Skip to contentSkip to conversation

ioctl_loop06: update loopback block size validation

PATCH Superseded Li Wang 4 participants 12 messages Mar 13, 2025 – Apr 11, 2025

lore
Share & tools
lore

Li Wang

author
#0 lore raw

PATCH v2 1/1 ioctl_loop06: update loopback block size validation

The kernel commit 47dd6753 ("block/bdev: lift block size restrictions to 64k") now supports block sizes larger than PAGE_SIZE, with a new upper limit of BLK_MAX_BLOCK_SIZE (64K). But ioctl_loop06 still assumes that PAGE_SIZE is the maximum allowed block size, causing failures on newer kernels(>= 6.14):

ioctl_loop06.c:74: TINFO: Using LOOP_SET_BLOCK_SIZE with arg > PAGE_SIZE ioctl_loop06.c:57: TFAIL: Set block size succeed unexpectedly
...
ioctl_loop06.c:74: TINFO: Using LOOP_CONFIGURE with block_size > PAGE_SIZE ioctl_loop06.c:57: TFAIL: Set block size succeed unexpectedly

This patch updates ioctl_loop06 to use BLK_MAX_BLOCK_SIZE instead of PAGE_SIZE for block size validation.

And, we give up the test size between (PAGE_SIZE, BLK_MAX_BLOCK_SIZE) even on old kernels becase it makes no sense.

Cc Luis Chamberlain <[email protected]>
Cc Christoph Hellwig <[email protected]>
Cc Hannes Reinecke <[email protected]>
---
 testcases/kernel/syscalls/ioctl/ioctl_loop06.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)
testcases/kernel/syscalls/ioctl/ioctl_loop06.c +5 -3
index 573871bc1..238a7c6f5 100644
@@ -17,6 +17,8 @@
 #include "lapi/loop.h"
 #include "tst_test.h"
 
+#define BLK_MAX_BLOCK_SIZE 0x00010000 /* SZ_64K */
+
 static char dev_path[1024];
 static int dev_num, dev_fd, file_fd, attach_flag, loop_configure_sup = 1;
 static unsigned int invalid_value, half_value, unalign_value;
@@ -31,7 +33,7 @@ static struct tcase {
 	"Using LOOP_SET_BLOCK_SIZE with arg < 512"},
 
 	{&invalid_value, LOOP_SET_BLOCK_SIZE,
-	"Using LOOP_SET_BLOCK_SIZE with arg > PAGE_SIZE"},
+	"Using LOOP_SET_BLOCK_SIZE with arg > BLK_MAX_BLOCK_SIZE"},
 
 	{&unalign_value, LOOP_SET_BLOCK_SIZE,
 	"Using LOOP_SET_BLOCK_SIZE with arg != power_of_2"},
@@ -40,7 +42,7 @@ static struct tcase {
 	"Using LOOP_CONFIGURE with block_size < 512"},
 
 	{&invalid_value, LOOP_CONFIGURE,
-	"Using LOOP_CONFIGURE with block_size > PAGE_SIZE"},
+	"Using LOOP_CONFIGURE with block_size > BLK_MAX_BLOCK_SIZE"},
 
 	{&unalign_value, LOOP_CONFIGURE,
 	"Using LOOP_CONFIGURE with block_size != power_of_2"},
@@ -106,7 +108,7 @@ static void setup(void)
 	tst_fill_file("test.img", 0, 1024, 1024);
 	half_value = 256;
 	pg_size = getpagesize();
-	invalid_value = pg_size * 2 ;
+	invalid_value = BLK_MAX_BLOCK_SIZE + 1;
 	unalign_value = pg_size - 1;
 
 	dev_fd = SAFE_OPEN(dev_path, O_RDWR);
Signature · 5 lines
2.48.1


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
#1 lore raw
Quoted 30 lines · Li Wang

The kernel commit 47dd6753 ("block/bdev: lift block size restrictions to 64k") now supports block sizes larger than PAGE_SIZE, with a new upper limit of BLK_MAX_BLOCK_SIZE (64K). But ioctl_loop06 still assumes that PAGE_SIZE is the maximum allowed block size, causing failures on newer kernels(>= 6.14):

ioctl_loop06.c:74: TINFO: Using LOOP_SET_BLOCK_SIZE with arg > PAGE_SIZE ioctl_loop06.c:57: TFAIL: Set block size succeed unexpectedly
...
ioctl_loop06.c:74: TINFO: Using LOOP_CONFIGURE with block_size > PAGE_SIZE ioctl_loop06.c:57: TFAIL: Set block size succeed unexpectedly

This patch updates ioctl_loop06 to use BLK_MAX_BLOCK_SIZE instead of PAGE_SIZE for block size validation.

