_zeroout_test = 1,
+	  .nr_exp_data_segs = 2,
+	  .exp_data_state = { { .exp_char = 'X', .off_blk = 0, .len_blk = 1 },
+			      { .exp_char = 0, .off_blk = 1, .len_blk = 2 } } },
+	{ .desc = "split writ extent to 3 extents and convert 2nd half unwrit (zeroout)",
+	  .is_unwrit_at_start = 0,
+	  .split_flags = EXT4_GET_BLOCKS_CONVERT_UNWRITTEN,
+	  .split_map = { .m_lblk = 11, .m_len = 1 },
+	  .nr_exp_ext = 1,
+	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 3, .is_unwrit = 0 } },
+	  .is_zeroout_test = 1,
+	  .nr_exp_data_segs = 3,
+	  .exp_data_state = { { .exp_char = 'X', .off_blk = 0, .len_blk = 1 },
+			      { .exp_char = 0, .off_blk = 1, .len_blk = 1 },
+			      { .exp_char = 'X', .off_blk = 2, .len_blk = 1 }}},
 };
 
 static const struct kunit_ext_test_param

fs/ext4/extents.c +15 -8

index 9fb8a3220ae2..95dd88df8fe4 100644
@@ -3485,7 +3485,19 @@ static struct ext4_ext_path *ext4_split_extent(handle_t *handle,
 	 * to initialize as a last resort
 	 */
 	if (split_flag & EXT4_EXT_MAY_ZEROOUT) {
-		path = ext4_find_extent(inode, map->m_lblk, NULL, flags);
+		int max_zeroout_blks =
+			EXT4_SB(inode->i_sb)->s_extent_max_zeroout_kb >>
+			(inode->i_sb->s_blocksize_bits - 10);
+		if (flags & EXT4_GET_BLOCKS_CONVERT_UNWRITTEN &&
+		    map->m_len > max_zeroout_blks)
+			/*
+			 * Written to unwritten extent is not a critical path so
+			 * lets respect the max zeroout
+			 */
+			return ERR_PTR(orig_err);
+
+		path = ext4_find_extent(inode, map->m_lblk, NULL,
+						flags);
 		if (IS_ERR(path))
 			return path;

@@ -3863,15 +3875,10 @@ static struct ext4_ext_path *ext4_split_convert_extents(handle_t *handle,
 		goto convert;
 
 	/*
-	 * We don't use zeroout fallback for written to unwritten conversion as
-	 * it is not as critical as endio and it might take unusually long.
-	 * Also, it is only safe to convert extent to initialized via explicit
+	 * It is only safe to convert extent to initialized via explicit
 	 * zeroout only if extent is fully inside i_size or new_size.
 	 */
-	if (!(flags & EXT4_GET_BLOCKS_CONVERT_UNWRITTEN))
-		split_flag |= ee_block + ee_len <= eof_block ?
-				      EXT4_EXT_MAY_ZEROOUT :
-				      0;
+	split_flag |= ee_block + ee_len <= eof_block ? EXT4_EXT_MAY_ZEROOUT : 0;
 
 	/*
 	 * pass SPLIT_NOMERGE explicitly so we don't end up merging extents we

Signature · 1 line

2.51.0

Ojaswin Mujoo

author reviewer

2026-01-04 12:19 UTC #10 lore raw

PATCH 5/7 ext4: Refactor zeroout path and handle all cases

In reply to Ojaswin Mujoo

Currently, zeroout is used as a fallback in case we fail to split/convert extents in the "traditional" modify-the-extent-tree way. This is essential to mitigate failures in critical paths like extent splitting during endio. However, the logic is very messy and not easy to follow. Further, the fragile use of various flags has made it prone to errors.

Refactor zeroout out logic by moving it up to ext4_split_extents(). Further, zeroout correctly based on the type of conversion we want, ie:
- unwritten to written: Zeroout everything around the mapped range.
- unwritten to unwritten: Zeroout everything
- written to unwritten: Zeroout only the mapped range.

Signed-off-by Ojaswin Mujoo <[email protected]>

---
 fs/ext4/extents.c | 287 +++++++++++++++++++++++++++++++---------------
 1 file changed, 195 insertions(+), 92 deletions(-)

fs/ext4/extents.c +195 ~~-92~~

index 460a70e6dae0..8082e1d93bbf 100644
@@ -44,14 +44,6 @@
 #define EXT4_EXT_MARK_UNWRIT1	0x2  /* mark first half unwritten */
 #define EXT4_EXT_MARK_UNWRIT2	0x4  /* mark second half unwritten */
 
-/* first half contains valid data */
-#define EXT4_EXT_DATA_ENTIRE_VALID1	0x8   /* has entirely valid data */
-#define EXT4_EXT_DATA_PARTIAL_VALID1	0x10  /* has partially valid data */
-#define EXT4_EXT_DATA_VALID1		(EXT4_EXT_DATA_ENTIRE_VALID1 | \
-					 EXT4_EXT_DATA_PARTIAL_VALID1)
-
-#define EXT4_EXT_DATA_VALID2	0x20 /* second half contains valid data */
-
 static __le32 ext4_extent_block_csum(struct inode *inode,
 				     struct ext4_extent_header *eh)
 {

@@ -3194,7 +3186,8 @@ static int ext4_ext_zeroout(struct inode *inode, struct ext4_extent *ex)
  *  a> the extent are splitted into two extent.
  *  b> split is not needed, and just mark the extent.
  *
- * Return an extent path pointer on success, or an error pointer on failure.
+ * Return an extent path pointer on success, or an error pointer on failure. On
+ * failure, the extent is restored to original state.
  */
 static struct ext4_ext_path *ext4_split_extent_at(handle_t *handle,
 						  struct inode *inode,

@@ -3204,14 +3197,10 @@ static struct ext4_ext_path *ext4_split_extent_at(handle_t *handle,
 {
 	ext4_fsblk_t newblock;
 	ext4_lblk_t ee_block;
-	struct ext4_extent *ex, newex, orig_ex, zero_ex;
+	struct ext4_extent *ex, newex, orig_ex;
 	struct ext4_extent *ex2 = NULL;
 	unsigned int ee_len, depth;
-	int err = 0;
-
-	BUG_ON((split_flag & EXT4_EXT_DATA_VALID1) == EXT4_EXT_DATA_VALID1);
-	BUG_ON((split_flag & EXT4_EXT_DATA_VALID1) &&
-	       (split_flag & EXT4_EXT_DATA_VALID2));
+	int err = 0, insert_err = 0;
 
 	/* Do not cache extents that are in the process of being modified. */
 	flags |= EXT4_EX_NOCACHE;

@@ -3277,11 +3266,10 @@ static struct ext4_ext_path *ext4_split_extent_at(handle_t *handle,
 
 	path = ext4_ext_insert_extent(handle, inode, path, &newex, flags);
 	if (!IS_ERR(path))
-		goto out;
+		return path;
 
-	err = PTR_ERR(path);
-	if (err != -ENOSPC && err != -EDQUOT && err != -ENOMEM)
-		goto out_path;
+	insert_err = PTR_ERR(path);
+	err = 0;
 
 	/*
 	 * Get a new path to try to zeroout or fix the extent length.

@@ -3297,53 +3285,13 @@ static struct ext4_ext_path *ext4_split_extent_at(handle_t *handle,
 				 split, PTR_ERR(path));
 		goto out_path;
 	}
-	depth = ext_depth(inode);
-	ex = path[depth].p_ext;
-
-	if (EXT4_EXT_MAY_ZEROOUT & split_flag) {
-		if (split_flag & EXT4_EXT_DATA_VALID1)
-			memcpy(&zero_ex, ex2, sizeof(zero_ex));
-		else if (split_flag & EXT4_EXT_DATA_VALID2)
-			memcpy(&zero_ex, ex, sizeof(zero_ex));
-		else
-			memcpy(&zero_ex, &orig_ex, sizeof(zero_ex));
-		ext4_ext_mark_initialized(&zero_ex);
 
-		err = ext4_ext_zeroout(inode, &zero_ex);
-		if (err)
-			goto fix_extent_len;
-
-		/*
-		 * The first half contains partially valid data, the splitting
-		 * of this extent has not been completed, fix extent length
-		 * and ext4_split_extent() split will the first half again.
-		 */
-		if (split_flag & EXT4_EXT_DATA_PARTIAL_VALID1) {
-			/*
-			 * Drop extent cache to prevent stale unwritten
-			 * extents remaining after zeroing out.
-			 */
-			ext4_es_remove_extent(inode,
-					le32_to_cpu(zero_ex.ee_block),
-					ext4_ext_get_actual_len(&zero_ex));
-			goto fix_extent_len;
-		}
-
-		/* update the extent length and mark as initialized */
-		ex->ee_len = cpu_to_le16(ee_len);
-		ext4_ext_try_to_merge(handle, inode, path, ex);
-		err = ext4_ext_dirty(handle, inode, path + path->p_depth);
-		if (!err)
-			/* update extent status tree */
-			ext4_zeroout_es(inode, &zero_ex);
-		/*
-		 * If we failed at this point, we don't know in which
-		 * state the extent tree exactly is so don't try to fix
-		 * length of the original extent as it may do even more
-		 * damage.
-		 */
+	err = ext4_ext_get_access(handle, inode, path + depth);
+	if (err)
 		goto out;
-	}
+
+	depth = ext_depth(inode);
+	ex = path[depth].p_ext;
 
 fix_extent_len:
 	ex->ee_len = orig_ex.ee_len;

@@ -3353,9 +3301,9 @@ static struct ext4_ext_path *ext4_split_extent_at(handle_t *handle,
 	 */
 	ext4_ext_dirty(handle, inode, path + path->p_depth);
 out:
-	if (err) {
+	if (err || insert_err) {
 		ext4_free_ext_path(path);
-		path = ERR_PTR(err);
+		path = err ? ERR_PTR(err) : ERR_PTR(insert_err);
 	}
 out_path:
 	if (IS_ERR(path))

@@ -3365,6 +3313,115 @@ static struct ext4_ext_path *ext4_split_extent_at(handle_t *handle,
 	return path;
 }
 
+static struct ext4_ext_path *
+ext4_split_extent_zeroout(handle_t *handle, struct inode *inode,
+			  struct ext4_ext_path *path,
+			  struct ext4_map_blocks *map, int flags)
+{
+	struct ext4_extent *ex;
+	unsigned int ee_len, depth;
+	ext4_lblk_t ee_block;
+	uint64_t lblk, pblk, len;
+	int is_unwrit;
+	int err = 0;
+
+	depth = ext_depth(inode);
+	ex = path[depth].p_ext;
+	ee_block = le32_to_cpu(ex->ee_block);
+	ee_len = ext4_ext_get_actual_len(ex);
+	is_unwrit = ext4_ext_is_unwritten(ex);
+
+	if (flags & EXT4_GET_BLOCKS_CONVERT) {
+		/*
+		 * EXT4_GET_BLOCKS_CONVERT: Caller wants the range specified by
+		 * map to be initialized. Zeroout everything except the map
+		 * range.
+		 */
+
+		loff_t map_end = (loff_t) map->m_lblk + map->m_len;
+		loff_t ex_end = (loff_t) ee_block + ee_len;
+
+		if (!is_unwrit)
+			/* Shouldn't happen. Just exit */
+			return ERR_PTR(-EINVAL);
+
+		/* zeroout left */
+		if (map->m_lblk > ee_block) {
+			lblk = ee_block;
+			len = map->m_lblk - ee_block;
+			pblk = ext4_ext_pblock(ex);
+			err = ext4_issue_zeroout(inode, lblk, pblk, len);
+			if (err)
+				/* ZEROOUT failed, just return original error */
+				return ERR_PTR(err);
+		}
+
+		/* zeroout right */
+		if (map->m_lblk + map->m_len < ee_block + ee_len) {
+			lblk = map_end;
+			len = ex_end - map_end;
+			pblk = ext4_ext_pblock(ex) + (map_end - ee_block);
+			err = ext4_issue_zeroout(inode, lblk, pblk, len);
+			if (err)
+				/* ZEROOUT failed, just return original error */
+				return ERR_PTR(err);
+		}
+	} else if (flags & EXT4_GET_BLOCKS_CONVERT_UNWRITTEN) {
+		/*
+		 * EXT4_GET_BLOCKS_CONVERT_UNWRITTEN: Caller wants the
+		 * range specified by map to be marked unwritten.
+		 * Zeroout the map range leaving rest as it is.
+		 */
+
+		if (is_unwrit)
+			/* Shouldn't happen. Just exit */
+			return ERR_PTR(-EINVAL);
+
+		lblk = map->m_lblk;
+		len = map->m_len;
+		pblk = ext4_ext_pblock(ex) + (map->m_lblk - ee_block);
+		err = ext4_issue_zeroout(inode, lblk, pblk, len);
+		if (err)
+			/* ZEROOUT failed, just return original error */
+			return ERR_PTR(err);
+	} else if (flags & EXT4_GET_BLOCKS_UNWRIT_EXT) {
+		/*
+		 * EXT4_GET_BLOCKS_UNWRIT_EXT: Today, this flag
+		 * implicitly implies that callers when wanting an
+		 * unwritten to unwritten split. So zeroout the whole
+		 * extent.
+		 *
+		 * TODO: The implicit meaning of the flag is not ideal
+		 * and eventually we should aim for a more well defined
+		 * behavior
+		 */
+
+		if (!is_unwrit)
+			/* Shouldn't happen. Just exit */
+			return ERR_PTR(-EINVAL);
+
+		lblk = ee_block;
+		len = ee_len;
+		pblk = ext4_ext_pblock(ex);
+		err = ext4_issue_zeroout(inode, lblk, pblk, len);
+		if (err)
+			/* ZEROOUT failed, just return original error */
+			return ERR_PTR(err);
+	}
+
+	err = ext4_ext_get_access(handle, inode, path + depth);
+	if (err)
+		return ERR_PTR(err);
+
+	ext4_ext_mark_initialized(ex);
+
+	ext4_ext_dirty(handle, inode, path + path->p_depth);
+	if (err)
+		return ERR_PTR(err);
+
+	return 0;
+}
+
 /*
  * ext4_split_extent() splits an extent and mark extent which is covered
  * by @map as split_flags indicates

@@ -3383,11 +3440,12 @@ static struct ext4_ext_path *ext4_split_extent(handle_t *handle,
 					       int split_flag, int flags,
 					       unsigned int *allocated)
 {
-	ext4_lblk_t ee_block;
+	ext4_lblk_t ee_block, orig_ee_block;
 	struct ext4_extent *ex;
-	unsigned int ee_len, depth;
-	int unwritten;
-	int split_flag1, flags1;
+	unsigned int ee_len, orig_ee_len, depth;
+	int unwritten, orig_unwritten;
+	int split_flag1 = 0, flags1 = 0;
+	int err = 0, orig_err;
 
 	depth = ext_depth(inode);
 	ex = path[depth].p_ext;

@@ -3395,23 +3453,29 @@ static struct ext4_ext_path *ext4_split_extent(handle_t *handle,
 	ee_len = ext4_ext_get_actual_len(ex);
 	unwritten = ext4_ext_is_unwritten(ex);
 
+	orig_ee_block = ee_block;
+	orig_ee_len = ee_len;
+	orig_unwritten = unwritten;
+
 	/* Do not cache extents that are in the process of being modified. */
 	flags |= EXT4_EX_NOCACHE;
 
 	if (map->m_lblk + map->m_len < ee_block + ee_len) {
-		split_flag1 = split_flag & EXT4_EXT_MAY_ZEROOUT;
 		flags1 = flags | EXT4_GET_BLOCKS_SPLIT_NOMERGE;
 		if (unwritten)
 			split_flag1 |= EXT4_EXT_MARK_UNWRIT1 |
 				       EXT4_EXT_MARK_UNWRIT2;
-		if (split_flag & EXT4_EXT_DATA_VALID2)
-			split_flag1 |= map->m_lblk > ee_block ?
-				       EXT4_EXT_DATA_PARTIAL_VALID1 :
-				       EXT4_EXT_DATA_ENTIRE_VALID1;
 		path = ext4_split_extent_at(handle, inode, path,
 				map->m_lblk + map->m_len, split_flag1, flags1);
-		if (IS_ERR(path))
-			return path;
+
+		if (IS_ERR(path)) {
+			orig_err = PTR_ERR(path);
+			if (orig_err != -ENOSPC && orig_err != -EDQUOT &&
+			    orig_err != -ENOMEM)
+				return path;
+
+			goto try_zeroout;
+		}
 		/*
 		 * Update path is required because previous ext4_split_extent_at
 		 * may result in split of original leaf or extent zeroout.