And, we give up the test size between (PAGE_SIZE, BLK_MAX_BLOCK_SIZE) even on old kernels becase it makes no sense.

Cc Luis Chamberlain <[email protected]>
Cc Christoph Hellwig <[email protected]>
Cc Hannes Reinecke <[email protected]>
---
 testcases/kernel/syscalls/ioctl/ioctl_loop06.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)
testcases/kernel/syscalls/ioctl/ioctl_loop06.c +0 -0
index 573871bc1..238a7c6f5 100644

testcases/kernel/syscalls/ioctl/ioctl_loop06.c line 20

Quoted 4 lines
testcases/kernel/syscalls/ioctl/ioctl_loop06.c +1 -0
@@ -17,6 +17,8 @@
 #include "lapi/loop.h"
 #include "tst_test.h"
 
+#define BLK_MAX_BLOCK_SIZE 0x00010000 /* SZ_64K */

Today its SZ_64K, tomorrow another value. If you have an ifdef check for kernel versions it may be good to do that. So if >= v6.15 then 64k othersize PAGE_SIZE is fine.

Quoted 23 lines
testcases/kernel/syscalls/ioctl/ioctl_loop06.c +3 -2
+
 static char dev_path[1024];
 static int dev_num, dev_fd, file_fd, attach_flag, loop_configure_sup = 1;
 static unsigned int invalid_value, half_value, unalign_value;
@@ -31,7 +33,7 @@ static struct tcase {
 	"Using LOOP_SET_BLOCK_SIZE with arg < 512"},
 
 	{&invalid_value, LOOP_SET_BLOCK_SIZE,
-	"Using LOOP_SET_BLOCK_SIZE with arg > PAGE_SIZE"},
+	"Using LOOP_SET_BLOCK_SIZE with arg > BLK_MAX_BLOCK_SIZE"},
 
 	{&unalign_value, LOOP_SET_BLOCK_SIZE,
 	"Using LOOP_SET_BLOCK_SIZE with arg != power_of_2"},
@@ -40,7 +42,7 @@ static struct tcase {
 	"Using LOOP_CONFIGURE with block_size < 512"},
 
 	{&invalid_value, LOOP_CONFIGURE,
-	"Using LOOP_CONFIGURE with block_size > PAGE_SIZE"},
+	"Using LOOP_CONFIGURE with block_size > BLK_MAX_BLOCK_SIZE"},
 
 	{&unalign_value, LOOP_CONFIGURE,
 	"Using LOOP_CONFIGURE with block_size != power_of_2"},

testcases/kernel/syscalls/ioctl/ioctl_loop06.c static void setup(void) · line 111

Quoted 5 lines
testcases/kernel/syscalls/ioctl/ioctl_loop06.c +1 -1
@@ -106,7 +108,7 @@ static void setup(void)
 	tst_fill_file("test.img", 0, 1024, 1024);
 	half_value = 256;
 	pg_size = getpagesize();
-	invalid_value = pg_size * 2 ;
+	invalid_value = BLK_MAX_BLOCK_SIZE + 1;

I'd use BLK_MAX_BLOCK_SIZE * 2 as we use power of 2 values;

Luis

Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp
author
#2 lore raw

Hi Luis,

Quoted 8 lines · Luis Chamberlain
Quoted 1 line · Li Wang

The kernel commit 47dd6753 ("block/bdev: lift block size restrictions to

64k")

Quoted 2 lines

now supports block sizes larger than PAGE_SIZE, with a new upper limit of BLK_MAX_BLOCK_SIZE (64K). But ioctl_loop06 still assumes that PAGE_SIZE

is the

Quoted 1 line

maximum allowed block size, causing failures on newer kernels(>= 6.14):

Well, this sounds like we need to go back to patch v1:
https://lists.linux.it/pipermail/ltp/2025-March/042599.html

Note: LTP has a function tst_kvercmp() for comparing kernel versions.

Signature · 5 lines
Regards,
Li Wang

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
author
#3 lore raw

PATCH v3 1/1 ioctl_loop06: update loopback block size validation

The kernel commit 47dd6753 ("block/bdev: lift block size restrictions to 64k") now supports block sizes larger than PAGE_SIZE, with a new upper limit of BLK_MAX_BLOCK_SIZE (64K). But ioctl_loop06 still assumes that PAGE_SIZE is the maximum allowed block size, causing failures on newer kernels(>= 6.14):

ioctl_loop06.c:74: TINFO: Using LOOP_SET_BLOCK_SIZE with arg > PAGE_SIZE ioctl_loop06.c:57: TFAIL: Set block size succeed unexpectedly
...
ioctl_loop06.c:74: TINFO: Using LOOP_CONFIGURE with block_size > PAGE_SIZE ioctl_loop06.c:57: TFAIL: Set block size succeed unexpectedly

This patch updates ioctl_loop06 to use BLK_MAX_BLOCK_SIZE instead of PAGE_SIZE for block size validation. Also adjust failure expectations based on the running kernel version.

Cc Luis Chamberlain <[email protected]>
Cc Christoph Hellwig <[email protected]>
Cc Hannes Reinecke <[email protected]>

---

Notes:
v2 --> v3:
* adding back the kernel version check to compare max value
* use BLK_MAX_BLOCK_SIZE * 2 as Luis suggestted

 testcases/kernel/syscalls/ioctl/ioctl_loop06.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)
testcases/kernel/syscalls/ioctl/ioctl_loop06.c +5 -3
index 573871bc1..be0aa2506 100644
@@ -17,6 +17,8 @@
 #include "lapi/loop.h"
 #include "tst_test.h"
 
+#define BLK_MAX_BLOCK_SIZE 0x00010000 /* SZ_64K */
+
 static char dev_path[1024];
 static int dev_num, dev_fd, file_fd, attach_flag, loop_configure_sup = 1;
 static unsigned int invalid_value, half_value, unalign_value;
@@ -31,7 +33,7 @@ static struct tcase {
 	"Using LOOP_SET_BLOCK_SIZE with arg < 512"},
 
 	{&invalid_value, LOOP_SET_BLOCK_SIZE,
-	"Using LOOP_SET_BLOCK_SIZE with arg > PAGE_SIZE"},
+	"Using LOOP_SET_BLOCK_SIZE with arg > BLK_MAX_BLOCK_SIZE"},
 
 	{&unalign_value, LOOP_SET_BLOCK_SIZE,
 	"Using LOOP_SET_BLOCK_SIZE with arg != power_of_2"},
@@ -40,7 +42,7 @@ static struct tcase {
 	"Using LOOP_CONFIGURE with block_size < 512"},
 
 	{&invalid_value, LOOP_CONFIGURE,
-	"Using LOOP_CONFIGURE with block_size > PAGE_SIZE"},
+	"Using LOOP_CONFIGURE with block_size > BLK_MAX_BLOCK_SIZE"},
 
 	{&unalign_value, LOOP_CONFIGURE,
 	"Using LOOP_CONFIGURE with block_size != power_of_2"},
@@ -106,8 +108,8 @@ static void setup(void)
 	tst_fill_file("test.img", 0, 1024, 1024);
 	half_value = 256;
 	pg_size = getpagesize();
-	invalid_value = pg_size * 2 ;
 	unalign_value = pg_size - 1;
+	invalid_value = (tst_kvercmp(6, 14, 0) < 0) ? pg_size * 2 : BLK_MAX_BLOCK_SIZE * 2;
 
 	dev_fd = SAFE_OPEN(dev_path, O_RDWR);
 	ret = ioctl(dev_fd, LOOP_SET_BLOCK_SIZE, 512);
Signature · 5 lines
2.48.1


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
author
#4 lore raw

PATCH v4 1/1 ioctl_loop06: update loopback block size validation

The kernel commit 47dd6753 ("block/bdev: lift block size restrictions to 64k") now supports block sizes larger than PAGE_SIZE, with a new upper limit of BLK_MAX_BLOCK_SIZE (64K). But ioctl_loop06 still assumes that PAGE_SIZE is the maximum allowed block size, causing failures on newer kernels(>= 6.14):

ioctl_loop06.c:74: TINFO: Using LOOP_SET_BLOCK_SIZE with arg > PAGE_SIZE ioctl_loop06.c:57: TFAIL: Set block size succeed unexpectedly
...
ioctl_loop06.c:74: TINFO: Using LOOP_CONFIGURE with block_size > PAGE_SIZE ioctl_loop06.c:57: TFAIL: Set block size succeed unexpectedly

This patch updates ioctl_loop06 to use BLK_MAX_BLOCK_SIZE instead of PAGE_SIZE for block size validation.