@@ -3427,22 +3491,68 @@ static struct ext4_ext_path *ext4_split_extent(handle_t *handle,
 			ext4_free_ext_path(path);
 			return ERR_PTR(-EFSCORRUPTED);
 		}
-		unwritten = ext4_ext_is_unwritten(ex);
 	}
 
 	if (map->m_lblk >= ee_block) {
-		split_flag1 = split_flag & EXT4_EXT_DATA_VALID2;
+		split_flag1 = 0;
 		if (unwritten) {
 			split_flag1 |= EXT4_EXT_MARK_UNWRIT1;
-			split_flag1 |= split_flag & (EXT4_EXT_MAY_ZEROOUT |
-						     EXT4_EXT_MARK_UNWRIT2);
+			split_flag1 |= split_flag & EXT4_EXT_MARK_UNWRIT2;
 		}
-		path = ext4_split_extent_at(handle, inode, path,
-				map->m_lblk, split_flag1, flags);
+		path = ext4_split_extent_at(handle, inode, path, map->m_lblk,
+					    split_flag1, flags);
+
+		if (IS_ERR(path)) {
+			orig_err = PTR_ERR(path);
+			if (orig_err != -ENOSPC && orig_err != -EDQUOT &&
+			    orig_err != -ENOMEM)
+				return path;
+
+			goto try_zeroout;
+		}
+	}
+
+	if (!err)
+		goto out;
+
+try_zeroout:
+	/*
+	 * There was an error in splitting the extent, just zeroout and convert
+	 * to initialize as a last resort
+	 */
+	if (split_flag & EXT4_EXT_MAY_ZEROOUT) {
+		path = ext4_find_extent(inode, map->m_lblk, NULL, flags);
 		if (IS_ERR(path))
 			return path;
+
+		depth = ext_depth(inode);
+		ex = path[depth].p_ext;
+		ee_block = le32_to_cpu(ex->ee_block);
+		ee_len = ext4_ext_get_actual_len(ex);
+		unwritten = ext4_ext_is_unwritten(ex);
+
+		/*
+		 * The extent to zeroout should have been unchanged
+		 * but its not, just return error to caller
+		 */
+		if (WARN_ON(ee_block != orig_ee_block ||
+			    ee_len != orig_ee_len ||
+			    unwritten != orig_unwritten))
+			return ERR_PTR(orig_err);
+
+		/*
+		 * Something went wrong in zeroout, just return the
+		 * original error
+		 */
+		if (ext4_split_extent_zeroout(handle, inode, path, map, flags))
+			return ERR_PTR(orig_err);
 	}
 
+	/* There's an error and we can't zeroout, just return the err */
+	return ERR_PTR(orig_err);
+
+out:
+
 	if (allocated) {
 		if (map->m_lblk + map->m_len > ee_block + ee_len)
 			*allocated = ee_len - (map->m_lblk - ee_block);

@@ -3486,7 +3596,7 @@ ext4_ext_convert_to_initialized(handle_t *handle, struct inode *inode,
 	ext4_lblk_t ee_block, eof_block;
 	unsigned int ee_len, depth, map_len = map->m_len;
 	int err = 0;
-	int split_flag = EXT4_EXT_DATA_VALID2;
+	int split_flag = 0;
 	unsigned int max_zeroout = 0;
 
 	ext_debug(inode, "logical block %llu, max_blocks %u\n",

@@ -3760,11 +3870,7 @@ static struct ext4_ext_path *ext4_split_convert_extents(handle_t *handle,
 	ee_block = le32_to_cpu(ex->ee_block);
 	ee_len = ext4_ext_get_actual_len(ex);
 
-	/* Convert to unwritten */
-	if (flags & EXT4_GET_BLOCKS_CONVERT_UNWRITTEN) {
-		split_flag |= EXT4_EXT_DATA_ENTIRE_VALID1;
-	/* Split the existing unwritten extent */
-	} else if (flags & (EXT4_GET_BLOCKS_UNWRIT_EXT |
+	if (flags & (EXT4_GET_BLOCKS_UNWRIT_EXT |
 			    EXT4_GET_BLOCKS_CONVERT)) {
 		/*
 		 * It is safe to convert extent to initialized via explicit

@@ -3773,9 +3879,6 @@ static struct ext4_ext_path *ext4_split_convert_extents(handle_t *handle,
 		split_flag |= ee_block + ee_len <= eof_block ?
 			      EXT4_EXT_MAY_ZEROOUT : 0;
 		split_flag |= EXT4_EXT_MARK_UNWRIT2;
-		/* Convert to initialized */
-		if (flags & EXT4_GET_BLOCKS_CONVERT)
-			split_flag |= EXT4_EXT_DATA_VALID2;
 	}
 	flags |= EXT4_GET_BLOCKS_SPLIT_NOMERGE;
 	return ext4_split_extent(handle, inode, path, map, split_flag, flags,

Signature · 1 line

2.51.0

Ojaswin Mujoo

author reviewer

2026-01-04 12:19 UTC #17 lore raw

PATCH 6/7 ext4: Refactor split and convert extents

In reply to Ojaswin Mujoo

ext4_split_convert_extents() has been historically prone to subtle bugs and inconsistent behavior due to the way all the various flags interact with the extent split and conversion process. For example, callers like ext4_convert_unwritten_extents_endio() and convert_initialized_extents() needed to open code extent conversion despite passing CONVERT or CONVERT_UNWRITTEN flags because ext4_split_convert_extents() wasn't performing the conversion.

Hence, refactor ext4_split_convert_extents() to clearly enforce the semantics of each flag. The major changes here are:

* Clearly separate the split and convert process:
* ext4_split_extent() and ext4_split_extent_at() are now only responsible to perform the split.
* ext4_split_convert_extents() is now responsible to perform extent conversion after calling ext4_split_extent() for splitting.
* This helps get rid of all the MARK_UNWRIT* flags.

* Clearly enforce the semantics of flags passed to ext4_split_convert_extents():

* EXT4_GET_BLOCKS_CONVERT: Will convert the split extent to written
* EXT4_GET_BLOCKS_CONVERT_UNWRITTEN: Will convert the split extent to unwritten
* Passing neither of the above means we only want a split.
* Modify all callers to enforce the above semantics.

* Use ext4_split_convert_extents() instead of ext4_split_extents()
* in ext4_ext_convert_to_initialized() for uniformity.

* Cleanup all callers open coding the conversion logic.
* Further, modify kuniy tests to pass flags based on the new semantics.

From an end user point of view, we should not see any changes in behavior of ext4.

Signed-off-by Ojaswin Mujoo <[email protected]>

---
 fs/ext4/extents-test.c |  12 +-
 fs/ext4/extents.c      | 299 +++++++++++++++++++----------------------
 2 files changed, 145 insertions(+), 166 deletions(-)

fs/ext4/extents-test.c +6 -6

index 54aed3eabfe2..725d5e79be96 100644
@@ -567,7 +567,7 @@ static const struct kunit_ext_test_param test_split_convert_params[] = {
 	/* unwrit to unwrit splits */
 	{ .desc = "split unwrit extent to 2 unwrit extents",
 	  .is_unwrit_at_start = 1,
-	  .split_flags = EXT4_GET_BLOCKS_UNWRIT_EXT,
+	  .split_flags = 0,
 	  .split_map = { .m_lblk = 10, .m_len = 1 },
 	  .nr_exp_ext = 2,
 	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 1, .is_unwrit = 1 },

@@ -575,7 +575,7 @@ static const struct kunit_ext_test_param test_split_convert_params[] = {
 	  .is_zeroout_test = 0 },
 	{ .desc = "split unwrit extent to 2 extents (2)",
 	  .is_unwrit_at_start = 1,
-	  .split_flags = EXT4_GET_BLOCKS_UNWRIT_EXT,
+	  .split_flags = 0,
 	  .split_map = { .m_lblk = 11, .m_len = 2 },
 	  .nr_exp_ext = 2,
 	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 1, .is_unwrit = 1 },

@@ -583,7 +583,7 @@ static const struct kunit_ext_test_param test_split_convert_params[] = {
 	  .is_zeroout_test = 0 },
 	{ .desc = "split unwrit extent to 3 unwrit extents",
 	  .is_unwrit_at_start = 1,
-	  .split_flags = EXT4_GET_BLOCKS_UNWRIT_EXT,
+	  .split_flags = 0,
 	  .split_map = { .m_lblk = 11, .m_len = 1 },
 	  .nr_exp_ext = 3,
 	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 1, .is_unwrit = 1 },

@@ -660,7 +660,7 @@ static const struct kunit_ext_test_param test_split_convert_params[] = {
 	/* unwrit to unwrit splits */
 	{ .desc = "split unwrit extent to 2 unwrit extents (zeroout)",
 	  .is_unwrit_at_start = 1,
-	  .split_flags = EXT4_GET_BLOCKS_UNWRIT_EXT,
+	  .split_flags = 0,
 	  .split_map = { .m_lblk = 10, .m_len = 1 },
 	  .nr_exp_ext = 1,
 	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 3, .is_unwrit = 0 } },

@@ -669,7 +669,7 @@ static const struct kunit_ext_test_param test_split_convert_params[] = {
 	  .exp_data_state = { { .exp_char = 0, .off_blk = 0, .len_blk = 3 } } },
 	{ .desc = "split unwrit extent to 2 unwrit extents (2) (zeroout)",
 	  .is_unwrit_at_start = 1,
-	  .split_flags = EXT4_GET_BLOCKS_UNWRIT_EXT,
+	  .split_flags = 0,
 	  .split_map = { .m_lblk = 11, .m_len = 2 },
 	  .nr_exp_ext = 1,
 	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 3, .is_unwrit = 0 } },

@@ -678,7 +678,7 @@ static const struct kunit_ext_test_param test_split_convert_params[] = {
 	  .exp_data_state = { { .exp_char = 0, .off_blk = 0, .len_blk = 3 } } },
 	{ .desc = "split unwrit extent to 3 unwrit extents (zeroout)",
 	  .is_unwrit_at_start = 1,
-	  .split_flags = EXT4_GET_BLOCKS_UNWRIT_EXT,
+	  .split_flags = 0,
 	  .split_map = { .m_lblk = 11, .m_len = 1 },
 	  .nr_exp_ext = 1,
 	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 3, .is_unwrit = 0 } },

fs/ext4/extents.c +139 ~~-160~~

493 lines not on this page · raw message

Signature · 1 line

2.51.0

Ojaswin Mujoo

author reviewer

2026-01-05 06:34 UTC #23 lore raw

In reply to Ojaswin Mujoo

Quoted 31 lines · Ojaswin Mujoo

Offlate we've have seen multiple issues and inconsistencies in the our extent splitting and conversion logic causing subtle bugs. Recent patches by Yhang Zi [1] helped address some of the issues however the messy use of EXT4_EXT_DATA_VALID* and EXT4_EXT_MARK_UNWRIT* flags made the implementation confusing and error prone.
This patchset aims to refactor the explent split and convert code paths to make the code simpler and the behavior consistent and easy to understand. It also adds several Kunit tests to stress various permutations of extent splitting and conversion.
I've rebased this over [2] since it seems like it'll go in first.
Another idea I want to try out after this is proactively zeroout before even trying to split, as Jan suggested here [3], but before trying to do that I wanted to refactor and add some tests hence sending these patches out first.
[1] https://lore.kernel.org/linux-ext4/[email protected]/
[2] https://lore.kernel.org/linux-ext4/[email protected]/T/#t
[3] https://lore.kernel.org/linux-ext4/yro4hwpttmy6e2zspvwjfdbpej6qvhlqjvlr5kp3nwffqgcnfd@z6qual55zhfq/
Rest of the details can be found in the commit messages.
Patch 1-2: new kunit tests
Patch 3-4: minor fixes
Patch 5: refactoring zeroout and making sure zeroout handles all
permutations correctly
Patch 6: Refactoring ext4_split_* functions.
Patch 7: Enable zeroout for writ to unwrit case

Quoted 5 lines

Testing:
- I've run kvm-xfstests for 4k, 1k and bigalloc_4k with -g quick and I see no failures.
- Some new kunit tests that were failing due to inconsistencies are not passing

Are *now* passing with the patches* :)

Quoted 25 lines

- Due to dev servers being down for eoy maintanence I couldn't run more rigourous tests. In the coming week I'll run stress and auto group as well.
Thoughts and comments are welcome.
Ojaswin Mujoo (7):
ext4: kunit tests for extent splitting and conversion
ext4: kunit tests for higher level extent manipulation functions
ext4: propagate flags to convert_initialized_extent()
ext4: propagate flags to ext4_convert_unwritten_extents_endio()
ext4: Refactor zeroout path and handle all cases
ext4: Refactor split and convert extents
ext4: Allow zeroout when doing written to unwritten split
 fs/ext4/extents-test.c   | 871 +++++++++++++++++++++++++++++++++++++++
 fs/ext4/extents.c        | 591 +++++++++++++++-----------
 fs/ext4/extents_status.c |   3 +
 fs/ext4/inode.c          |   4 +
 4 files changed, 1226 insertions(+), 243 deletions(-)
 create mode 100644 fs/ext4/extents-test.c
Signature · 1 line
2.51.0

Jan Kara

reviewer

2026-01-06 14:33 UTC #4 lore raw

Re: [PATCH 3/7] ext4: propagate flags to convert_initialized_extent()

In reply to Ojaswin Mujoo

Quoted 38 lines · Ojaswin Mujoo

Currently, ext4_zero_range passes EXT4_EX_NOCACHE flag to avoid caching extents however this is not respected by convert_initialized_extent(). Hence, modify it to accept flags from the caller and to pass the flags on to other extent manipulation functions it calls. This makes sure the NOCACHE flag is respected throughout the code path.
Signed-off-by Ojaswin Mujoo <[email protected]>
---
 fs/ext4/extents-test.c | 2 +-
 fs/ext4/extents.c      | 5 +++--
 2 files changed, 4 insertions(+), 3 deletions(-)
fs/ext4/extents-test.c +1 -1
index 4fb94d3c8a1e..54aed3eabfe2 100644
@@ -422,7 +422,7 @@ static void test_convert_initialized(struct kunit *test)
 
 	map.m_lblk = param->split_map.m_lblk;
 	map.m_len = param->split_map.m_len;
-	convert_initialized_extent(NULL, inode, &map, path, &allocated);
+	convert_initialized_extent(NULL, inode, &map, path, 0, &allocated);
 
 	path = ext4_find_extent(inode, EX_DATA_LBLK, NULL, 0);
 	ex = path->p_ext;
fs/ext4/extents.c +1 -0
index 0ad0a9f2e3d4..5228196f5ad4 100644
@@ -3845,6 +3845,7 @@ static struct ext4_ext_path *
 convert_initialized_extent(handle_t *handle, struct inode *inode,
 			   struct ext4_map_blocks *map,
 			   struct ext4_ext_path *path,
+			   int flags,
 			   unsigned int *allocated)
 {
 	struct ext4_extent *ex;

fs/ext4/extents.c convert_initialized_extent(handle_t *handle, struct inode *inode, · line 3874

Quoted 5 lines

fs/ext4/extents.c +1 -1

@@ -3870,7 +3871,7 @@ convert_initialized_extent(handle_t *handle, struct inode *inode,
 
 	if (ee_block != map->m_lblk || ee_len > map->m_len) {
 		path = ext4_split_convert_extents(handle, inode, map, path,
-				EXT4_GET_BLOCKS_CONVERT_UNWRITTEN, NULL);
+				flags | EXT4_GET_BLOCKS_CONVERT_UNWRITTEN, NULL);

No need to keep EXT4_GET_BLOCKS_CONVERT_UNWRITTEN here as the caller has it in flags already? Otherwise the patch looks good.

Honza

Quoted 15 lines

fs/ext4/extents.c +1 -1

 		if (IS_ERR(path))
 			return path;

@@ -4264,7 +4265,7 @@ int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
 			if ((!ext4_ext_is_unwritten(ex)) &&
 			    (flags & EXT4_GET_BLOCKS_CONVERT_UNWRITTEN)) {
 				path = convert_initialized_extent(handle,
-					inode, map, path, &allocated);
+					inode, map, path, flags, &allocated);
 				if (IS_ERR(path))
 					err = PTR_ERR(path);
 				goto out;

Signature · 1 line

2.51.0

Signature · 2 lines

Jan Kara <[email protected]>
SUSE Labs, CR

Jan Kara

reviewer Reviewed-by on v1 4/7

2026-01-06 14:34 UTC #7 lore raw

Re: [PATCH 4/7] ext4: propagate flags to ext4_convert_unwritten_extents_endio()

In reply to Ojaswin Mujoo

Quoted 11 lines · Ojaswin Mujoo

Currently, callers like ext4_convert_unwritten_extents() pass EXT4_EX_NOCACHE flag to avoid caching extents however this is not respected by ext4_convert_unwritten_extents_endio(). Hence, modify it to accept flags from the caller and to pass the flags on to other extent manipulation functions it calls. This makes sure the NOCACHE flag is respected throughout the code path.
Also, since the caller already passes METADATA_NOFAIL and CONVERT flags we don't need to explicitly pass it anymore.
Signed-off-by Ojaswin Mujoo <[email protected]>

Looks good. Feel free to add:

Reviewed-by Jan Kara <[email protected]>

Honza

Quoted 39 lines

---
 fs/ext4/extents.c | 7 ++-----
 1 file changed, 2 insertions(+), 5 deletions(-)

fs/ext4/extents.c +2 -5

index 5228196f5ad4..460a70e6dae0 100644
@@ -3785,7 +3785,7 @@ static struct ext4_ext_path *ext4_split_convert_extents(handle_t *handle,
 static struct ext4_ext_path *
 ext4_convert_unwritten_extents_endio(handle_t *handle, struct inode *inode,
 				     struct ext4_map_blocks *map,
-				     struct ext4_ext_path *path)
+				     struct ext4_ext_path *path, int flags)
 {
 	struct ext4_extent *ex;
 	ext4_lblk_t ee_block;

@@ -3802,9 +3802,6 @@ ext4_convert_unwritten_extents_endio(handle_t *handle, struct inode *inode,
 		  (unsigned long long)ee_block, ee_len);
 
 	if (ee_block != map->m_lblk || ee_len > map->m_len) {
-		int flags = EXT4_GET_BLOCKS_CONVERT |
-			    EXT4_GET_BLOCKS_METADATA_NOFAIL;
-
 		path = ext4_split_convert_extents(handle, inode, map, path,
 						  flags, NULL);
 		if (IS_ERR(path))

@@ -3943,7 +3940,7 @@ ext4_ext_handle_unwritten_extents(handle_t *handle, struct inode *inode,
 	/* IO end_io complete, convert the filled extent to written */
 	if (flags & EXT4_GET_BLOCKS_CONVERT) {
 		path = ext4_convert_unwritten_extents_endio(handle, inode,
-							    map, path);
+							    map, path, flags);
 		if (IS_ERR(path))
 			return path;
 		ext4_update_inode_fsync_trans(handle, inode, 1);

Signature · 1 line

2.51.0

Signature · 2 lines

Jan Kara <[email protected]>
SUSE Labs, CR

Jan Kara

reviewer

2026-01-06 15:31 UTC #11 lore raw

Re: [PATCH 5/7] ext4: Refactor zeroout path and handle all cases

In reply to Ojaswin Mujoo

Quoted 14 lines · Ojaswin Mujoo

Currently, zeroout is used as a fallback in case we fail to split/convert extents in the "traditional" modify-the-extent-tree way. This is essential to mitigate failures in critical paths like extent splitting during endio. However, the logic is very messy and not easy to follow. Further, the fragile use of various flags has made it prone to errors.
Refactor zeroout out logic by moving it up to ext4_split_extents(). Further, zeroout correctly based on the type of conversion we want, ie:
- unwritten to written: Zeroout everything around the mapped range.
- unwritten to unwritten: Zeroout everything
- written to unwritten: Zeroout only the mapped range.
Signed-off-by Ojaswin Mujoo <[email protected]>

...

Quoted 4 lines

fs/ext4/extents.c +0 -0

@@ -3383,11 +3440,12 @@ static struct ext4_ext_path *ext4_split_extent(handle_t *handle,
 					       int split_flag, int flags,
 					       unsigned int *allocated)
 {

fs/ext4/extents.c static struct ext4_ext_path *ext4_split_extent(handle_t *handle, · line 3448

Quoted 10 lines

fs/ext4/extents.c +5 -4

-	ext4_lblk_t ee_block;
+	ext4_lblk_t ee_block, orig_ee_block;
 	struct ext4_extent *ex;
-	unsigned int ee_len, depth;
-	int unwritten;
-	int split_flag1, flags1;
+	unsigned int ee_len, orig_ee_len, depth;
+	int unwritten, orig_unwritten;
+	int split_flag1 = 0, flags1 = 0;
+	int err = 0, orig_err;

Cannot orig_err be used uninitialized in this function? At least it isn't obvious to me some of the branches setting it is always taken.

Quoted 59 lines

fs/ext4/extents.c +19 ~~-13~~

@@ -3395,23 +3453,29 @@ static struct ext4_ext_path *ext4_split_extent(handle_t *handle,
 	ee_len = ext4_ext_get_actual_len(ex);
 	unwritten = ext4_ext_is_unwritten(ex);
 
+	orig_ee_block = ee_block;
+	orig_ee_len = ee_len;
+	orig_unwritten = unwritten;
+
 	/* Do not cache extents that are in the process of being modified. */
 	flags |= EXT4_EX_NOCACHE;
 
 	if (map->m_lblk + map->m_len < ee_block + ee_len) {
-		split_flag1 = split_flag & EXT4_EXT_MAY_ZEROOUT;
 		flags1 = flags | EXT4_GET_BLOCKS_SPLIT_NOMERGE;
 		if (unwritten)
 			split_flag1 |= EXT4_EXT_MARK_UNWRIT1 |
 				       EXT4_EXT_MARK_UNWRIT2;
-		if (split_flag & EXT4_EXT_DATA_VALID2)
-			split_flag1 |= map->m_lblk > ee_block ?
-				       EXT4_EXT_DATA_PARTIAL_VALID1 :
-				       EXT4_EXT_DATA_ENTIRE_VALID1;
 		path = ext4_split_extent_at(handle, inode, path,
 				map->m_lblk + map->m_len, split_flag1, flags1);
-		if (IS_ERR(path))
-			return path;
+
+		if (IS_ERR(path)) {
+			orig_err = PTR_ERR(path);
+			if (orig_err != -ENOSPC && orig_err != -EDQUOT &&
+			    orig_err != -ENOMEM)
+				return path;
+
+			goto try_zeroout;
+		}
 		/*
 		 * Update path is required because previous ext4_split_extent_at
 		 * may result in split of original leaf or extent zeroout.

@@ -3427,22 +3491,68 @@ static struct ext4_ext_path *ext4_split_extent(handle_t *handle,
 			ext4_free_ext_path(path);
 			return ERR_PTR(-EFSCORRUPTED);
 		}
-		unwritten = ext4_ext_is_unwritten(ex);
 	}
 
 	if (map->m_lblk >= ee_block) {
-		split_flag1 = split_flag & EXT4_EXT_DATA_VALID2;
+		split_flag1 = 0;
 		if (unwritten) {
 			split_flag1 |= EXT4_EXT_MARK_UNWRIT1;
-			split_flag1 |= split_flag & (EXT4_EXT_MAY_ZEROOUT |
-						     EXT4_EXT_MARK_UNWRIT2);
+			split_flag1 |= split_flag & EXT4_EXT_MARK_UNWRIT2;
 		}
-		path = ext4_split_extent_at(handle, inode, path,
-				map->m_lblk, split_flag1, flags);
+		path = ext4_split_extent_at(handle, inode, path, map->m_lblk,
+					    split_flag1, flags);
+
+		if (IS_ERR(path)) {

fs/ext4/extents.c static struct ext4_ext_path *ext4_split_extent(handle_t *handle, · line 3515

Quoted 10 lines

fs/ext4/extents.c +10 -0

+			orig_err = PTR_ERR(path);
+			if (orig_err != -ENOSPC && orig_err != -EDQUOT &&
+			    orig_err != -ENOMEM)
+				return path;
+
+			goto try_zeroout;
+		}
+	}
+
+	if (!err)

Nothing touches 'err' in this function...

Quoted 24 lines

fs/ext4/extents.c +22 -0

+		goto out;
+
+try_zeroout:
+	/*
+	 * There was an error in splitting the extent, just zeroout and convert
+	 * to initialize as a last resort
+	 */
+	if (split_flag & EXT4_EXT_MAY_ZEROOUT) {
+		path = ext4_find_extent(inode, map->m_lblk, NULL, flags);
 		if (IS_ERR(path))
 			return path;
+
+		depth = ext_depth(inode);
+		ex = path[depth].p_ext;
+		ee_block = le32_to_cpu(ex->ee_block);
+		ee_len = ext4_ext_get_actual_len(ex);
+		unwritten = ext4_ext_is_unwritten(ex);
+
+		/*
+		 * The extent to zeroout should have been unchanged
+		 * but its not, just return error to caller
+		 */
+		if (WARN_ON(ee_block != orig_ee_block ||
+			    ee_len != orig_ee_len ||

fs/ext4/extents.c static struct ext4_ext_path *ext4_split_extent(handle_t *handle, · line 3549

Quoted 10 lines

fs/ext4/extents.c +9 -0

+			    unwritten != orig_unwritten))
+			return ERR_PTR(orig_err);
+
+		/*
+		 * Something went wrong in zeroout, just return the
+		 * original error
+		 */
+		if (ext4_split_extent_zeroout(handle, inode, path, map, flags))
+			return ERR_PTR(orig_err);
 	}

Also nothing seems to zero out orig_err in case zero out above succeeded. What am I missing?

Honza

Quoted 9 lines

fs/ext4/extents.c +5 -0

+	/* There's an error and we can't zeroout, just return the err */
+	return ERR_PTR(orig_err);
+
+out:
+
 	if (allocated) {
 		if (map->m_lblk + map->m_len > ee_block + ee_len)
 			*allocated = ee_len - (map->m_lblk - ee_block);

Signature · 2 lines

Jan Kara <[email protected]>
SUSE Labs, CR

Jan Kara

reviewer Reviewed-by on v1 7/7

2026-01-06 15:41 UTC #22 lore raw

Re: [PATCH 7/7] ext4: Allow zeroout when doing written to unwritten split

In reply to Ojaswin Mujoo

Quoted 6 lines · Ojaswin Mujoo

Currently, when we are doing an extent split and convert operation of written to unwritten extent (example, as done by ZERO_RANGE), we don't allow the zeroout fallback in case the extent tree manipulation fails. This is mostly because zeroout might take unsually long and the fact that this code path is more tolerant to failures than endio.

Quoted 7 lines

Since we have zeroout machinery in place, we might as well use it hence lift this restriction. To mitigate zeroout taking too long respect the max zeroout limit here so that the operation finishes relatively fast.
Also, add kunit tests for this case.
Signed-off-by Ojaswin Mujoo <[email protected]>

Looks good. Feel free to add:

Reviewed-by Jan Kara <[email protected]>

Honza

Quoted 95 lines

---
 fs/ext4/extents-test.c | 33 +++++++++++++++++++++++++++++++++
 fs/ext4/extents.c      | 23 +++++++++++++++--------
 2 files changed, 48 insertions(+), 8 deletions(-)

fs/ext4/extents-test.c +33 -0

index 725d5e79be96..3b5274297fe9 100644
@@ -685,6 +685,39 @@ static const struct kunit_ext_test_param test_split_convert_params[] = {
 	  .is_zeroout_test = 1,
 	  .nr_exp_data_segs = 1,
 	  .exp_data_state = { { .exp_char = 0, .off_blk = 0, .len_blk = 3 } } },
+
+	/* writ to unwrit splits */
+	{ .desc = "split writ extent to 2 extents and convert 1st half unwrit (zeroout)",
+	  .is_unwrit_at_start = 0,
+	  .split_flags = EXT4_GET_BLOCKS_CONVERT_UNWRITTEN,
+	  .split_map = { .m_lblk = 10, .m_len = 1 },
+	  .nr_exp_ext = 1,
+	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 3, .is_unwrit = 0 } },
+	  .is_zeroout_test = 1,
+	  .nr_exp_data_segs = 2,
+	  .exp_data_state = { { .exp_char = 0, .off_blk = 0, .len_blk = 1 },
+			      { .exp_char = 'X', .off_blk = 1, .len_blk = 2 }}},
+	{ .desc = "split writ extent to 2 extents and convert 2nd half unwrit (zeroout)",
+	  .is_unwrit_at_start = 0,
+	  .split_flags = EXT4_GET_BLOCKS_CONVERT_UNWRITTEN,
+	  .split_map = { .m_lblk = 11, .m_len = 2 },
+	  .nr_exp_ext = 1,
+	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 3, .is_unwrit = 0 } },
+	  .is_zeroout_test = 1,
+	  .nr_exp_data_segs = 2,
+	  .exp_data_state = { { .exp_char = 'X', .off_blk = 0, .len_blk = 1 },
+			      { .exp_char = 0, .off_blk = 1, .len_blk = 2 } } },
+	{ .desc = "split writ extent to 3 extents and convert 2nd half unwrit (zeroout)",
+	  .is_unwrit_at_start = 0,
+	  .split_flags = EXT4_GET_BLOCKS_CONVERT_UNWRITTEN,
+	  .split_map = { .m_lblk = 11, .m_len = 1 },
+	  .nr_exp_ext = 1,
+	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 3, .is_unwrit = 0 } },
+	  .is_zeroout_test = 1,
+	  .nr_exp_data_segs = 3,
+	  .exp_data_state = { { .exp_char = 'X', .off_blk = 0, .len_blk = 1 },
+			      { .exp_char = 0, .off_blk = 1, .len_blk = 1 },
+			      { .exp_char = 'X', .off_blk = 2, .len_blk = 1 }}},
 };
 
 static const struct kunit_ext_test_param

fs/ext4/extents.c +15 -8

index 9fb8a3220ae2..95dd88df8fe4 100644
@@ -3485,7 +3485,19 @@ static struct ext4_ext_path *ext4_split_extent(handle_t *handle,
 	 * to initialize as a last resort
 	 */
 	if (split_flag & EXT4_EXT_MAY_ZEROOUT) {
-		path = ext4_find_extent(inode, map->m_lblk, NULL, flags);
+		int max_zeroout_blks =
+			EXT4_SB(inode->i_sb)->s_extent_max_zeroout_kb >>
+			(inode->i_sb->s_blocksize_bits - 10);
+		if (flags & EXT4_GET_BLOCKS_CONVERT_UNWRITTEN &&
+		    map->m_len > max_zeroout_blks)
+			/*
+			 * Written to unwritten extent is not a critical path so
+			 * lets respect the max zeroout
+			 */
+			return ERR_PTR(orig_err);
+
+		path = ext4_find_extent(inode, map->m_lblk, NULL,
+						flags);
 		if (IS_ERR(path))
 			return path;

@@ -3863,15 +3875,10 @@ static struct ext4_ext_path *ext4_split_convert_extents(handle_t *handle,
 		goto convert;
 
 	/*
-	 * We don't use zeroout fallback for written to unwritten conversion as
-	 * it is not as critical as endio and it might take unusually long.
-	 * Also, it is only safe to convert extent to initialized via explicit
+	 * It is only safe to convert extent to initialized via explicit
 	 * zeroout only if extent is fully inside i_size or new_size.
 	 */
-	if (!(flags & EXT4_GET_BLOCKS_CONVERT_UNWRITTEN))
-		split_flag |= ee_block + ee_len <= eof_block ?
-				      EXT4_EXT_MAY_ZEROOUT :
-				      0;
+	split_flag |= ee_block + ee_len <= eof_block ? EXT4_EXT_MAY_ZEROOUT : 0;
 
 	/*
 	 * pass SPLIT_NOMERGE explicitly so we don't end up merging extents we

Signature · 1 line

2.51.0

Signature · 2 lines

Jan Kara <[email protected]>
SUSE Labs, CR

Jan Kara

reviewer Reviewed-by on v1 6/7

2026-01-06 15:56 UTC #18 lore raw

Re: [PATCH 6/7] ext4: Refactor split and convert extents

In reply to Ojaswin Mujoo

Quoted 30 lines · Ojaswin Mujoo

ext4_split_convert_extents() has been historically prone to subtle bugs and inconsistent behavior due to the way all the various flags interact with the extent split and conversion process. For example, callers like ext4_convert_unwritten_extents_endio() and convert_initialized_extents() needed to open code extent conversion despite passing CONVERT or CONVERT_UNWRITTEN flags because ext4_split_convert_extents() wasn't performing the conversion.
Hence, refactor ext4_split_convert_extents() to clearly enforce the semantics of each flag. The major changes here are:
* Clearly separate the split and convert process:
* ext4_split_extent() and ext4_split_extent_at() are now only responsible to perform the split.
* ext4_split_convert_extents() is now responsible to perform extent conversion after calling ext4_split_extent() for splitting.
* This helps get rid of all the MARK_UNWRIT* flags.
* Clearly enforce the semantics of flags passed to ext4_split_convert_extents():
* EXT4_GET_BLOCKS_CONVERT: Will convert the split extent to written
* EXT4_GET_BLOCKS_CONVERT_UNWRITTEN: Will convert the split extent to unwritten
* Passing neither of the above means we only want a split.
* Modify all callers to enforce the above semantics.
* Use ext4_split_convert_extents() instead of ext4_split_extents()
* in ext4_ext_convert_to_initialized() for uniformity.

Quoted 7 lines

* Cleanup all callers open coding the conversion logic.
* Further, modify kuniy tests to pass flags based on the new semantics.
From an end user point of view, we should not see any changes in behavior of ext4.
Signed-off-by Ojaswin Mujoo <[email protected]>

Looks good. Feel free to add:

Reviewed-by Jan Kara <[email protected]>

Honza

Quoted 589 lines

---
 fs/ext4/extents-test.c |  12 +-
 fs/ext4/extents.c      | 299 +++++++++++++++++++----------------------
 2 files changed, 145 insertions(+), 166 deletions(-)

fs/ext4/extents-test.c +6 -6

index 54aed3eabfe2..725d5e79be96 100644
@@ -567,7 +567,7 @@ static const struct kunit_ext_test_param test_split_convert_params[] = {
 	/* unwrit to unwrit splits */
 	{ .desc = "split unwrit extent to 2 unwrit extents",
 	  .is_unwrit_at_start = 1,
-	  .split_flags = EXT4_GET_BLOCKS_UNWRIT_EXT,
+	  .split_flags = 0,
 	  .split_map = { .m_lblk = 10, .m_len = 1 },
 	  .nr_exp_ext = 2,
 	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 1, .is_unwrit = 1 },

@@ -575,7 +575,7 @@ static const struct kunit_ext_test_param test_split_convert_params[] = {
 	  .is_zeroout_test = 0 },
 	{ .desc = "split unwrit extent to 2 extents (2)",
 	  .is_unwrit_at_start = 1,
-	  .split_flags = EXT4_GET_BLOCKS_UNWRIT_EXT,
+	  .split_flags = 0,
 	  .split_map = { .m_lblk = 11, .m_len = 2 },
 	  .nr_exp_ext = 2,
 	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 1, .is_unwrit = 1 },

@@ -583,7 +583,7 @@ static const struct kunit_ext_test_param test_split_convert_params[] = {
 	  .is_zeroout_test = 0 },
 	{ .desc = "split unwrit extent to 3 unwrit extents",
 	  .is_unwrit_at_start = 1,
-	  .split_flags = EXT4_GET_BLOCKS_UNWRIT_EXT,
+	  .split_flags = 0,
 	  .split_map = { .m_lblk = 11, .m_len = 1 },
 	  .nr_exp_ext = 3,
 	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 1, .is_unwrit = 1 },

@@ -660,7 +660,7 @@ static const struct kunit_ext_test_param test_split_convert_params[] = {
 	/* unwrit to unwrit splits */
 	{ .desc = "split unwrit extent to 2 unwrit extents (zeroout)",
 	  .is_unwrit_at_start = 1,
-	  .split_flags = EXT4_GET_BLOCKS_UNWRIT_EXT,
+	  .split_flags = 0,
 	  .split_map = { .m_lblk = 10, .m_len = 1 },
 	  .nr_exp_ext = 1,
 	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 3, .is_unwrit = 0 } },

@@ -669,7 +669,7 @@ static const struct kunit_ext_test_param test_split_convert_params[] = {
 	  .exp_data_state = { { .exp_char = 0, .off_blk = 0, .len_blk = 3 } } },
 	{ .desc = "split unwrit extent to 2 unwrit extents (2) (zeroout)",
 	  .is_unwrit_at_start = 1,
-	  .split_flags = EXT4_GET_BLOCKS_UNWRIT_EXT,
+	  .split_flags = 0,
 	  .split_map = { .m_lblk = 11, .m_len = 2 },
 	  .nr_exp_ext = 1,
 	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 3, .is_unwrit = 0 } },

@@ -678,7 +678,7 @@ static const struct kunit_ext_test_param test_split_convert_params[] = {
 	  .exp_data_state = { { .exp_char = 0, .off_blk = 0, .len_blk = 3 } } },
 	{ .desc = "split unwrit extent to 3 unwrit extents (zeroout)",
 	  .is_unwrit_at_start = 1,
-	  .split_flags = EXT4_GET_BLOCKS_UNWRIT_EXT,
+	  .split_flags = 0,
 	  .split_map = { .m_lblk = 11, .m_len = 1 },
 	  .nr_exp_ext = 1,
 	  .exp_ext_state = { { .ex_lblk = 10, .ex_len = 3, .is_unwrit = 0 } },

fs/ext4/extents.c +139 ~~-160~~

index 8082e1d93bbf..9fb8a3220ae2 100644
@@ -41,8 +41,9 @@
  */
 #define EXT4_EXT_MAY_ZEROOUT	0x1  /* safe to zeroout if split fails \
 					due to ENOSPC */
-#define EXT4_EXT_MARK_UNWRIT1	0x2  /* mark first half unwritten */
-#define EXT4_EXT_MARK_UNWRIT2	0x4  /* mark second half unwritten */
+static struct ext4_ext_path *ext4_split_convert_extents(
+	handle_t *handle, struct inode *inode, struct ext4_map_blocks *map,
+	struct ext4_ext_path *path, int flags, unsigned int *allocated);
 
 static __le32 ext4_extent_block_csum(struct inode *inode,
 				     struct ext4_extent_header *eh)

@@ -84,8 +85,7 @@ static void ext4_extent_block_csum_set(struct inode *inode,
 static struct ext4_ext_path *ext4_split_extent_at(handle_t *handle,
 						  struct inode *inode,
 						  struct ext4_ext_path *path,
-						  ext4_lblk_t split,
-						  int split_flag, int flags);
+						  ext4_lblk_t split, int flags);
 
 static int ext4_ext_trunc_restart_fn(struct inode *inode, int *dropped)
 {

@@ -333,15 +333,12 @@ ext4_force_split_extent_at(handle_t *handle, struct inode *inode,
 			   struct ext4_ext_path *path, ext4_lblk_t lblk,
 			   int nofail)
 {
-	int unwritten = ext4_ext_is_unwritten(path[path->p_depth].p_ext);
 	int flags = EXT4_EX_NOCACHE | EXT4_GET_BLOCKS_SPLIT_NOMERGE;
 
 	if (nofail)
 		flags |= EXT4_GET_BLOCKS_METADATA_NOFAIL | EXT4_EX_NOFAIL;
 
-	return ext4_split_extent_at(handle, inode, path, lblk, unwritten ?
-			EXT4_EXT_MARK_UNWRIT1|EXT4_EXT_MARK_UNWRIT2 : 0,
-			flags);
+	return ext4_split_extent_at(handle, inode, path, lblk, flags);
 }
 
 static int

@@ -3174,17 +3171,11 @@ static int ext4_ext_zeroout(struct inode *inode, struct ext4_extent *ex)
  * @inode: the file inode
  * @path: the path to the extent
  * @split: the logical block where the extent is splitted.
- * @split_flags: indicates if the extent could be zeroout if split fails, and
- *		 the states(init or unwritten) of new extents.
  * @flags: flags used to insert new extent to extent tree.
  *
  *
  * Splits extent [a, b] into two extents [a, @split) and [@split, b], states
- * of which are determined by split_flag.
- *
- * There are two cases:
- *  a> the extent are splitted into two extent.
- *  b> split is not needed, and just mark the extent.
+ * of which are same as the original extent. No conversion is performed.
  *
  * Return an extent path pointer on success, or an error pointer on failure. On
  * failure, the extent is restored to original state.