And, dynamically sets bs based on BLK_MAX_BLOCK_SIZE, using 1024 bytes if it's below 1MB or scaling it otherwise. Ensures tst_fill_file() writes efficiently while maintaining compatibility across different kernel versions.

Cc Luis Chamberlain <[email protected]>
Cc Christoph Hellwig <[email protected]>
Cc Hannes Reinecke <[email protected]>
---
 configure.ac                                  |  1 +
 include/lapi/blkdev.h                         | 19 +++++++++++++++++++
 .../kernel/syscalls/ioctl/ioctl_loop06.c      | 12 ++++++++----
 3 files changed, 28 insertions(+), 4 deletions(-)
 create mode 100644 include/lapi/blkdev.h
configure.ac +1 -0
index 0f2b6f332..5538d88d5 100644
@@ -52,6 +52,7 @@ AC_CHECK_HEADERS_ONCE([ \
     emmintrin.h \
     ifaddrs.h \
     keyutils.h \
+    linux/blkdev.h \
     linux/can.h \
     linux/cgroupstats.h \
     linux/cryptouser.h \
include/lapi/blkdev.h new file +19 -0
new file mode 100644
index 000000000..3ee058ce0
@@ -0,0 +1,19 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Copyright (c) 2025 Linux Test Project
+ *  Li Wang <[email protected]>
+ */
+
+#ifndef LAPI_BLKDEV_H__
+#define LAPI_BLKDEV_H__
+
+#ifdef HAVE_LINUX_BLKDEV_H
+#include <linux/blkdev.h>
+#endif
+
+/* Define BLK_MAX_BLOCK_SIZE for older kernels */
+#ifndef BLK_MAX_BLOCK_SIZE
+#define BLK_MAX_BLOCK_SIZE 0x00010000 /* 64K */
+#endif
+
+#endif /* LAPI_BLKDEV_H */
testcases/kernel/syscalls/ioctl/ioctl_loop06.c +8 -4
index 573871bc1..35e9e79e9 100644
@@ -14,7 +14,9 @@
 #include <unistd.h>
 #include <sys/types.h>
 #include <stdlib.h>
+#include "lapi/blkdev.h"
 #include "lapi/loop.h"
+#include "tst_fs.h"
 #include "tst_test.h"
 
 static char dev_path[1024];
@@ -31,7 +33,7 @@ static struct tcase {
 	"Using LOOP_SET_BLOCK_SIZE with arg < 512"},
 
 	{&invalid_value, LOOP_SET_BLOCK_SIZE,
-	"Using LOOP_SET_BLOCK_SIZE with arg > PAGE_SIZE"},
+	"Using LOOP_SET_BLOCK_SIZE with arg > BLK_MAX_BLOCK_SIZE"},
 
 	{&unalign_value, LOOP_SET_BLOCK_SIZE,
 	"Using LOOP_SET_BLOCK_SIZE with arg != power_of_2"},
@@ -40,7 +42,7 @@ static struct tcase {
 	"Using LOOP_CONFIGURE with block_size < 512"},
 
 	{&invalid_value, LOOP_CONFIGURE,
-	"Using LOOP_CONFIGURE with block_size > PAGE_SIZE"},
+	"Using LOOP_CONFIGURE with block_size > BLK_MAX_BLOCK_SIZE"},
 
 	{&unalign_value, LOOP_CONFIGURE,
 	"Using LOOP_CONFIGURE with block_size != power_of_2"},
@@ -103,10 +105,12 @@ static void setup(void)
 	if (dev_num < 0)
 		tst_brk(TBROK, "Failed to find free loop device");
 
-	tst_fill_file("test.img", 0, 1024, 1024);
+	size_t bs = (BLK_MAX_BLOCK_SIZE < TST_MB) ? 1024 : 4 * BLK_MAX_BLOCK_SIZE / 1024;
+	tst_fill_file("test.img", 0, bs, 1024);
+
 	half_value = 256;
 	pg_size = getpagesize();
-	invalid_value = pg_size * 2 ;
+	invalid_value = BLK_MAX_BLOCK_SIZE * 2;
 	unalign_value = pg_size - 1;
 
 	dev_fd = SAFE_OPEN(dev_path, O_RDWR);
Signature · 5 lines
2.48.1


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
Reviewed-by on v4 1/1
#5 lore raw

Hi Li,

Quoted 11 lines

The kernel commit 47dd6753 ("block/bdev: lift block size restrictions to 64k") now supports block sizes larger than PAGE_SIZE, with a new upper limit of BLK_MAX_BLOCK_SIZE (64K). But ioctl_loop06 still assumes that PAGE_SIZE is the maximum allowed block size, causing failures on newer kernels(>= 6.14):

ioctl_loop06.c:74: TINFO: Using LOOP_SET_BLOCK_SIZE with arg > PAGE_SIZE ioctl_loop06.c:57: TFAIL: Set block size succeed unexpectedly
...
ioctl_loop06.c:74: TINFO: Using LOOP_CONFIGURE with block_size > PAGE_SIZE ioctl_loop06.c:57: TFAIL: Set block size succeed unexpectedly

Quoted 6 lines

This patch updates ioctl_loop06 to use BLK_MAX_BLOCK_SIZE instead of PAGE_SIZE for block size validation.

And, dynamically sets bs based on BLK_MAX_BLOCK_SIZE, using 1024 bytes if it's below 1MB or scaling it otherwise. Ensures tst_fill_file() writes efficiently while maintaining compatibility across different kernel versions.

LGTM.

Kind regards,
Petr

Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp
Tested-by on v4 1/1
#6 lore raw
In reply to Li Wang
Quoted 14 lines · CEST Li Wang

The kernel commit 47dd6753 ("block/bdev: lift block size restrictions to 64k") now supports block sizes larger than PAGE_SIZE, with a new upper limit of BLK_MAX_BLOCK_SIZE (64K). But ioctl_loop06 still assumes that PAGE_SIZE is the maximum allowed block size, causing failures on newer kernels(>= 6.14):

ioctl_loop06.c:74: TINFO: Using LOOP_SET_BLOCK_SIZE with arg > PAGE_SIZE ioctl_loop06.c:57: TFAIL: Set block size succeed unexpectedly
...
ioctl_loop06.c:74: TINFO: Using LOOP_CONFIGURE with block_size > PAGE_SIZE ioctl_loop06.c:57: TFAIL: Set block size succeed unexpectedly

This patch updates ioctl_loop06 to use BLK_MAX_BLOCK_SIZE instead of PAGE_SIZE for block size validation.

Quoted 8 lines

And, dynamically sets bs based on BLK_MAX_BLOCK_SIZE, using 1024 bytes if it's below 1MB or scaling it otherwise. Ensures tst_fill_file() writes efficiently while maintaining compatibility across different kernel versions.

Cc Luis Chamberlain <[email protected]>
Cc Christoph Hellwig <[email protected]>
Cc Hannes Reinecke <[email protected]>

tst_tmpdir.c:316: TINFO: Using /tmp/LTP_iocxHZLn0 as tmpdir (tmpfs filesystem)
tst_test.c:1905: TINFO: LTP version: 20250130
tst_test.c:1909: TINFO: Tested kernel: 6.15.0-rc1-1.ge4c97fd-default #1 SMP PREEMPT_DYNAMIC Mon Apr 7 21:58:09 UTC 2025 (e4c97fd) x86_64 tst_kconfig.c:88: TINFO: Parsing kernel config '/proc/config.gz'
tst_kconfig.c:676: TINFO: CONFIG_FAULT_INJECTION kernel option detected which might slow the execution tst_test.c:1722: TINFO: Overall timeout per run is 0h 02m 00s
tst_device.c:98: TINFO: Found free device 1 '/dev/loop1'
ioctl_loop06_new.c:76: TINFO: Using LOOP_SET_BLOCK_SIZE with arg < 512
ioctl_loop06_new.c:67: TPASS: Set block size failed as expected: EINVAL (22)
ioctl_loop06_new.c:76: TINFO: Using LOOP_SET_BLOCK_SIZE with arg > BLK_MAX_BLOCK_SIZE
ioctl_loop06_new.c:67: TPASS: Set block size failed as expected: EINVAL (22)
ioctl_loop06_new.c:76: TINFO: Using LOOP_SET_BLOCK_SIZE with arg != power_of_2
ioctl_loop06_new.c:67: TPASS: Set block size failed as expected: EINVAL (22)
ioctl_loop06_new.c:76: TINFO: Using LOOP_CONFIGURE with block_size < 512
tst_device.c:255: TINFO: ioctl return: 0: ECHILD (10)
tst_device.c:255: TINFO: ioctl return: 0: ECHILD (10)
tst_device.c:255: TINFO: ioctl return: -1: ENXIO (6)
ioctl_loop06_new.c:67: TPASS: Set block size failed as expected: EINVAL (22)
ioctl_loop06_new.c:76: TINFO: Using LOOP_CONFIGURE with block_size > BLK_MAX_BLOCK_SIZE
ioctl_loop06_new.c:67: TPASS: Set block size failed as expected: EINVAL (22)
ioctl_loop06_new.c:76: TINFO: Using LOOP_CONFIGURE with block_size != power_of_2
ioctl_loop06_new.c:67: TPASS: Set block size failed as expected: EINVAL (22)