@@ -3193,14 +3184,14 @@ static struct ext4_ext_path *ext4_split_extent_at(handle_t *handle,
 						  struct inode *inode,
 						  struct ext4_ext_path *path,
 						  ext4_lblk_t split,
-						  int split_flag, int flags)
+						  int flags)
 {
 	ext4_fsblk_t newblock;
 	ext4_lblk_t ee_block;
 	struct ext4_extent *ex, newex, orig_ex;
 	struct ext4_extent *ex2 = NULL;
 	unsigned int ee_len, depth;
-	int err = 0, insert_err = 0;
+	int err = 0, insert_err = 0, is_unwrit = 0;
 
 	/* Do not cache extents that are in the process of being modified. */
 	flags |= EXT4_EX_NOCACHE;

@@ -3214,39 +3205,24 @@ static struct ext4_ext_path *ext4_split_extent_at(handle_t *handle,
 	ee_block = le32_to_cpu(ex->ee_block);
 	ee_len = ext4_ext_get_actual_len(ex);
 	newblock = split - ee_block + ext4_ext_pblock(ex);
+	is_unwrit = ext4_ext_is_unwritten(ex);
 
 	BUG_ON(split < ee_block || split >= (ee_block + ee_len));
-	BUG_ON(!ext4_ext_is_unwritten(ex) &&
-	       split_flag & (EXT4_EXT_MAY_ZEROOUT |
-			     EXT4_EXT_MARK_UNWRIT1 |
-			     EXT4_EXT_MARK_UNWRIT2));
 
-	err = ext4_ext_get_access(handle, inode, path + depth);
-	if (err)
+	/*
+	 * No split needed
+	 */
+	if (split == ee_block)
 		goto out;
 
-	if (split == ee_block) {
-		/*
-		 * case b: block @split is the block that the extent begins with
-		 * then we just change the state of the extent, and splitting
-		 * is not needed.
-		 */
-		if (split_flag & EXT4_EXT_MARK_UNWRIT2)
-			ext4_ext_mark_unwritten(ex);
-		else
-			ext4_ext_mark_initialized(ex);
-
-		if (!(flags & EXT4_GET_BLOCKS_SPLIT_NOMERGE))
-			ext4_ext_try_to_merge(handle, inode, path, ex);
-
-		err = ext4_ext_dirty(handle, inode, path + path->p_depth);
+	err = ext4_ext_get_access(handle, inode, path + depth);
+	if (err)
 		goto out;
-	}
 
 	/* case a */
 	memcpy(&orig_ex, ex, sizeof(orig_ex));
 	ex->ee_len = cpu_to_le16(split - ee_block);
-	if (split_flag & EXT4_EXT_MARK_UNWRIT1)
+	if (is_unwrit)
 		ext4_ext_mark_unwritten(ex);
 
 	/*

@@ -3261,7 +3237,7 @@ static struct ext4_ext_path *ext4_split_extent_at(handle_t *handle,
 	ex2->ee_block = cpu_to_le32(split);
 	ex2->ee_len   = cpu_to_le16(ee_len - (split - ee_block));
 	ext4_ext_store_pblock(ex2, newblock);
-	if (split_flag & EXT4_EXT_MARK_UNWRIT2)
+	if (is_unwrit)
 		ext4_ext_mark_unwritten(ex2);
 
 	path = ext4_ext_insert_extent(handle, inode, path, &newex, flags);

@@ -3384,16 +3360,11 @@ ext4_split_extent_zeroout(handle_t *handle, struct inode *inode,
 		if (err)
 			/* ZEROOUT failed, just return original error */
 			return ERR_PTR(err);
-	} else if (flags & EXT4_GET_BLOCKS_UNWRIT_EXT) {
+	} else {
 		/*
-		 * EXT4_GET_BLOCKS_UNWRIT_EXT: Today, this flag
-		 * implicitly implies that callers when wanting an
-		 * unwritten to unwritten split. So zeroout the whole
-		 * extent.
-		 *
-		 * TODO: The implicit meaning of the flag is not ideal
-		 * and eventually we should aim for a more well defined
-		 * behavior
+		 * None of the convert flags imply we just want a split.
+		 * In this case we can only zeroout if an unwritten split
+		 * was needed.
 		 */
 
 		if (!is_unwrit)

@@ -3415,7 +3386,7 @@ ext4_split_extent_zeroout(handle_t *handle, struct inode *inode,
 
 	ext4_ext_mark_initialized(ex);
 
-	ext4_ext_dirty(handle, inode, path + path->p_depth);
+	ext4_ext_dirty(handle, inode, path + depth);
 	if (err)
 		return ERR_PTR(err);

@@ -3438,13 +3409,13 @@ static struct ext4_ext_path *ext4_split_extent(handle_t *handle,
 					       struct ext4_ext_path *path,
 					       struct ext4_map_blocks *map,
 					       int split_flag, int flags,
-					       unsigned int *allocated)
+					       unsigned int *allocated, bool *did_zeroout)
 {
 	ext4_lblk_t ee_block, orig_ee_block;
 	struct ext4_extent *ex;
 	unsigned int ee_len, orig_ee_len, depth;
 	int unwritten, orig_unwritten;
-	int split_flag1 = 0, flags1 = 0;
+	int flags1 = 0;
 	int err = 0, orig_err;
 
 	depth = ext_depth(inode);

@@ -3462,11 +3433,9 @@ static struct ext4_ext_path *ext4_split_extent(handle_t *handle,
 
 	if (map->m_lblk + map->m_len < ee_block + ee_len) {
 		flags1 = flags | EXT4_GET_BLOCKS_SPLIT_NOMERGE;
-		if (unwritten)
-			split_flag1 |= EXT4_EXT_MARK_UNWRIT1 |
-				       EXT4_EXT_MARK_UNWRIT2;
+
 		path = ext4_split_extent_at(handle, inode, path,
-				map->m_lblk + map->m_len, split_flag1, flags1);
+					    map->m_lblk + map->m_len, flags1);
 
 		if (IS_ERR(path)) {
 			orig_err = PTR_ERR(path);

@@ -3494,13 +3463,8 @@ static struct ext4_ext_path *ext4_split_extent(handle_t *handle,
 	}
 
 	if (map->m_lblk >= ee_block) {
-		split_flag1 = 0;
-		if (unwritten) {
-			split_flag1 |= EXT4_EXT_MARK_UNWRIT1;
-			split_flag1 |= split_flag & EXT4_EXT_MARK_UNWRIT2;
-		}
 		path = ext4_split_extent_at(handle, inode, path, map->m_lblk,
-					    split_flag1, flags);
+					    flags);
 
 		if (IS_ERR(path)) {
 			orig_err = PTR_ERR(path);

@@ -3546,6 +3510,11 @@ static struct ext4_ext_path *ext4_split_extent(handle_t *handle,
 		 */
 		if (ext4_split_extent_zeroout(handle, inode, path, map, flags))
 			return ERR_PTR(orig_err);
+
+		/* zeroout succeeded */
+		if (did_zeroout)
+			*did_zeroout = true;
+		goto out;
 	}
 
 	/* There's an error and we can't zeroout, just return the err */

@@ -3596,7 +3565,6 @@ ext4_ext_convert_to_initialized(handle_t *handle, struct inode *inode,
 	ext4_lblk_t ee_block, eof_block;
 	unsigned int ee_len, depth, map_len = map->m_len;
 	int err = 0;
-	int split_flag = 0;
 	unsigned int max_zeroout = 0;
 
 	ext_debug(inode, "logical block %llu, max_blocks %u\n",

@@ -3748,9 +3716,7 @@ ext4_ext_convert_to_initialized(handle_t *handle, struct inode *inode,
 	 * It is safe to convert extent to initialized via explicit
 	 * zeroout only if extent is fully inside i_size or new_size.
 	 */
-	split_flag |= ee_block + ee_len <= eof_block ? EXT4_EXT_MAY_ZEROOUT : 0;
-
-	if (EXT4_EXT_MAY_ZEROOUT & split_flag)
+	if (ee_block + ee_len <= eof_block)
 		max_zeroout = sbi->s_extent_max_zeroout_kb >>
 			(inode->i_sb->s_blocksize_bits - 10);

@@ -3805,8 +3771,8 @@ ext4_ext_convert_to_initialized(handle_t *handle, struct inode *inode,
 	}
 
 fallback:
-	path = ext4_split_extent(handle, inode, path, &split_map, split_flag,
-				 flags, NULL);
+	path = ext4_split_convert_extents(handle, inode, &split_map, path,
+					  flags | EXT4_GET_BLOCKS_CONVERT, NULL);
 	if (IS_ERR(path))
 		return path;
 out:

@@ -3820,6 +3786,26 @@ ext4_ext_convert_to_initialized(handle_t *handle, struct inode *inode,
 	return ERR_PTR(err);
 }
 
+static bool ext4_ext_needs_conv(struct inode *inode, struct ext4_ext_path *path,
+				int flags)
+{
+	struct ext4_extent *ex;
+	bool is_unwrit;
+	int depth;
+
+	depth = ext_depth(inode);
+	ex = path[depth].p_ext;
+	is_unwrit = ext4_ext_is_unwritten(ex);
+
+	if (is_unwrit && (flags & EXT4_GET_BLOCKS_CONVERT))
+		return true;
+
+	if (!is_unwrit && (flags & EXT4_GET_BLOCKS_CONVERT_UNWRITTEN))
+		return true;
+
+	return false;
+}
+
 /*
  * This function is called by ext4_ext_map_blocks() from
  * ext4_get_blocks_dio_write() when DIO to write

@@ -3856,7 +3842,9 @@ static struct ext4_ext_path *ext4_split_convert_extents(handle_t *handle,
 	ext4_lblk_t ee_block;
 	struct ext4_extent *ex;
 	unsigned int ee_len;
-	int split_flag = 0, depth;
+	int split_flag = 0, depth, err = 0;
+	bool did_zeroout = false;
+	bool needs_conv = ext4_ext_needs_conv(inode, path, flags);
 
 	ext_debug(inode, "logical block %llu, max_blocks %u\n",
 		  (unsigned long long)map->m_lblk, map->m_len);

@@ -3870,19 +3858,81 @@ static struct ext4_ext_path *ext4_split_convert_extents(handle_t *handle,
 	ee_block = le32_to_cpu(ex->ee_block);
 	ee_len = ext4_ext_get_actual_len(ex);
 
-	if (flags & (EXT4_GET_BLOCKS_UNWRIT_EXT |
-			    EXT4_GET_BLOCKS_CONVERT)) {
+	/* No split needed */
+	if (ee_block == map->m_lblk && ee_len == map->m_len)
+		goto convert;
+
+	/*
+	 * We don't use zeroout fallback for written to unwritten conversion as
+	 * it is not as critical as endio and it might take unusually long.
+	 * Also, it is only safe to convert extent to initialized via explicit
+	 * zeroout only if extent is fully inside i_size or new_size.
+	 */
+	if (!(flags & EXT4_GET_BLOCKS_CONVERT_UNWRITTEN))
+		split_flag |= ee_block + ee_len <= eof_block ?
+				      EXT4_EXT_MAY_ZEROOUT :
+				      0;
+
+	/*
+	 * pass SPLIT_NOMERGE explicitly so we don't end up merging extents we
+	 * just split.
+	 */
+	path = ext4_split_extent(handle, inode, path, map, split_flag,
+				 flags | EXT4_GET_BLOCKS_SPLIT_NOMERGE,
+				 allocated, &did_zeroout);
+
+convert:
+	/*
+	 * We don't need a conversion if:
+	 * 1. There was an error in split.
+	 * 2. We split via zeroout.
+	 * 3. None of the convert flags were passed.
+	 */
+	if (IS_ERR(path) || did_zeroout || !needs_conv)
+		return path;
+
+	path = ext4_find_extent(inode, map->m_lblk, path, flags);
+	if (IS_ERR(path))
+		return path;
+
+	depth = ext_depth(inode);
+	ex = path[depth].p_ext;
+
+	err = ext4_ext_get_access(handle, inode, path + depth);
+	if (err)
+		goto err;
+
+	if (flags & EXT4_GET_BLOCKS_CONVERT)
+		ext4_ext_mark_initialized(ex);
+	else if (flags & EXT4_GET_BLOCKS_CONVERT_UNWRITTEN)
+		ext4_ext_mark_unwritten(ex);
+
+	if (!(flags & EXT4_GET_BLOCKS_SPLIT_NOMERGE))
 		/*
-		 * It is safe to convert extent to initialized via explicit
-		 * zeroout only if extent is fully inside i_size or new_size.
+		 * note: ext4_ext_correct_indexes() isn't needed here because
+		 * borders are not changed
 		 */
-		split_flag |= ee_block + ee_len <= eof_block ?
-			      EXT4_EXT_MAY_ZEROOUT : 0;
-		split_flag |= EXT4_EXT_MARK_UNWRIT2;
+		ext4_ext_try_to_merge(handle, inode, path, ex);
+
+	err = ext4_ext_dirty(handle, inode, path + depth);
+	if (err)
+		goto err;
+
+	/* Lets update the extent status tree after conversion */
+	ext4_es_insert_extent(inode, le32_to_cpu(ex->ee_block),
+			      ext4_ext_get_actual_len(ex), ext4_ext_pblock(ex),
+			      ext4_ext_is_unwritten(ex) ?
+				      EXTENT_STATUS_UNWRITTEN :
+				      EXTENT_STATUS_WRITTEN,
+			      false);
+
+err:
+	if (err) {
+		ext4_free_ext_path(path);
+		return ERR_PTR(err);
 	}
-	flags |= EXT4_GET_BLOCKS_SPLIT_NOMERGE;
-	return ext4_split_extent(handle, inode, path, map, split_flag, flags,
-				 allocated);
+
+	return path;
 }
 
 static struct ext4_ext_path *

@@ -3894,7 +3944,6 @@ ext4_convert_unwritten_extents_endio(handle_t *handle, struct inode *inode,
 	ext4_lblk_t ee_block;
 	unsigned int ee_len;
 	int depth;
-	int err = 0;
 
 	depth = ext_depth(inode);
 	ex = path[depth].p_ext;

@@ -3904,41 +3953,8 @@ ext4_convert_unwritten_extents_endio(handle_t *handle, struct inode *inode,
 	ext_debug(inode, "logical block %llu, max_blocks %u\n",
 		  (unsigned long long)ee_block, ee_len);
 
-	if (ee_block != map->m_lblk || ee_len > map->m_len) {
-		path = ext4_split_convert_extents(handle, inode, map, path,
-						  flags, NULL);
-		if (IS_ERR(path))
-			return path;
-
-		path = ext4_find_extent(inode, map->m_lblk, path, 0);
-		if (IS_ERR(path))
-			return path;
-		depth = ext_depth(inode);
-		ex = path[depth].p_ext;
-	}
-
-	err = ext4_ext_get_access(handle, inode, path + depth);
-	if (err)
-		goto errout;
-	/* first mark the extent as initialized */
-	ext4_ext_mark_initialized(ex);
-
-	/* note: ext4_ext_correct_indexes() isn't needed here because
-	 * borders are not changed
-	 */
-	ext4_ext_try_to_merge(handle, inode, path, ex);
-
-	/* Mark modified extent as dirty */
-	err = ext4_ext_dirty(handle, inode, path + path->p_depth);
-	if (err)
-		goto errout;
-
-	ext4_ext_show_leaf(inode, path);
-	return path;
-
-errout:
-	ext4_free_ext_path(path);
-	return ERR_PTR(err);
+	return ext4_split_convert_extents(handle, inode, map, path, flags,
+					  NULL);
 }
 
 static struct ext4_ext_path *

@@ -3952,7 +3968,6 @@ convert_initialized_extent(handle_t *handle, struct inode *inode,
 	ext4_lblk_t ee_block;
 	unsigned int ee_len;
 	int depth;
-	int err = 0;
 
 	/*
 	 * Make sure that the extent is no bigger than we support with

@@ -3969,40 +3984,12 @@ convert_initialized_extent(handle_t *handle, struct inode *inode,
 	ext_debug(inode, "logical block %llu, max_blocks %u\n",
 		  (unsigned long long)ee_block, ee_len);
 
-	if (ee_block != map->m_lblk || ee_len > map->m_len) {
-		path = ext4_split_convert_extents(handle, inode, map, path,
-				flags | EXT4_GET_BLOCKS_CONVERT_UNWRITTEN, NULL);
-		if (IS_ERR(path))
-			return path;
-
-		path = ext4_find_extent(inode, map->m_lblk, path, 0);
-		if (IS_ERR(path))
-			return path;
-		depth = ext_depth(inode);
-		ex = path[depth].p_ext;
-		if (!ex) {
-			EXT4_ERROR_INODE(inode, "unexpected hole at %lu",
-					 (unsigned long) map->m_lblk);
-			err = -EFSCORRUPTED;
-			goto errout;
-		}
-	}
-
-	err = ext4_ext_get_access(handle, inode, path + depth);
-	if (err)
-		goto errout;
-	/* first mark the extent as unwritten */
-	ext4_ext_mark_unwritten(ex);
-
-	/* note: ext4_ext_correct_indexes() isn't needed here because
-	 * borders are not changed
-	 */
-	ext4_ext_try_to_merge(handle, inode, path, ex);
+	path = ext4_split_convert_extents(
+		handle, inode, map, path,
+		flags | EXT4_GET_BLOCKS_CONVERT_UNWRITTEN, NULL);
+	if (IS_ERR(path))
+		return path;
 
-	/* Mark modified extent as dirty */
-	err = ext4_ext_dirty(handle, inode, path + path->p_depth);
-	if (err)
-		goto errout;
 	ext4_ext_show_leaf(inode, path);
 
 	ext4_update_inode_fsync_trans(handle, inode, 1);

@@ -4012,10 +3999,6 @@ convert_initialized_extent(handle_t *handle, struct inode *inode,
 		*allocated = map->m_len;
 	map->m_len = *allocated;
 	return path;
-
-errout:
-	ext4_free_ext_path(path);
-	return ERR_PTR(err);
 }
 
 static struct ext4_ext_path *

@@ -5649,7 +5632,7 @@ static int ext4_insert_range(struct file *file, loff_t offset, loff_t len)
 	struct ext4_extent *extent;
 	ext4_lblk_t start_lblk, len_lblk, ee_start_lblk = 0;
 	unsigned int credits, ee_len;
-	int ret, depth, split_flag = 0;
+	int ret, depth;
 	loff_t start;
 
 	trace_ext4_insert_range(inode, offset, len);

@@ -5720,12 +5703,8 @@ static int ext4_insert_range(struct file *file, loff_t offset, loff_t len)
 		 */
 		if ((start_lblk > ee_start_lblk) &&
 				(start_lblk < (ee_start_lblk + ee_len))) {
-			if (ext4_ext_is_unwritten(extent))
-				split_flag = EXT4_EXT_MARK_UNWRIT1 |
-					EXT4_EXT_MARK_UNWRIT2;
 			path = ext4_split_extent_at(handle, inode, path,
-					start_lblk, split_flag,
-					EXT4_EX_NOCACHE |
+					start_lblk, EXT4_EX_NOCACHE |
 					EXT4_GET_BLOCKS_SPLIT_NOMERGE |
 					EXT4_GET_BLOCKS_METADATA_NOFAIL);
 		}

Signature · 1 line

2.51.0

Signature · 2 lines

Jan Kara <[email protected]>
SUSE Labs, CR

Zhang Yi

2026-01-07 06:33 UTC #8 lore raw

Re: [PATCH 4/7] ext4: propagate flags to ext4_convert_unwritten_extents_endio()

In reply to Ojaswin Mujoo

Hi, Ojaswin!

Quoted 9 lines · Ojaswin Mujoo

Currently, callers like ext4_convert_unwritten_extents() pass EXT4_EX_NOCACHE flag to avoid caching extents however this is not respected by ext4_convert_unwritten_extents_endio(). Hence, modify it to accept flags from the caller and to pass the flags on to other extent manipulation functions it calls. This makes sure the NOCACHE flag is respected throughout the code path.
Also, since the caller already passes METADATA_NOFAIL and CONVERT flags we don't need to explicitly pass it anymore.

Thank you for the refactor! One comment below.

Quoted 20 lines

Signed-off-by Ojaswin Mujoo <[email protected]>

---
 fs/ext4/extents.c | 7 ++-----
 1 file changed, 2 insertions(+), 5 deletions(-)

fs/ext4/extents.c +1 -1

index 5228196f5ad4..460a70e6dae0 100644
@@ -3785,7 +3785,7 @@ static struct ext4_ext_path *ext4_split_convert_extents(handle_t *handle,
 static struct ext4_ext_path *
 ext4_convert_unwritten_extents_endio(handle_t *handle, struct inode *inode,
 				     struct ext4_map_blocks *map,
-				     struct ext4_ext_path *path)
+				     struct ext4_ext_path *path, int flags)
 {
 	struct ext4_extent *ex;
 	ext4_lblk_t ee_block;

fs/ext4/extents.c ext4_convert_unwritten_extents_endio(handle_t *handle, struct inode *inode, · line 3807

Quoted 9 lines

fs/ext4/extents.c +0 -3

@@ -3802,9 +3802,6 @@ ext4_convert_unwritten_extents_endio(handle_t *handle, struct inode *inode,
 		  (unsigned long long)ee_block, ee_len);
 
 	if (ee_block != map->m_lblk || ee_len > map->m_len) {
-		int flags = EXT4_GET_BLOCKS_CONVERT |
-			    EXT4_GET_BLOCKS_METADATA_NOFAIL;
-
 		path = ext4_split_convert_extents(handle, inode, map, path,
 						  flags, NULL);
 		if (IS_ERR(path))

There is another instance of ext4_find_extent() below that does not respect the EXT4_EX_NOCACHE flag. I think we should pass the flag as well.

Thanks,
Yi.

Quoted 9 lines

fs/ext4/extents.c +1 -1

@@ -3943,7 +3940,7 @@ ext4_ext_handle_unwritten_extents(handle_t *handle, struct inode *inode,
 	/* IO end_io complete, convert the filled extent to written */
 	if (flags & EXT4_GET_BLOCKS_CONVERT) {
 		path = ext4_convert_unwritten_extents_endio(handle, inode,
-							    map, path);
+							    map, path, flags);
 		if (IS_ERR(path))
 			return path;
 		ext4_update_inode_fsync_trans(handle, inode, 1);

Ojaswin Mujoo

author reviewer

2026-01-07 07:00 UTC #9 lore raw

Re: [PATCH 4/7] ext4: propagate flags to ext4_convert_unwritten_extents_endio()

Quoted 48 lines · Zhang Yi

Hi, Ojaswin!
Quoted 9 lines · Ojaswin Mujoo
Currently, callers like ext4_convert_unwritten_extents() pass EXT4_EX_NOCACHE flag to avoid caching extents however this is not respected by ext4_convert_unwritten_extents_endio(). Hence, modify it to accept flags from the caller and to pass the flags on to other extent manipulation functions it calls. This makes sure the NOCACHE flag is respected throughout the code path.
Also, since the caller already passes METADATA_NOFAIL and CONVERT flags we don't need to explicitly pass it anymore.
Thank you for the refactor! One comment below.
Quoted 20 lines
Signed-off-by Ojaswin Mujoo <[email protected]>
---
 fs/ext4/extents.c | 7 ++-----
 1 file changed, 2 insertions(+), 5 deletions(-)
fs/ext4/extents.c +1 -1
index 5228196f5ad4..460a70e6dae0 100644
@@ -3785,7 +3785,7 @@ static struct ext4_ext_path *ext4_split_convert_extents(handle_t *handle,
 static struct ext4_ext_path *
 ext4_convert_unwritten_extents_endio(handle_t *handle, struct inode *inode,
 				     struct ext4_map_blocks *map,
-				     struct ext4_ext_path *path)
+				     struct ext4_ext_path *path, int flags)
 {
 	struct ext4_extent *ex;
 	ext4_lblk_t ee_block;
fs/ext4/extents.c ext4_convert_unwritten_extents_endio(handle_t *handle, struct inode *inode, · line 3807
Quoted 9 lines
fs/ext4/extents.c +0 -3
@@ -3802,9 +3802,6 @@ ext4_convert_unwritten_extents_endio(handle_t *handle, struct inode *inode,
 		  (unsigned long long)ee_block, ee_len);
 
 	if (ee_block != map->m_lblk || ee_len > map->m_len) {
-		int flags = EXT4_GET_BLOCKS_CONVERT |
-			    EXT4_GET_BLOCKS_METADATA_NOFAIL;
-
 		path = ext4_split_convert_extents(handle, inode, map, path,
 						  flags, NULL);
 		if (IS_ERR(path))
There is another instance of ext4_find_extent() below that does not respect the EXT4_EX_NOCACHE flag. I think we should pass the flag as well.

Quoted 2 lines

Thanks,
Yi.

Hey Yi, thanks for the review.

Yes you are right, its removed in later commits but for completeness I'll add it there.

Thanks,
ojaswin

Quoted 11 lines

Quoted 10 lines

fs/ext4/extents.c +1 -1

@@ -3943,7 +3940,7 @@ ext4_ext_handle_unwritten_extents(handle_t *handle, struct inode *inode,
 	/* IO end_io complete, convert the filled extent to written */
 	if (flags & EXT4_GET_BLOCKS_CONVERT) {
 		path = ext4_convert_unwritten_extents_endio(handle, inode,
-							    map, path);
+							    map, path, flags);
 		if (IS_ERR(path))
 			return path;
 		ext4_update_inode_fsync_trans(handle, inode, 1);

Ojaswin Mujoo

author reviewer

2026-01-07 07:19 UTC #5 lore raw

Re: [PATCH 3/7] ext4: propagate flags to convert_initialized_extent()

In reply to Jan Kara

Quoted 40 lines · Jan Kara

Quoted 39 lines · Ojaswin Mujoo

Signed-off-by Ojaswin Mujoo <[email protected]>

---
 fs/ext4/extents-test.c | 2 +-
 fs/ext4/extents.c      | 5 +++--
 2 files changed, 4 insertions(+), 3 deletions(-)

fs/ext4/extents-test.c +1 -1

index 4fb94d3c8a1e..54aed3eabfe2 100644
@@ -422,7 +422,7 @@ static void test_convert_initialized(struct kunit *test)
 
 	map.m_lblk = param->split_map.m_lblk;
 	map.m_len = param->split_map.m_len;
-	convert_initialized_extent(NULL, inode, &map, path, &allocated);
+	convert_initialized_extent(NULL, inode, &map, path, 0, &allocated);
 
 	path = ext4_find_extent(inode, EX_DATA_LBLK, NULL, 0);
 	ex = path->p_ext;

fs/ext4/extents.c +1 -0

index 0ad0a9f2e3d4..5228196f5ad4 100644
@@ -3845,6 +3845,7 @@ static struct ext4_ext_path *
 convert_initialized_extent(handle_t *handle, struct inode *inode,
 			   struct ext4_map_blocks *map,
 			   struct ext4_ext_path *path,
+			   int flags,
 			   unsigned int *allocated)
 {
 	struct ext4_extent *ex;

@@ -3870,7 +3871,7 @@ convert_initialized_extent(handle_t *handle, struct inode *inode,

Quoted 7 lines

fs/ext4/extents.c convert_initialized_extent(handle_t *handle, struct inode *inode, · line 3874
Quoted 4 lines
fs/ext4/extents.c +1 -1
 	if (ee_block != map->m_lblk || ee_len > map->m_len) {
 		path = ext4_split_convert_extents(handle, inode, map, path,
-				EXT4_GET_BLOCKS_CONVERT_UNWRITTEN, NULL);
+				flags | EXT4_GET_BLOCKS_CONVERT_UNWRITTEN, NULL);
No need to keep EXT4_GET_BLOCKS_CONVERT_UNWRITTEN here as the caller has it in flags already? Otherwise the patch looks good.

Hi Jan,

Right, I'll make this change in v2.

Thanks again!
Ojaswin

Quoted 21 lines

Honza

Quoted 15 lines

fs/ext4/extents.c +1 -1

 		if (IS_ERR(path))
 			return path;

@@ -4264,7 +4265,7 @@ int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
 			if ((!ext4_ext_is_unwritten(ex)) &&
 			    (flags & EXT4_GET_BLOCKS_CONVERT_UNWRITTEN)) {
 				path = convert_initialized_extent(handle,
-					inode, map, path, &allocated);
+					inode, map, path, flags, &allocated);
 				if (IS_ERR(path))
 					err = PTR_ERR(path);
 				goto out;

Signature · 1 line

2.51.0

Signature · 2 lines

Jan Kara <[email protected]>
SUSE Labs, CR

Ojaswin Mujoo

author reviewer

2026-01-07 07:21 UTC #12 lore raw

Re: [PATCH 5/7] ext4: Refactor zeroout path and handle all cases

In reply to Jan Kara

Quoted 27 lines · Jan Kara

Quoted 14 lines · Ojaswin Mujoo
Currently, zeroout is used as a fallback in case we fail to split/convert extents in the "traditional" modify-the-extent-tree way. This is essential to mitigate failures in critical paths like extent splitting during endio. However, the logic is very messy and not easy to follow. Further, the fragile use of various flags has made it prone to errors.
Refactor zeroout out logic by moving it up to ext4_split_extents(). Further, zeroout correctly based on the type of conversion we want, ie:
- unwritten to written: Zeroout everything around the mapped range.
- unwritten to unwritten: Zeroout everything
- written to unwritten: Zeroout only the mapped range.
Signed-off-by Ojaswin Mujoo <[email protected]>
...
Quoted 9 lines
fs/ext4/extents.c +1 -3
@@ -3383,11 +3440,12 @@ static struct ext4_ext_path *ext4_split_extent(handle_t *handle,
 					       int split_flag, int flags,
 					       unsigned int *allocated)
 {
-	ext4_lblk_t ee_block;
+	ext4_lblk_t ee_block, orig_ee_block;
 	struct ext4_extent *ex;
-	unsigned int ee_len, depth;
-	int unwritten;

Quoted 8 lines

fs/ext4/extents.c static struct ext4_ext_path *ext4_split_extent(handle_t *handle, · line 3448
Quoted 5 lines
fs/ext4/extents.c +4 -1
-	int split_flag1, flags1;
+	unsigned int ee_len, orig_ee_len, depth;
+	int unwritten, orig_unwritten;
+	int split_flag1 = 0, flags1 = 0;
+	int err = 0, orig_err;
Cannot orig_err be used uninitialized in this function? At least it isn't obvious to me some of the branches setting it is always taken.

Hi Jan, thanks for the reviews. Yes orig_err is always initialized before it is used (initialized on error and used in zeroout path which is only taked on error), but I agree that we can just init it to 0.

Quoted 64 lines

fs/ext4/extents.c +23 ~~-13~~

@@ -3395,23 +3453,29 @@ static struct ext4_ext_path *ext4_split_extent(handle_t *handle,
 	ee_len = ext4_ext_get_actual_len(ex);
 	unwritten = ext4_ext_is_unwritten(ex);
 
+	orig_ee_block = ee_block;
+	orig_ee_len = ee_len;
+	orig_unwritten = unwritten;
+
 	/* Do not cache extents that are in the process of being modified. */
 	flags |= EXT4_EX_NOCACHE;
 
 	if (map->m_lblk + map->m_len < ee_block + ee_len) {
-		split_flag1 = split_flag & EXT4_EXT_MAY_ZEROOUT;
 		flags1 = flags | EXT4_GET_BLOCKS_SPLIT_NOMERGE;
 		if (unwritten)
 			split_flag1 |= EXT4_EXT_MARK_UNWRIT1 |
 				       EXT4_EXT_MARK_UNWRIT2;
-		if (split_flag & EXT4_EXT_DATA_VALID2)
-			split_flag1 |= map->m_lblk > ee_block ?
-				       EXT4_EXT_DATA_PARTIAL_VALID1 :
-				       EXT4_EXT_DATA_ENTIRE_VALID1;
 		path = ext4_split_extent_at(handle, inode, path,
 				map->m_lblk + map->m_len, split_flag1, flags1);
-		if (IS_ERR(path))
-			return path;
+
+		if (IS_ERR(path)) {
+			orig_err = PTR_ERR(path);
+			if (orig_err != -ENOSPC && orig_err != -EDQUOT &&
+			    orig_err != -ENOMEM)
+				return path;
+
+			goto try_zeroout;
+		}
 		/*
 		 * Update path is required because previous ext4_split_extent_at
 		 * may result in split of original leaf or extent zeroout.

@@ -3427,22 +3491,68 @@ static struct ext4_ext_path *ext4_split_extent(handle_t *handle,
 			ext4_free_ext_path(path);
 			return ERR_PTR(-EFSCORRUPTED);
 		}
-		unwritten = ext4_ext_is_unwritten(ex);
 	}
 
 	if (map->m_lblk >= ee_block) {
-		split_flag1 = split_flag & EXT4_EXT_DATA_VALID2;
+		split_flag1 = 0;
 		if (unwritten) {
 			split_flag1 |= EXT4_EXT_MARK_UNWRIT1;
-			split_flag1 |= split_flag & (EXT4_EXT_MAY_ZEROOUT |
-						     EXT4_EXT_MARK_UNWRIT2);
+			split_flag1 |= split_flag & EXT4_EXT_MARK_UNWRIT2;
 		}
-		path = ext4_split_extent_at(handle, inode, path,
-				map->m_lblk, split_flag1, flags);
+		path = ext4_split_extent_at(handle, inode, path, map->m_lblk,
+					    split_flag1, flags);
+
+		if (IS_ERR(path)) {
+			orig_err = PTR_ERR(path);
+			if (orig_err != -ENOSPC && orig_err != -EDQUOT &&
+			    orig_err != -ENOMEM)
+				return path;

Quoted 8 lines

fs/ext4/extents.c static struct ext4_ext_path *ext4_split_extent(handle_t *handle, · line 3515
Quoted 6 lines
fs/ext4/extents.c +6 -0
+
+			goto try_zeroout;
+		}
+	}
+
+	if (!err)
Nothing touches 'err' in this function...

Yes :), I'll remove this.

Quoted 30 lines

fs/ext4/extents.c +27 -0

+		goto out;
+
+try_zeroout:
+	/*
+	 * There was an error in splitting the extent, just zeroout and convert
+	 * to initialize as a last resort
+	 */
+	if (split_flag & EXT4_EXT_MAY_ZEROOUT) {
+		path = ext4_find_extent(inode, map->m_lblk, NULL, flags);
 		if (IS_ERR(path))
 			return path;
+
+		depth = ext_depth(inode);
+		ex = path[depth].p_ext;
+		ee_block = le32_to_cpu(ex->ee_block);
+		ee_len = ext4_ext_get_actual_len(ex);
+		unwritten = ext4_ext_is_unwritten(ex);
+
+		/*
+		 * The extent to zeroout should have been unchanged
+		 * but its not, just return error to caller
+		 */
+		if (WARN_ON(ee_block != orig_ee_block ||
+			    ee_len != orig_ee_len ||
+			    unwritten != orig_unwritten))
+			return ERR_PTR(orig_err);
+
+		/*
+		 * Something went wrong in zeroout, just return the

Quoted 8 lines

fs/ext4/extents.c static struct ext4_ext_path *ext4_split_extent(handle_t *handle, · line 3549
Quoted 5 lines
fs/ext4/extents.c +4 -0
+		 * original error
+		 */
+		if (ext4_split_extent_zeroout(handle, inode, path, map, flags))
+			return ERR_PTR(orig_err);
 	}
Also nothing seems to zero out orig_err in case zero out above succeeded. What am I missing?

So if zeroout here succeeds we just goto out and return path, we never use orig_err. Not the best practice and admittedly, I seem to have complicated the error handling a bit. I will streamline it in v2.

Thanks for pointing this out.
Ojaswin.

Quoted 15 lines

Honza

Quoted 9 lines

fs/ext4/extents.c +5 -0

+	/* There's an error and we can't zeroout, just return the err */
+	return ERR_PTR(orig_err);
+
+out:
+
 	if (allocated) {
 		if (map->m_lblk + map->m_len > ee_block + ee_len)
 			*allocated = ee_len - (map->m_lblk - ee_block);

Signature · 2 lines

Jan Kara <[email protected]>
SUSE Labs, CR

Zhang Yi

2026-01-08 11:58 UTC #13 lore raw

Re: [PATCH 5/7] ext4: Refactor zeroout path and handle all cases

In reply to Ojaswin Mujoo

Quoted 7 lines · Ojaswin Mujoo

Currently, zeroout is used as a fallback in case we fail to split/convert extents in the "traditional" modify-the-extent-tree way. This is essential to mitigate failures in critical paths like extent splitting during endio. However, the logic is very messy and not easy to follow. Further, the fragile use of various flags has made it prone to errors.

Quoted 7 lines

Refactor zeroout out logic by moving it up to ext4_split_extents(). Further, zeroout correctly based on the type of conversion we want, ie:
- unwritten to written: Zeroout everything around the mapped range.
- unwritten to unwritten: Zeroout everything
- written to unwritten: Zeroout only the mapped range.
Signed-off-by Ojaswin Mujoo <[email protected]>

Hi, Ojaswin!

The refactor overall looks good to me. After this series, the split logic becomes more straightforward and clear. :)

I have some comments below.