Summary:

passed   6
failed   0
broken   0
skipped  0

warnings 0

Quoted 92 lines
---
 configure.ac                                  |  1 +
 include/lapi/blkdev.h                         | 19 +++++++++++++++++++
 .../kernel/syscalls/ioctl/ioctl_loop06.c      | 12 ++++++++----
 3 files changed, 28 insertions(+), 4 deletions(-)
 create mode 100644 include/lapi/blkdev.h
configure.ac +1 -0
index 0f2b6f332..5538d88d5 100644
@@ -52,6 +52,7 @@ AC_CHECK_HEADERS_ONCE([ \
     emmintrin.h \
     ifaddrs.h \
     keyutils.h \
+    linux/blkdev.h \
     linux/can.h \
     linux/cgroupstats.h \
     linux/cryptouser.h \
include/lapi/blkdev.h new file +19 -0
new file mode 100644
index 000000000..3ee058ce0
@@ -0,0 +1,19 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Copyright (c) 2025 Linux Test Project
+ *  Li Wang <[email protected]>
+ */
+
+#ifndef LAPI_BLKDEV_H__
+#define LAPI_BLKDEV_H__
+
+#ifdef HAVE_LINUX_BLKDEV_H
+#include <linux/blkdev.h>
+#endif
+
+/* Define BLK_MAX_BLOCK_SIZE for older kernels */
+#ifndef BLK_MAX_BLOCK_SIZE
+#define BLK_MAX_BLOCK_SIZE 0x00010000 /* 64K */
+#endif
+
+#endif /* LAPI_BLKDEV_H */
testcases/kernel/syscalls/ioctl/ioctl_loop06.c +8 -4
index 573871bc1..35e9e79e9 100644
@@ -14,7 +14,9 @@
 #include <unistd.h>
 #include <sys/types.h>
 #include <stdlib.h>
+#include "lapi/blkdev.h"
 #include "lapi/loop.h"
+#include "tst_fs.h"
 #include "tst_test.h"
 
 static char dev_path[1024];
@@ -31,7 +33,7 @@ static struct tcase {
 	"Using LOOP_SET_BLOCK_SIZE with arg < 512"},
 
 	{&invalid_value, LOOP_SET_BLOCK_SIZE,
-	"Using LOOP_SET_BLOCK_SIZE with arg > PAGE_SIZE"},
+	"Using LOOP_SET_BLOCK_SIZE with arg > BLK_MAX_BLOCK_SIZE"},
 
 	{&unalign_value, LOOP_SET_BLOCK_SIZE,
 	"Using LOOP_SET_BLOCK_SIZE with arg != power_of_2"},
@@ -40,7 +42,7 @@ static struct tcase {
 	"Using LOOP_CONFIGURE with block_size < 512"},
 
 	{&invalid_value, LOOP_CONFIGURE,
-	"Using LOOP_CONFIGURE with block_size > PAGE_SIZE"},
+	"Using LOOP_CONFIGURE with block_size > BLK_MAX_BLOCK_SIZE"},
 
 	{&unalign_value, LOOP_CONFIGURE,
 	"Using LOOP_CONFIGURE with block_size != power_of_2"},
@@ -103,10 +105,12 @@ static void setup(void)
 	if (dev_num < 0)
 		tst_brk(TBROK, "Failed to find free loop device");
 