Quoted 8 lines

---
 fs/ext4/extents.c | 287 +++++++++++++++++++++++++++++++---------------
 1 file changed, 195 insertions(+), 92 deletions(-)

fs/ext4/extents.c +0 -0

index 460a70e6dae0..8082e1d93bbf 100644

[...]

Quoted 25 lines

fs/ext4/extents.c +21 -0

@@ -3365,6 +3313,115 @@ static struct ext4_ext_path *ext4_split_extent_at(handle_t *handle,
 	return path;
 }
 
+static struct ext4_ext_path *
+ext4_split_extent_zeroout(handle_t *handle, struct inode *inode,
+			  struct ext4_ext_path *path,
+			  struct ext4_map_blocks *map, int flags)
+{
+	struct ext4_extent *ex;
+	unsigned int ee_len, depth;
+	ext4_lblk_t ee_block;
+	uint64_t lblk, pblk, len;
+	int is_unwrit;
+	int err = 0;
+
+	depth = ext_depth(inode);
+	ex = path[depth].p_ext;
+	ee_block = le32_to_cpu(ex->ee_block);
+	ee_len = ext4_ext_get_actual_len(ex);
+	is_unwrit = ext4_ext_is_unwritten(ex);
+
+	if (flags & EXT4_GET_BLOCKS_CONVERT) {
+		/*
+		 * EXT4_GET_BLOCKS_CONVERT: Caller wants the range specified by

fs/ext4/extents.c static struct ext4_ext_path *ext4_split_extent_at(handle_t *handle, · line 3346

Quoted 10 lines

fs/ext4/extents.c +10 -0

+		 * map to be initialized. Zeroout everything except the map
+		 * range.
+		 */
+
+		loff_t map_end = (loff_t) map->m_lblk + map->m_len;
+		loff_t ex_end = (loff_t) ee_block + ee_len;
+
+		if (!is_unwrit)
+			/* Shouldn't happen. Just exit */
+			return ERR_PTR(-EINVAL);

For cases that are should not happen, I'd suggest adding a WARN_ON_ONCE or a message to facilitate future problem identification. Same as below.

Quoted 42 lines

fs/ext4/extents.c +42 -0

+
+		/* zeroout left */
+		if (map->m_lblk > ee_block) {
+			lblk = ee_block;
+			len = map->m_lblk - ee_block;
+			pblk = ext4_ext_pblock(ex);
+			err = ext4_issue_zeroout(inode, lblk, pblk, len);
+			if (err)
+				/* ZEROOUT failed, just return original error */
+				return ERR_PTR(err);
+		}
+
+		/* zeroout right */
+		if (map->m_lblk + map->m_len < ee_block + ee_len) {
+			lblk = map_end;
+			len = ex_end - map_end;
+			pblk = ext4_ext_pblock(ex) + (map_end - ee_block);
+			err = ext4_issue_zeroout(inode, lblk, pblk, len);
+			if (err)
+				/* ZEROOUT failed, just return original error */
+				return ERR_PTR(err);
+		}
+	} else if (flags & EXT4_GET_BLOCKS_CONVERT_UNWRITTEN) {
+		/*
+		 * EXT4_GET_BLOCKS_CONVERT_UNWRITTEN: Caller wants the
+		 * range specified by map to be marked unwritten.
+		 * Zeroout the map range leaving rest as it is.
+		 */
+
+		if (is_unwrit)
+			/* Shouldn't happen. Just exit */
+			return ERR_PTR(-EINVAL);
+
+		lblk = map->m_lblk;
+		len = map->m_len;
+		pblk = ext4_ext_pblock(ex) + (map->m_lblk - ee_block);
+		err = ext4_issue_zeroout(inode, lblk, pblk, len);
+		if (err)
+			/* ZEROOUT failed, just return original error */
+			return ERR_PTR(err);
+	} else if (flags & EXT4_GET_BLOCKS_UNWRIT_EXT) {
+		/*

fs/ext4/extents.c static struct ext4_ext_path *ext4_split_extent_at(handle_t *handle, · line 3398

Quoted 10 lines

fs/ext4/extents.c +10 -0

+		 * EXT4_GET_BLOCKS_UNWRIT_EXT: Today, this flag
+		 * implicitly implies that callers when wanting an
+		 * unwritten to unwritten split. So zeroout the whole
+		 * extent.
+		 *
+		 * TODO: The implicit meaning of the flag is not ideal
+		 * and eventually we should aim for a more well defined
+		 * behavior
+		 */
+

I don't think we need this branch anymore. After applying my patch "ext4:
don't split extent before submitting I/O", we will no longer encounter situations where doing an unwritten to unwritten split. It means that at all call sites of ext4_split_extent(), only EXT4_GET_BLOCKS_CONVERT or EXT4_GET_BLOCKS_CONVERT_UNWRITTEN flags are passed. What do you think?

Thanks,
Yi.

Quoted 29 lines

fs/ext4/extents.c +26 -0

+		if (!is_unwrit)
+			/* Shouldn't happen. Just exit */
+			return ERR_PTR(-EINVAL);
+
+		lblk = ee_block;
+		len = ee_len;
+		pblk = ext4_ext_pblock(ex);
+		err = ext4_issue_zeroout(inode, lblk, pblk, len);
+		if (err)
+			/* ZEROOUT failed, just return original error */
+			return ERR_PTR(err);
+	}
+
+	err = ext4_ext_get_access(handle, inode, path + depth);
+	if (err)
+		return ERR_PTR(err);
+
+	ext4_ext_mark_initialized(ex);
+
+	ext4_ext_dirty(handle, inode, path + path->p_depth);
+	if (err)
+		return ERR_PTR(err);
+
+	return 0;
+}
+
 /*
  * ext4_split_extent() splits an extent and mark extent which is covered
  * by @map as split_flags indicates

[...]

Zhang Yi

2026-01-08 12:34 UTC #19 lore raw

Re: [PATCH 6/7] ext4: Refactor split and convert extents

In reply to Ojaswin Mujoo

Quoted 30 lines · Ojaswin Mujoo

ext4_split_convert_extents() has been historically prone to subtle bugs and inconsistent behavior due to the way all the various flags interact with the extent split and conversion process. For example, callers like ext4_convert_unwritten_extents_endio() and convert_initialized_extents() needed to open code extent conversion despite passing CONVERT or CONVERT_UNWRITTEN flags because ext4_split_convert_extents() wasn't performing the conversion.
Hence, refactor ext4_split_convert_extents() to clearly enforce the semantics of each flag. The major changes here are:
* Clearly separate the split and convert process:
* ext4_split_extent() and ext4_split_extent_at() are now only responsible to perform the split.
* ext4_split_convert_extents() is now responsible to perform extent conversion after calling ext4_split_extent() for splitting.
* This helps get rid of all the MARK_UNWRIT* flags.
* Clearly enforce the semantics of flags passed to ext4_split_convert_extents():
* EXT4_GET_BLOCKS_CONVERT: Will convert the split extent to written
* EXT4_GET_BLOCKS_CONVERT_UNWRITTEN: Will convert the split extent to unwritten
* Passing neither of the above means we only want a split.
* Modify all callers to enforce the above semantics.
* Use ext4_split_convert_extents() instead of ext4_split_extents()
* in ext4_ext_convert_to_initialized() for uniformity.

Quoted 7 lines

* Cleanup all callers open coding the conversion logic.
* Further, modify kuniy tests to pass flags based on the new semantics.
From an end user point of view, we should not see any changes in behavior of ext4.
Signed-off-by Ojaswin Mujoo <[email protected]>

Some comments below.

Quoted 5 lines

---
 fs/ext4/extents-test.c |  12 +-
 fs/ext4/extents.c      | 299 +++++++++++++++++++----------------------
 2 files changed, 145 insertions(+), 166 deletions(-)

[..]

Quoted 28 lines

fs/ext4/extents.c +20 -0

@@ -3820,6 +3786,26 @@ ext4_ext_convert_to_initialized(handle_t *handle, struct inode *inode,
 	return ERR_PTR(err);
 }
 
+static bool ext4_ext_needs_conv(struct inode *inode, struct ext4_ext_path *path,
+				int flags)
+{
+	struct ext4_extent *ex;
+	bool is_unwrit;
+	int depth;
+
+	depth = ext_depth(inode);
+	ex = path[depth].p_ext;
+	is_unwrit = ext4_ext_is_unwritten(ex);
+
+	if (is_unwrit && (flags & EXT4_GET_BLOCKS_CONVERT))
+		return true;
+
+	if (!is_unwrit && (flags & EXT4_GET_BLOCKS_CONVERT_UNWRITTEN))
+		return true;
+
+	return false;
+}
+
 /*
  * This function is called by ext4_ext_map_blocks() from
  * ext4_get_blocks_dio_write() when DIO to write

fs/ext4/extents.c static struct ext4_ext_path *ext4_split_convert_extents(handle_t *handle, · line 3847

Quoted 7 lines

fs/ext4/extents.c +3 -1

@@ -3856,7 +3842,9 @@ static struct ext4_ext_path *ext4_split_convert_extents(handle_t *handle,
 	ext4_lblk_t ee_block;
 	struct ext4_extent *ex;
 	unsigned int ee_len;
-	int split_flag = 0, depth;
+	int split_flag = 0, depth, err = 0;
+	bool did_zeroout = false;
+	bool needs_conv = ext4_ext_needs_conv(inode, path, flags);

As I described in Patch 05, there is currently no situation where splitting occurs without conversion, so I don't think we need this check. Is it right?

Quoted 71 lines

fs/ext4/extents.c +55 -7

 	ext_debug(inode, "logical block %llu, max_blocks %u\n",
 		  (unsigned long long)map->m_lblk, map->m_len);

@@ -3870,19 +3858,81 @@ static struct ext4_ext_path *ext4_split_convert_extents(handle_t *handle,
 	ee_block = le32_to_cpu(ex->ee_block);
 	ee_len = ext4_ext_get_actual_len(ex);
 
-	if (flags & (EXT4_GET_BLOCKS_UNWRIT_EXT |
-			    EXT4_GET_BLOCKS_CONVERT)) {
+	/* No split needed */
+	if (ee_block == map->m_lblk && ee_len == map->m_len)
+		goto convert;
+
+	/*
+	 * We don't use zeroout fallback for written to unwritten conversion as
+	 * it is not as critical as endio and it might take unusually long.
+	 * Also, it is only safe to convert extent to initialized via explicit
+	 * zeroout only if extent is fully inside i_size or new_size.
+	 */
+	if (!(flags & EXT4_GET_BLOCKS_CONVERT_UNWRITTEN))
+		split_flag |= ee_block + ee_len <= eof_block ?
+				      EXT4_EXT_MAY_ZEROOUT :
+				      0;
+
+	/*
+	 * pass SPLIT_NOMERGE explicitly so we don't end up merging extents we
+	 * just split.
+	 */
+	path = ext4_split_extent(handle, inode, path, map, split_flag,
+				 flags | EXT4_GET_BLOCKS_SPLIT_NOMERGE,
+				 allocated, &did_zeroout);
+
+convert:
+	/*
+	 * We don't need a conversion if:
+	 * 1. There was an error in split.
+	 * 2. We split via zeroout.
+	 * 3. None of the convert flags were passed.
+	 */
+	if (IS_ERR(path) || did_zeroout || !needs_conv)
+		return path;
+
+	path = ext4_find_extent(inode, map->m_lblk, path, flags);
+	if (IS_ERR(path))
+		return path;
+
+	depth = ext_depth(inode);
+	ex = path[depth].p_ext;
+
+	err = ext4_ext_get_access(handle, inode, path + depth);
+	if (err)
+		goto err;
+
+	if (flags & EXT4_GET_BLOCKS_CONVERT)
+		ext4_ext_mark_initialized(ex);
+	else if (flags & EXT4_GET_BLOCKS_CONVERT_UNWRITTEN)
+		ext4_ext_mark_unwritten(ex);
+
+	if (!(flags & EXT4_GET_BLOCKS_SPLIT_NOMERGE))
 		/*
-		 * It is safe to convert extent to initialized via explicit
-		 * zeroout only if extent is fully inside i_size or new_size.
+		 * note: ext4_ext_correct_indexes() isn't needed here because
+		 * borders are not changed
 		 */
-		split_flag |= ee_block + ee_len <= eof_block ?
-			      EXT4_EXT_MAY_ZEROOUT : 0;
-		split_flag |= EXT4_EXT_MARK_UNWRIT2;
+		ext4_ext_try_to_merge(handle, inode, path, ex);
+
+	err = ext4_ext_dirty(handle, inode, path + depth);

fs/ext4/extents.c static struct ext4_ext_path *ext4_split_convert_extents(handle_t *handle, · line 3927

Quoted 10 lines

fs/ext4/extents.c +10 -0

+	if (err)
+		goto err;
+
+	/* Lets update the extent status tree after conversion */
+	ext4_es_insert_extent(inode, le32_to_cpu(ex->ee_block),
+			      ext4_ext_get_actual_len(ex), ext4_ext_pblock(ex),
+			      ext4_ext_is_unwritten(ex) ?
+				      EXTENT_STATUS_UNWRITTEN :
+				      EXTENT_STATUS_WRITTEN,
+			      false);

I think the did_zeroout case also should update the extent status tree (and it should be better to be added in the previous patch). Otherwise, this would lead to residual stale unwritten extents in the zeroed range.

We should be careful about the extent status tree. I'd suggest that pay close attention to the following error output when running tests,

EXT4-fs warning (device pmem2s): ext4_es_cache_extent:1079: inode #718: comm 108573.fsstress: ES cache extent failed: add [55,22,12260,0x1] conflict with existing [75,2,12280,0x2]

or directly add a WARN_ON_ONCE(1) at the end of ext4_es_cache_extent(). There may be other problems related to the stale extents that could arise.

Thanks,
Yi.

Quoted 154 lines

fs/ext4/extents.c +16 ~~-83~~

+
+err:
+	if (err) {
+		ext4_free_ext_path(path);
+		return ERR_PTR(err);
 	}
-	flags |= EXT4_GET_BLOCKS_SPLIT_NOMERGE;
-	return ext4_split_extent(handle, inode, path, map, split_flag, flags,
-				 allocated);
+
+	return path;
 }
 
 static struct ext4_ext_path *

@@ -3894,7 +3944,6 @@ ext4_convert_unwritten_extents_endio(handle_t *handle, struct inode *inode,
 	ext4_lblk_t ee_block;
 	unsigned int ee_len;
 	int depth;
-	int err = 0;
 
 	depth = ext_depth(inode);
 	ex = path[depth].p_ext;

@@ -3904,41 +3953,8 @@ ext4_convert_unwritten_extents_endio(handle_t *handle, struct inode *inode,
 	ext_debug(inode, "logical block %llu, max_blocks %u\n",
 		  (unsigned long long)ee_block, ee_len);
 
-	if (ee_block != map->m_lblk || ee_len > map->m_len) {
-		path = ext4_split_convert_extents(handle, inode, map, path,
-						  flags, NULL);
-		if (IS_ERR(path))
-			return path;
-
-		path = ext4_find_extent(inode, map->m_lblk, path, 0);
-		if (IS_ERR(path))
-			return path;
-		depth = ext_depth(inode);
-		ex = path[depth].p_ext;
-	}
-
-	err = ext4_ext_get_access(handle, inode, path + depth);
-	if (err)
-		goto errout;
-	/* first mark the extent as initialized */
-	ext4_ext_mark_initialized(ex);
-
-	/* note: ext4_ext_correct_indexes() isn't needed here because
-	 * borders are not changed
-	 */
-	ext4_ext_try_to_merge(handle, inode, path, ex);
-
-	/* Mark modified extent as dirty */
-	err = ext4_ext_dirty(handle, inode, path + path->p_depth);
-	if (err)
-		goto errout;
-
-	ext4_ext_show_leaf(inode, path);
-	return path;
-
-errout:
-	ext4_free_ext_path(path);
-	return ERR_PTR(err);
+	return ext4_split_convert_extents(handle, inode, map, path, flags,
+					  NULL);
 }
 
 static struct ext4_ext_path *

@@ -3952,7 +3968,6 @@ convert_initialized_extent(handle_t *handle, struct inode *inode,
 	ext4_lblk_t ee_block;
 	unsigned int ee_len;
 	int depth;
-	int err = 0;
 
 	/*
 	 * Make sure that the extent is no bigger than we support with

@@ -3969,40 +3984,12 @@ convert_initialized_extent(handle_t *handle, struct inode *inode,
 	ext_debug(inode, "logical block %llu, max_blocks %u\n",
 		  (unsigned long long)ee_block, ee_len);
 
-	if (ee_block != map->m_lblk || ee_len > map->m_len) {
-		path = ext4_split_convert_extents(handle, inode, map, path,
-				flags | EXT4_GET_BLOCKS_CONVERT_UNWRITTEN, NULL);
-		if (IS_ERR(path))
-			return path;
-
-		path = ext4_find_extent(inode, map->m_lblk, path, 0);
-		if (IS_ERR(path))
-			return path;
-		depth = ext_depth(inode);
-		ex = path[depth].p_ext;
-		if (!ex) {
-			EXT4_ERROR_INODE(inode, "unexpected hole at %lu",
-					 (unsigned long) map->m_lblk);
-			err = -EFSCORRUPTED;
-			goto errout;
-		}
-	}
-
-	err = ext4_ext_get_access(handle, inode, path + depth);
-	if (err)
-		goto errout;
-	/* first mark the extent as unwritten */
-	ext4_ext_mark_unwritten(ex);
-
-	/* note: ext4_ext_correct_indexes() isn't needed here because
-	 * borders are not changed
-	 */
-	ext4_ext_try_to_merge(handle, inode, path, ex);
+	path = ext4_split_convert_extents(
+		handle, inode, map, path,
+		flags | EXT4_GET_BLOCKS_CONVERT_UNWRITTEN, NULL);
+	if (IS_ERR(path))
+		return path;
 
-	/* Mark modified extent as dirty */
-	err = ext4_ext_dirty(handle, inode, path + path->p_depth);
-	if (err)
-		goto errout;
 	ext4_ext_show_leaf(inode, path);
 
 	ext4_update_inode_fsync_trans(handle, inode, 1);

@@ -4012,10 +3999,6 @@ convert_initialized_extent(handle_t *handle, struct inode *inode,
 		*allocated = map->m_len;
 	map->m_len = *allocated;
 	return path;
-
-errout:
-	ext4_free_ext_path(path);
-	return ERR_PTR(err);
 }
 
 static struct ext4_ext_path *

@@ -5649,7 +5632,7 @@ static int ext4_insert_range(struct file *file, loff_t offset, loff_t len)
 	struct ext4_extent *extent;
 	ext4_lblk_t start_lblk, len_lblk, ee_start_lblk = 0;
 	unsigned int credits, ee_len;
-	int ret, depth, split_flag = 0;
+	int ret, depth;
 	loff_t start;
 
 	trace_ext4_insert_range(inode, offset, len);

@@ -5720,12 +5703,8 @@ static int ext4_insert_range(struct file *file, loff_t offset, loff_t len)
 		 */
 		if ((start_lblk > ee_start_lblk) &&
 				(start_lblk < (ee_start_lblk + ee_len))) {
-			if (ext4_ext_is_unwritten(extent))
-				split_flag = EXT4_EXT_MARK_UNWRIT1 |
-					EXT4_EXT_MARK_UNWRIT2;
 			path = ext4_split_extent_at(handle, inode, path,
-					start_lblk, split_flag,
-					EXT4_EX_NOCACHE |
+					start_lblk, EXT4_EX_NOCACHE |
 					EXT4_GET_BLOCKS_SPLIT_NOMERGE |
 					EXT4_GET_BLOCKS_METADATA_NOFAIL);
 		}

Ojaswin Mujoo

author reviewer

2026-01-08 12:43 UTC #14 lore raw

Re: [PATCH 5/7] ext4: Refactor zeroout path and handle all cases

In reply to Zhang Yi

Quoted 64 lines · Zhang Yi

Quoted 7 lines · Ojaswin Mujoo
Currently, zeroout is used as a fallback in case we fail to split/convert extents in the "traditional" modify-the-extent-tree way. This is essential to mitigate failures in critical paths like extent splitting during endio. However, the logic is very messy and not easy to follow. Further, the fragile use of various flags has made it prone to errors.
Quoted 7 lines
Refactor zeroout out logic by moving it up to ext4_split_extents(). Further, zeroout correctly based on the type of conversion we want, ie:
- unwritten to written: Zeroout everything around the mapped range.
- unwritten to unwritten: Zeroout everything
- written to unwritten: Zeroout only the mapped range.
Signed-off-by Ojaswin Mujoo <[email protected]>
Hi, Ojaswin!
The refactor overall looks good to me. After this series, the split logic becomes more straightforward and clear. :)
I have some comments below.
Quoted 8 lines
---
 fs/ext4/extents.c | 287 +++++++++++++++++++++++++++++++---------------
 1 file changed, 195 insertions(+), 92 deletions(-)
fs/ext4/extents.c +0 -0
index 460a70e6dae0..8082e1d93bbf 100644
[...]
Quoted 30 lines
fs/ext4/extents.c +26 -0
@@ -3365,6 +3313,115 @@ static struct ext4_ext_path *ext4_split_extent_at(handle_t *handle,
 	return path;
 }
 
+static struct ext4_ext_path *
+ext4_split_extent_zeroout(handle_t *handle, struct inode *inode,
+			  struct ext4_ext_path *path,
+			  struct ext4_map_blocks *map, int flags)
+{
+	struct ext4_extent *ex;
+	unsigned int ee_len, depth;
+	ext4_lblk_t ee_block;
+	uint64_t lblk, pblk, len;
+	int is_unwrit;
+	int err = 0;
+
+	depth = ext_depth(inode);
+	ex = path[depth].p_ext;
+	ee_block = le32_to_cpu(ex->ee_block);
+	ee_len = ext4_ext_get_actual_len(ex);
+	is_unwrit = ext4_ext_is_unwritten(ex);
+
+	if (flags & EXT4_GET_BLOCKS_CONVERT) {
+		/*
+		 * EXT4_GET_BLOCKS_CONVERT: Caller wants the range specified by
+		 * map to be initialized. Zeroout everything except the map
+		 * range.
+		 */
+
+		loff_t map_end = (loff_t) map->m_lblk + map->m_len;

Quoted 8 lines

fs/ext4/extents.c static struct ext4_ext_path *ext4_split_extent_at(handle_t *handle, · line 3346
Quoted 5 lines
fs/ext4/extents.c +5 -0
+		loff_t ex_end = (loff_t) ee_block + ee_len;
+
+		if (!is_unwrit)
+			/* Shouldn't happen. Just exit */
+			return ERR_PTR(-EINVAL);
For cases that are should not happen, I'd suggest adding a WARN_ON_ONCE or a message to facilitate future problem identification. Same as below.

Hi Yi,

Thanks for the review! Sure I can do that in v2.

Quoted 51 lines

fs/ext4/extents.c +50 -0

+
+		/* zeroout left */
+		if (map->m_lblk > ee_block) {
+			lblk = ee_block;
+			len = map->m_lblk - ee_block;
+			pblk = ext4_ext_pblock(ex);
+			err = ext4_issue_zeroout(inode, lblk, pblk, len);
+			if (err)
+				/* ZEROOUT failed, just return original error */
+				return ERR_PTR(err);
+		}
+
+		/* zeroout right */
+		if (map->m_lblk + map->m_len < ee_block + ee_len) {
+			lblk = map_end;
+			len = ex_end - map_end;
+			pblk = ext4_ext_pblock(ex) + (map_end - ee_block);
+			err = ext4_issue_zeroout(inode, lblk, pblk, len);
+			if (err)
+				/* ZEROOUT failed, just return original error */
+				return ERR_PTR(err);
+		}
+	} else if (flags & EXT4_GET_BLOCKS_CONVERT_UNWRITTEN) {
+		/*
+		 * EXT4_GET_BLOCKS_CONVERT_UNWRITTEN: Caller wants the
+		 * range specified by map to be marked unwritten.
+		 * Zeroout the map range leaving rest as it is.
+		 */
+
+		if (is_unwrit)
+			/* Shouldn't happen. Just exit */
+			return ERR_PTR(-EINVAL);
+
+		lblk = map->m_lblk;
+		len = map->m_len;
+		pblk = ext4_ext_pblock(ex) + (map->m_lblk - ee_block);
+		err = ext4_issue_zeroout(inode, lblk, pblk, len);
+		if (err)
+			/* ZEROOUT failed, just return original error */
+			return ERR_PTR(err);
+	} else if (flags & EXT4_GET_BLOCKS_UNWRIT_EXT) {
+		/*
+		 * EXT4_GET_BLOCKS_UNWRIT_EXT: Today, this flag
+		 * implicitly implies that callers when wanting an
+		 * unwritten to unwritten split. So zeroout the whole
+		 * extent.
+		 *
+		 * TODO: The implicit meaning of the flag is not ideal
+		 * and eventually we should aim for a more well defined
+		 * behavior

Quoted 8 lines

fs/ext4/extents.c static struct ext4_ext_path *ext4_split_extent_at(handle_t *handle, · line 3398
Quoted 2 lines
fs/ext4/extents.c +2 -0
+		 */
+
I don't think we need this branch anymore. After applying my patch "ext4:
don't split extent before submitting I/O", we will no longer encounter situations where doing an unwritten to unwritten split. It means that at all call sites of ext4_split_extent(), only EXT4_GET_BLOCKS_CONVERT or EXT4_GET_BLOCKS_CONVERT_UNWRITTEN flags are passed. What do you think?

Yes, I did notice that as well after rebasing on your changes.

So the next patch enforces the behavior that if no flag is passed to ext4_split_extent() -> ext4_split_extent_zeroout() then we assume a split without conversion. As you mentioned, there is no remaining caller that does this but I thought of handling it here so that in future if we ever need to use unwrit to unwrit splits we handle it correctly.

Incase you still feel this makes it confusing or is uneccessary I can remove the else part altoghether and add a WARN_ON.

Thanks,
ojaswin

Quoted 36 lines

Thanks,
Yi.

Quoted 29 lines

fs/ext4/extents.c +26 -0

+		if (!is_unwrit)
+			/* Shouldn't happen. Just exit */
+			return ERR_PTR(-EINVAL);
+
+		lblk = ee_block;
+		len = ee_len;
+		pblk = ext4_ext_pblock(ex);
+		err = ext4_issue_zeroout(inode, lblk, pblk, len);
+		if (err)
+			/* ZEROOUT failed, just return original error */
+			return ERR_PTR(err);
+	}
+
+	err = ext4_ext_get_access(handle, inode, path + depth);
+	if (err)
+		return ERR_PTR(err);
+
+	ext4_ext_mark_initialized(ex);
+
+	ext4_ext_dirty(handle, inode, path + path->p_depth);
+	if (err)
+		return ERR_PTR(err);
+
+	return 0;
+}
+
 /*
  * ext4_split_extent() splits an extent and mark extent which is covered
  * by @map as split_flags indicates

[...]

Zhang Yi

2026-01-08 12:54 UTC #15 lore raw

Re: [PATCH 5/7] ext4: Refactor zeroout path and handle all cases

Quoted 139 lines · Ojaswin Mujoo

Quoted 64 lines · Zhang Yi
Quoted 7 lines · Ojaswin Mujoo
Currently, zeroout is used as a fallback in case we fail to split/convert extents in the "traditional" modify-the-extent-tree way. This is essential to mitigate failures in critical paths like extent splitting during endio. However, the logic is very messy and not easy to follow. Further, the fragile use of various flags has made it prone to errors.
Quoted 7 lines
Refactor zeroout out logic by moving it up to ext4_split_extents(). Further, zeroout correctly based on the type of conversion we want, ie:
- unwritten to written: Zeroout everything around the mapped range.
- unwritten to unwritten: Zeroout everything
- written to unwritten: Zeroout only the mapped range.
Signed-off-by Ojaswin Mujoo <[email protected]>
Hi, Ojaswin!
The refactor overall looks good to me. After this series, the split logic becomes more straightforward and clear. :)
I have some comments below.
Quoted 8 lines
---
 fs/ext4/extents.c | 287 +++++++++++++++++++++++++++++++---------------
 1 file changed, 195 insertions(+), 92 deletions(-)
fs/ext4/extents.c +0 -0
index 460a70e6dae0..8082e1d93bbf 100644
[...]
Quoted 30 lines
fs/ext4/extents.c +26 -0
@@ -3365,6 +3313,115 @@ static struct ext4_ext_path *ext4_split_extent_at(handle_t *handle,
 	return path;
 }
 
+static struct ext4_ext_path *
+ext4_split_extent_zeroout(handle_t *handle, struct inode *inode,
+			  struct ext4_ext_path *path,
+			  struct ext4_map_blocks *map, int flags)
+{
+	struct ext4_extent *ex;
+	unsigned int ee_len, depth;
+	ext4_lblk_t ee_block;
+	uint64_t lblk, pblk, len;
+	int is_unwrit;
+	int err = 0;
+
+	depth = ext_depth(inode);
+	ex = path[depth].p_ext;
+	ee_block = le32_to_cpu(ex->ee_block);
+	ee_len = ext4_ext_get_actual_len(ex);
+	is_unwrit = ext4_ext_is_unwritten(ex);
+
+	if (flags & EXT4_GET_BLOCKS_CONVERT) {
+		/*
+		 * EXT4_GET_BLOCKS_CONVERT: Caller wants the range specified by
+		 * map to be initialized. Zeroout everything except the map
+		 * range.
+		 */
+
+		loff_t map_end = (loff_t) map->m_lblk + map->m_len;
Quoted 8 lines
fs/ext4/extents.c static struct ext4_ext_path *ext4_split_extent_at(handle_t *handle, · line 3346
Quoted 5 lines
fs/ext4/extents.c +5 -0
+		loff_t ex_end = (loff_t) ee_block + ee_len;
+
+		if (!is_unwrit)
+			/* Shouldn't happen. Just exit */
+			return ERR_PTR(-EINVAL);
For cases that are should not happen, I'd suggest adding a WARN_ON_ONCE or a message to facilitate future problem identification. Same as below.
Hi Yi,
Thanks for the review! Sure I can do that in v2.
Quoted 51 lines
Quoted 51 lines
fs/ext4/extents.c +50 -0
+
+		/* zeroout left */
+		if (map->m_lblk > ee_block) {
+			lblk = ee_block;
+			len = map->m_lblk - ee_block;
+			pblk = ext4_ext_pblock(ex);
+			err = ext4_issue_zeroout(inode, lblk, pblk, len);
+			if (err)
+				/* ZEROOUT failed, just return original error */
+				return ERR_PTR(err);
+		}
+
+		/* zeroout right */
+		if (map->m_lblk + map->m_len < ee_block + ee_len) {
+			lblk = map_end;
+			len = ex_end - map_end;
+			pblk = ext4_ext_pblock(ex) + (map_end - ee_block);
+			err = ext4_issue_zeroout(inode, lblk, pblk, len);
+			if (err)
+				/* ZEROOUT failed, just return original error */
+				return ERR_PTR(err);
+		}
+	} else if (flags & EXT4_GET_BLOCKS_CONVERT_UNWRITTEN) {
+		/*
+		 * EXT4_GET_BLOCKS_CONVERT_UNWRITTEN: Caller wants the
+		 * range specified by map to be marked unwritten.
+		 * Zeroout the map range leaving rest as it is.
+		 */
+
+		if (is_unwrit)
+			/* Shouldn't happen. Just exit */
+			return ERR_PTR(-EINVAL);
+
+		lblk = map->m_lblk;
+		len = map->m_len;
+		pblk = ext4_ext_pblock(ex) + (map->m_lblk - ee_block);
+		err = ext4_issue_zeroout(inode, lblk, pblk, len);
+		if (err)
+			/* ZEROOUT failed, just return original error */
+			return ERR_PTR(err);
+	} else if (flags & EXT4_GET_BLOCKS_UNWRIT_EXT) {
+		/*
+		 * EXT4_GET_BLOCKS_UNWRIT_EXT: Today, this flag
+		 * implicitly implies that callers when wanting an
+		 * unwritten to unwritten split. So zeroout the whole
+		 * extent.
+		 *
+		 * TODO: The implicit meaning of the flag is not ideal
+		 * and eventually we should aim for a more well defined
+		 * behavior
Quoted 8 lines
fs/ext4/extents.c static struct ext4_ext_path *ext4_split_extent_at(handle_t *handle, · line 3398
Quoted 2 lines
fs/ext4/extents.c +2 -0
+		 */
+
I don't think we need this branch anymore. After applying my patch "ext4:
don't split extent before submitting I/O", we will no longer encounter situations where doing an unwritten to unwritten split. It means that at all call sites of ext4_split_extent(), only EXT4_GET_BLOCKS_CONVERT or EXT4_GET_BLOCKS_CONVERT_UNWRITTEN flags are passed. What do you think?
Yes, I did notice that as well after rebasing on your changes.

Quoted 9 lines

So the next patch enforces the behavior that if no flag is passed to ext4_split_extent() -> ext4_split_extent_zeroout() then we assume a split without conversion. As you mentioned, there is no remaining caller that does this but I thought of handling it here so that in future if we ever need to use unwrit to unwrit splits we handle it correctly.
Incase you still feel this makes it confusing or is uneccessary I can remove the else part altoghether and add a WARN_ON.

Yes, my personal suggestion is to add this part of the logic only when it is really needed. :)

Cheers,
Yi.

Quoted 39 lines

Thanks,
ojaswin

Quoted 36 lines

Thanks,
Yi.

Quoted 29 lines

fs/ext4/extents.c +26 -0

+		if (!is_unwrit)
+			/* Shouldn't happen. Just exit */
+			return ERR_PTR(-EINVAL);
+
+		lblk = ee_block;
+		len = ee_len;
+		pblk = ext4_ext_pblock(ex);
+		err = ext4_issue_zeroout(inode, lblk, pblk, len);
+		if (err)
+			/* ZEROOUT failed, just return original error */
+			return ERR_PTR(err);
+	}
+
+	err = ext4_ext_get_access(handle, inode, path + depth);
+	if (err)
+		return ERR_PTR(err);
+
+	ext4_ext_mark_initialized(ex);
+
+	ext4_ext_dirty(handle, inode, path + path->p_depth);
+	if (err)
+		return ERR_PTR(err);
+
+	return 0;
+}
+
 /*
  * ext4_split_extent() splits an extent and mark extent which is covered
  * by @map as split_flags indicates

[...]

Ojaswin Mujoo

author reviewer

2026-01-08 14:09 UTC #16 lore raw

Re: [PATCH 5/7] ext4: Refactor zeroout path and handle all cases

Quoted 146 lines · Zhang Yi

Quoted 145 lines · Ojaswin Mujoo
Quoted 64 lines · Zhang Yi
Quoted 7 lines · Ojaswin Mujoo
Currently, zeroout is used as a fallback in case we fail to split/convert extents in the "traditional" modify-the-extent-tree way. This is essential to mitigate failures in critical paths like extent splitting during endio. However, the logic is very messy and not easy to follow. Further, the fragile use of various flags has made it prone to errors.
Quoted 7 lines
Refactor zeroout out logic by moving it up to ext4_split_extents(). Further, zeroout correctly based on the type of conversion we want, ie:
- unwritten to written: Zeroout everything around the mapped range.
- unwritten to unwritten: Zeroout everything
- written to unwritten: Zeroout only the mapped range.
Signed-off-by Ojaswin Mujoo <[email protected]>
Hi, Ojaswin!
The refactor overall looks good to me. After this series, the split logic becomes more straightforward and clear. :)
I have some comments below.
Quoted 8 lines
---
 fs/ext4/extents.c | 287 +++++++++++++++++++++++++++++++---------------
 1 file changed, 195 insertions(+), 92 deletions(-)
fs/ext4/extents.c +0 -0
index 460a70e6dae0..8082e1d93bbf 100644
[...]
Quoted 30 lines
fs/ext4/extents.c +26 -0
@@ -3365,6 +3313,115 @@ static struct ext4_ext_path *ext4_split_extent_at(handle_t *handle,
 	return path;
 }
 
+static struct ext4_ext_path *
+ext4_split_extent_zeroout(handle_t *handle, struct inode *inode,
+			  struct ext4_ext_path *path,
+			  struct ext4_map_blocks *map, int flags)
+{
+	struct ext4_extent *ex;
+	unsigned int ee_len, depth;
+	ext4_lblk_t ee_block;
+	uint64_t lblk, pblk, len;
+	int is_unwrit;
+	int err = 0;
+
+	depth = ext_depth(inode);
+	ex = path[depth].p_ext;
+	ee_block = le32_to_cpu(ex->ee_block);
+	ee_len = ext4_ext_get_actual_len(ex);
+	is_unwrit = ext4_ext_is_unwritten(ex);
+
+	if (flags & EXT4_GET_BLOCKS_CONVERT) {
+		/*
+		 * EXT4_GET_BLOCKS_CONVERT: Caller wants the range specified by
+		 * map to be initialized. Zeroout everything except the map
+		 * range.
+		 */
+
+		loff_t map_end = (loff_t) map->m_lblk + map->m_len;
Quoted 8 lines
fs/ext4/extents.c static struct ext4_ext_path *ext4_split_extent_at(handle_t *handle, · line 3346
Quoted 5 lines
fs/ext4/extents.c +5 -0
+		loff_t ex_end = (loff_t) ee_block + ee_len;
+
+		if (!is_unwrit)
+			/* Shouldn't happen. Just exit */
+			return ERR_PTR(-EINVAL);
For cases that are should not happen, I'd suggest adding a WARN_ON_ONCE or a message to facilitate future problem identification. Same as below.
Hi Yi,
Thanks for the review! Sure I can do that in v2.
Quoted 51 lines
Quoted 51 lines
fs/ext4/extents.c +50 -0
+
+		/* zeroout left */
+		if (map->m_lblk > ee_block) {
+			lblk = ee_block;
+			len = map->m_lblk - ee_block;
+			pblk = ext4_ext_pblock(ex);
+			err = ext4_issue_zeroout(inode, lblk, pblk, len);
+			if (err)
+				/* ZEROOUT failed, just return original error */
+				return ERR_PTR(err);
+		}
+
+		/* zeroout right */
+		if (map->m_lblk + map->m_len < ee_block + ee_len) {
+			lblk = map_end;
+			len = ex_end - map_end;
+			pblk = ext4_ext_pblock(ex) + (map_end - ee_block);
+			err = ext4_issue_zeroout(inode, lblk, pblk, len);
+			if (err)
+				/* ZEROOUT failed, just return original error */
+				return ERR_PTR(err);
+		}
+	} else if (flags & EXT4_GET_BLOCKS_CONVERT_UNWRITTEN) {
+		/*
+		 * EXT4_GET_BLOCKS_CONVERT_UNWRITTEN: Caller wants the
+		 * range specified by map to be marked unwritten.
+		 * Zeroout the map range leaving rest as it is.
+		 */
+
+		if (is_unwrit)
+			/* Shouldn't happen. Just exit */
+			return ERR_PTR(-EINVAL);
+
+		lblk = map->m_lblk;
+		len = map->m_len;
+		pblk = ext4_ext_pblock(ex) + (map->m_lblk - ee_block);
+		err = ext4_issue_zeroout(inode, lblk, pblk, len);
+		if (err)
+			/* ZEROOUT failed, just return original error */
+			return ERR_PTR(err);
+	} else if (flags & EXT4_GET_BLOCKS_UNWRIT_EXT) {
+		/*
+		 * EXT4_GET_BLOCKS_UNWRIT_EXT: Today, this flag
+		 * implicitly implies that callers when wanting an
+		 * unwritten to unwritten split. So zeroout the whole
+		 * extent.
+		 *
+		 * TODO: The implicit meaning of the flag is not ideal
+		 * and eventually we should aim for a more well defined
+		 * behavior
Quoted 8 lines
fs/ext4/extents.c static struct ext4_ext_path *ext4_split_extent_at(handle_t *handle, · line 3398
Quoted 2 lines
fs/ext4/extents.c +2 -0
+		 */
+
I don't think we need this branch anymore. After applying my patch "ext4:
don't split extent before submitting I/O", we will no longer encounter situations where doing an unwritten to unwritten split. It means that at all call sites of ext4_split_extent(), only EXT4_GET_BLOCKS_CONVERT or EXT4_GET_BLOCKS_CONVERT_UNWRITTEN flags are passed. What do you think?
Yes, I did notice that as well after rebasing on your changes.
So the next patch enforces the behavior that if no flag is passed to ext4_split_extent() -> ext4_split_extent_zeroout() then we assume a split without conversion. As you mentioned, there is no remaining caller that does this but I thought of handling it here so that in future if we ever need to use unwrit to unwrit splits we handle it correctly.

Quoted 6 lines

Quoted 3 lines
Incase you still feel this makes it confusing or is uneccessary I can remove the else part altoghether and add a WARN_ON.
Yes, my personal suggestion is to add this part of the logic only when it is really needed. :)

Okay sounds good, will take this out in v2.

Regards,
ojaswin

Quoted 44 lines

Cheers,
Yi.

Quoted 39 lines

Thanks,
ojaswin

Quoted 36 lines

Thanks,
Yi.

Quoted 29 lines

fs/ext4/extents.c +26 -0

+		if (!is_unwrit)
+			/* Shouldn't happen. Just exit */
+			return ERR_PTR(-EINVAL);
+
+		lblk = ee_block;
+		len = ee_len;
+		pblk = ext4_ext_pblock(ex);
+		err = ext4_issue_zeroout(inode, lblk, pblk, len);
+		if (err)
+			/* ZEROOUT failed, just return original error */
+			return ERR_PTR(err);
+	}
+
+	err = ext4_ext_get_access(handle, inode, path + depth);
+	if (err)
+		return ERR_PTR(err);
+
+	ext4_ext_mark_initialized(ex);
+
+	ext4_ext_dirty(handle, inode, path + path->p_depth);
+	if (err)
+		return ERR_PTR(err);
+
+	return 0;
+}
+
 /*
  * ext4_split_extent() splits an extent and mark extent which is covered
  * by @map as split_flags indicates

[...]

Ojaswin Mujoo

author reviewer

2026-01-12 09:42 UTC #20 lore raw

Re: [PATCH 6/7] ext4: Refactor split and convert extents

In reply to Zhang Yi

Quoted 80 lines · Zhang Yi

Quoted 30 lines · Ojaswin Mujoo
ext4_split_convert_extents() has been historically prone to subtle bugs and inconsistent behavior due to the way all the various flags interact with the extent split and conversion process. For example, callers like ext4_convert_unwritten_extents_endio() and convert_initialized_extents() needed to open code extent conversion despite passing CONVERT or CONVERT_UNWRITTEN flags because ext4_split_convert_extents() wasn't performing the conversion.
Hence, refactor ext4_split_convert_extents() to clearly enforce the semantics of each flag. The major changes here are:
* Clearly separate the split and convert process:
* ext4_split_extent() and ext4_split_extent_at() are now only responsible to perform the split.
* ext4_split_convert_extents() is now responsible to perform extent conversion after calling ext4_split_extent() for splitting.
* This helps get rid of all the MARK_UNWRIT* flags.
* Clearly enforce the semantics of flags passed to ext4_split_convert_extents():
* EXT4_GET_BLOCKS_CONVERT: Will convert the split extent to written
* EXT4_GET_BLOCKS_CONVERT_UNWRITTEN: Will convert the split extent to unwritten
* Passing neither of the above means we only want a split.
* Modify all callers to enforce the above semantics.
* Use ext4_split_convert_extents() instead of ext4_split_extents()
* in ext4_ext_convert_to_initialized() for uniformity.
Quoted 7 lines
* Cleanup all callers open coding the conversion logic.
* Further, modify kuniy tests to pass flags based on the new semantics.
From an end user point of view, we should not see any changes in behavior of ext4.
Signed-off-by Ojaswin Mujoo <[email protected]>
Some comments below.
Quoted 5 lines
---
 fs/ext4/extents-test.c |  12 +-
 fs/ext4/extents.c      | 299 +++++++++++++++++++----------------------
 2 files changed, 145 insertions(+), 166 deletions(-)
[..]
Quoted 31 lines
fs/ext4/extents.c +20 -0
@@ -3820,6 +3786,26 @@ ext4_ext_convert_to_initialized(handle_t *handle, struct inode *inode,
 	return ERR_PTR(err);
 }
 
+static bool ext4_ext_needs_conv(struct inode *inode, struct ext4_ext_path *path,
+				int flags)
+{
+	struct ext4_extent *ex;
+	bool is_unwrit;
+	int depth;
+
+	depth = ext_depth(inode);
+	ex = path[depth].p_ext;
+	is_unwrit = ext4_ext_is_unwritten(ex);
+
+	if (is_unwrit && (flags & EXT4_GET_BLOCKS_CONVERT))
+		return true;
+
+	if (!is_unwrit && (flags & EXT4_GET_BLOCKS_CONVERT_UNWRITTEN))
+		return true;
+
+	return false;
+}
+
 /*
  * This function is called by ext4_ext_map_blocks() from
  * ext4_get_blocks_dio_write() when DIO to write
@@ -3856,7 +3842,9 @@ static struct ext4_ext_path *ext4_split_convert_extents(handle_t *handle,
 	ext4_lblk_t ee_block;
 	struct ext4_extent *ex;
 	unsigned int ee_len;

Quoted 8 lines

fs/ext4/extents.c static struct ext4_ext_path *ext4_split_convert_extents(handle_t *handle, · line 3847
Quoted 4 lines
fs/ext4/extents.c +3 -1
-	int split_flag = 0, depth;
+	int split_flag = 0, depth, err = 0;
+	bool did_zeroout = false;
+	bool needs_conv = ext4_ext_needs_conv(inode, path, flags);
As I described in Patch 05, there is currently no situation where splitting occurs without conversion, so I don't think we need this check. Is it right?

Hey Yi,

Yes I can get rid of this.

Quoted 87 lines

Quoted 72 lines

fs/ext4/extents.c +55 -7

 	ext_debug(inode, "logical block %llu, max_blocks %u\n",
 		  (unsigned long long)map->m_lblk, map->m_len);

@@ -3870,19 +3858,81 @@ static struct ext4_ext_path *ext4_split_convert_extents(handle_t *handle,
 	ee_block = le32_to_cpu(ex->ee_block);
 	ee_len = ext4_ext_get_actual_len(ex);
 
-	if (flags & (EXT4_GET_BLOCKS_UNWRIT_EXT |
-			    EXT4_GET_BLOCKS_CONVERT)) {
+	/* No split needed */
+	if (ee_block == map->m_lblk && ee_len == map->m_len)
+		goto convert;
+
+	/*
+	 * We don't use zeroout fallback for written to unwritten conversion as
+	 * it is not as critical as endio and it might take unusually long.
+	 * Also, it is only safe to convert extent to initialized via explicit
+	 * zeroout only if extent is fully inside i_size or new_size.
+	 */
+	if (!(flags & EXT4_GET_BLOCKS_CONVERT_UNWRITTEN))
+		split_flag |= ee_block + ee_len <= eof_block ?
+				      EXT4_EXT_MAY_ZEROOUT :
+				      0;
+
+	/*
+	 * pass SPLIT_NOMERGE explicitly so we don't end up merging extents we
+	 * just split.
+	 */
+	path = ext4_split_extent(handle, inode, path, map, split_flag,
+				 flags | EXT4_GET_BLOCKS_SPLIT_NOMERGE,
+				 allocated, &did_zeroout);
+
+convert:
+	/*
+	 * We don't need a conversion if:
+	 * 1. There was an error in split.
+	 * 2. We split via zeroout.
+	 * 3. None of the convert flags were passed.
+	 */
+	if (IS_ERR(path) || did_zeroout || !needs_conv)
+		return path;
+
+	path = ext4_find_extent(inode, map->m_lblk, path, flags);
+	if (IS_ERR(path))
+		return path;
+
+	depth = ext_depth(inode);
+	ex = path[depth].p_ext;
+
+	err = ext4_ext_get_access(handle, inode, path + depth);
+	if (err)
+		goto err;
+
+	if (flags & EXT4_GET_BLOCKS_CONVERT)
+		ext4_ext_mark_initialized(ex);
+	else if (flags & EXT4_GET_BLOCKS_CONVERT_UNWRITTEN)
+		ext4_ext_mark_unwritten(ex);
+
+	if (!(flags & EXT4_GET_BLOCKS_SPLIT_NOMERGE))
 		/*
-		 * It is safe to convert extent to initialized via explicit
-		 * zeroout only if extent is fully inside i_size or new_size.
+		 * note: ext4_ext_correct_indexes() isn't needed here because
+		 * borders are not changed
 		 */
-		split_flag |= ee_block + ee_len <= eof_block ?
-			      EXT4_EXT_MAY_ZEROOUT : 0;
-		split_flag |= EXT4_EXT_MARK_UNWRIT2;
+		ext4_ext_try_to_merge(handle, inode, path, ex);
+
+	err = ext4_ext_dirty(handle, inode, path + depth);

fs/ext4/extents.c static struct ext4_ext_path *ext4_split_convert_extents(handle_t *handle, · line 3927

Quoted 10 lines

fs/ext4/extents.c +10 -0

+	if (err)
+		goto err;
+
+	/* Lets update the extent status tree after conversion */
+	ext4_es_insert_extent(inode, le32_to_cpu(ex->ee_block),
+			      ext4_ext_get_actual_len(ex), ext4_ext_pblock(ex),
+			      ext4_ext_is_unwritten(ex) ?
+				      EXTENT_STATUS_UNWRITTEN :
+				      EXTENT_STATUS_WRITTEN,
+			      false);

Quoted 4 lines

We should be careful about the extent status tree. I'd suggest that pay close attention to the following error output when running tests,
EXT4-fs warning (device pmem2s): ext4_es_cache_extent:1079: inode #718: comm 108573.fsstress: ES cache extent failed: add [55,22,12260,0x1] conflict with existing [75,2,12280,0x2]

Yes I did see this line during testing earlier but caching it like above fixed all the of em. Yes I did seem to miss caching for the zeroout case. The thing is that as long as nocache is not passed we usually end up caching it correctly in one of the ext4_find_extent() calls and I guess that's why I might have missed it.

Also, I think I should also check for NOCACHE flag here before caching it here right, since many caller pass the NOCACHE flag.

I've also added es_cache support to kunit tests so hopefully we will catch any issues there. Will add these changes in v2.

Regards,
ojaswin

Quoted 163 lines

or directly add a WARN_ON_ONCE(1) at the end of ext4_es_cache_extent(). There may be other problems related to the stale extents that could arise.

Thanks,
Yi.

Quoted 154 lines

fs/ext4/extents.c +16 ~~-83~~

+
+err:
+	if (err) {
+		ext4_free_ext_path(path);
+		return ERR_PTR(err);
 	}
-	flags |= EXT4_GET_BLOCKS_SPLIT_NOMERGE;
-	return ext4_split_extent(handle, inode, path, map, split_flag, flags,
-				 allocated);
+
+	return path;
 }
 
 static struct ext4_ext_path *

@@ -3894,7 +3944,6 @@ ext4_convert_unwritten_extents_endio(handle_t *handle, struct inode *inode,
 	ext4_lblk_t ee_block;
 	unsigned int ee_len;
 	int depth;
-	int err = 0;
 
 	depth = ext_depth(inode);
 	ex = path[depth].p_ext;

@@ -3904,41 +3953,8 @@ ext4_convert_unwritten_extents_endio(handle_t *handle, struct inode *inode,
 	ext_debug(inode, "logical block %llu, max_blocks %u\n",
 		  (unsigned long long)ee_block, ee_len);
 
-	if (ee_block != map->m_lblk || ee_len > map->m_len) {
-		path = ext4_split_convert_extents(handle, inode, map, path,
-						  flags, NULL);
-		if (IS_ERR(path))
-			return path;
-
-		path = ext4_find_extent(inode, map->m_lblk, path, 0);
-		if (IS_ERR(path))
-			return path;
-		depth = ext_depth(inode);
-		ex = path[depth].p_ext;
-	}
-
-	err = ext4_ext_get_access(handle, inode, path + depth);
-	if (err)
-		goto errout;
-	/* first mark the extent as initialized */
-	ext4_ext_mark_initialized(ex);
-
-	/* note: ext4_ext_correct_indexes() isn't needed here because
-	 * borders are not changed
-	 */
-	ext4_ext_try_to_merge(handle, inode, path, ex);
-
-	/* Mark modified extent as dirty */
-	err = ext4_ext_dirty(handle, inode, path + path->p_depth);
-	if (err)
-		goto errout;
-
-	ext4_ext_show_leaf(inode, path);
-	return path;
-
-errout:
-	ext4_free_ext_path(path);
-	return ERR_PTR(err);
+	return ext4_split_convert_extents(handle, inode, map, path, flags,
+					  NULL);
 }
 
 static struct ext4_ext_path *

@@ -3952,7 +3968,6 @@ convert_initialized_extent(handle_t *handle, struct inode *inode,
 	ext4_lblk_t ee_block;
 	unsigned int ee_len;
 	int depth;
-	int err = 0;
 
 	/*
 	 * Make sure that the extent is no bigger than we support with

@@ -3969,40 +3984,12 @@ convert_initialized_extent(handle_t *handle, struct inode *inode,
 	ext_debug(inode, "logical block %llu, max_blocks %u\n",
 		  (unsigned long long)ee_block, ee_len);
 
-	if (ee_block != map->m_lblk || ee_len > map->m_len) {
-		path = ext4_split_convert_extents(handle, inode, map, path,
-				flags | EXT4_GET_BLOCKS_CONVERT_UNWRITTEN, NULL);
-		if (IS_ERR(path))
-			return path;
-
-		path = ext4_find_extent(inode, map->m_lblk, path, 0);
-		if (IS_ERR(path))
-			return path;
-		depth = ext_depth(inode);
-		ex = path[depth].p_ext;
-		if (!ex) {
-			EXT4_ERROR_INODE(inode, "unexpected hole at %lu",
-					 (unsigned long) map->m_lblk);
-			err = -EFSCORRUPTED;
-			goto errout;
-		}
-	}
-
-	err = ext4_ext_get_access(handle, inode, path + depth);
-	if (err)
-		goto errout;
-	/* first mark the extent as unwritten */
-	ext4_ext_mark_unwritten(ex);
-
-	/* note: ext4_ext_correct_indexes() isn't needed here because
-	 * borders are not changed
-	 */
-	ext4_ext_try_to_merge(handle, inode, path, ex);
+	path = ext4_split_convert_extents(
+		handle, inode, map, path,
+		flags | EXT4_GET_BLOCKS_CONVERT_UNWRITTEN, NULL);
+	if (IS_ERR(path))
+		return path;
 
-	/* Mark modified extent as dirty */
-	err = ext4_ext_dirty(handle, inode, path + path->p_depth);
-	if (err)
-		goto errout;
 	ext4_ext_show_leaf(inode, path);
 
 	ext4_update_inode_fsync_trans(handle, inode, 1);

@@ -4012,10 +3999,6 @@ convert_initialized_extent(handle_t *handle, struct inode *inode,
 		*allocated = map->m_len;
 	map->m_len = *allocated;
 	return path;
-
-errout:
-	ext4_free_ext_path(path);
-	return ERR_PTR(err);
 }
 
 static struct ext4_ext_path *

@@ -5649,7 +5632,7 @@ static int ext4_insert_range(struct file *file, loff_t offset, loff_t len)
 	struct ext4_extent *extent;
 	ext4_lblk_t start_lblk, len_lblk, ee_start_lblk = 0;
 	unsigned int credits, ee_len;
-	int ret, depth, split_flag = 0;
+	int ret, depth;
 	loff_t start;
 
 	trace_ext4_insert_range(inode, offset, len);

@@ -5720,12 +5703,8 @@ static int ext4_insert_range(struct file *file, loff_t offset, loff_t len)
 		 */
 		if ((start_lblk > ee_start_lblk) &&
 				(start_lblk < (ee_start_lblk + ee_len))) {
-			if (ext4_ext_is_unwritten(extent))
-				split_flag = EXT4_EXT_MARK_UNWRIT1 |
-					EXT4_EXT_MARK_UNWRIT2;
 			path = ext4_split_extent_at(handle, inode, path,
-					start_lblk, split_flag,
-					EXT4_EX_NOCACHE |
+					start_lblk, EXT4_EX_NOCACHE |
 					EXT4_GET_BLOCKS_SPLIT_NOMERGE |
 					EXT4_GET_BLOCKS_METADATA_NOFAIL);
 		}