-	tst_fill_file("test.img", 0, 1024, 1024);
+	size_t bs = (BLK_MAX_BLOCK_SIZE < TST_MB) ? 1024 : 4 * BLK_MAX_BLOCK_SIZE / 1024;
+	tst_fill_file("test.img", 0, bs, 1024);
+
 	half_value = 256;
 	pg_size = getpagesize();
-	invalid_value = pg_size * 2 ;
+	invalid_value = BLK_MAX_BLOCK_SIZE * 2;
 	unalign_value = pg_size - 1;
 
 	dev_fd = SAFE_OPEN(dev_path, O_RDWR);
Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp
#7 lore raw

Hi Avinesh, All,

Quoted 24 lines · Avinesh Kumar
Quoted 1 line · CEST Li Wang

The kernel commit 47dd6753 ("block/bdev: lift block size restrictions to

64k")

Quoted 2 lines

now supports block sizes larger than PAGE_SIZE, with a new upper limit of BLK_MAX_BLOCK_SIZE (64K). But ioctl_loop06 still assumes that PAGE_SIZE

is the

Quoted 3 lines

maximum allowed block size, causing failures on newer kernels(>= 6.14):

ioctl_loop06.c:74: TINFO: Using LOOP_SET_BLOCK_SIZE with arg >

PAGE_SIZE

Quoted 3 lines

ioctl_loop06.c:57: TFAIL: Set block size succeed unexpectedly
...
ioctl_loop06.c:74: TINFO: Using LOOP_CONFIGURE with block_size >

PAGE_SIZE

Quoted 3 lines

ioctl_loop06.c:57: TFAIL: Set block size succeed unexpectedly

This patch updates ioctl_loop06 to use BLK_MAX_BLOCK_SIZE instead of

PAGE_SIZE

Quoted 3 lines

for block size validation.

And, dynamically sets bs based on BLK_MAX_BLOCK_SIZE, using 1024 bytes

if it's

Quoted 1 line

below 1MB or scaling it otherwise. Ensures tst_fill_file() writes

efficiently

Quoted 9 lines
Quoted 6 lines

while maintaining compatibility across different kernel versions.

Cc Luis Chamberlain <[email protected]>
Cc Christoph Hellwig <[email protected]>
Cc Hannes Reinecke <[email protected]>

Thanks for testing. I thought this one was merged, but when I checked the log, it was not.
Anyway, I pushed it now.

Signature · 5 lines
Regards,
Li Wang

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
#8 lore raw
In reply to Li Wang
Quoted 9 lines · Li Wang

Hi Luis,

Quoted 5 lines · Luis Chamberlain
Quoted 1 line · Li Wang

The kernel commit 47dd6753 ("block/bdev: lift block size restrictions to

64k")

Quoted 2 lines

now supports block sizes larger than PAGE_SIZE, with a new upper limit of BLK_MAX_BLOCK_SIZE (64K). But ioctl_loop06 still assumes that PAGE_SIZE

Quoted 8 lines
Quoted 3 lines

is the

Quoted 1 line

maximum allowed block size, causing failures on newer kernels(>= 6.14):

Well, this sounds like we need to go back to patch v1:
https://lists.linux.it/pipermail/ltp/2025-March/042599.html

Note: LTP has a function tst_kvercmp() for comparing kernel versions.

Close.

BLK_MAX_BLOCK_SIZE may change in the future so the only real way to test for an invalid block size is having a check for linux/blkdev.h.

I see commit 7c84fa710f75 ("ioprio: use ioprio.h kernel header if it exists") is an example of using a kernel header file. I think similar thing can be done to use linux/blkdev.h and if you don't have HAVE_LINUX_BLKDEV_H then you can define BLK_MAX_BLOCK_SIZE to 64k for for older kernels.

A more appropriate invalid value would be BLK_MAX_BLOCK_SIZE * 2 as non power of 2 values will be outright disallowed.

Luis

Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp
#9 lore raw
Quoted 24 lines · Luis Chamberlain
Quoted 9 lines · Li Wang

Hi Luis,

Quoted 5 lines · Luis Chamberlain
Quoted 1 line · Li Wang

The kernel commit 47dd6753 ("block/bdev: lift block size restrictions to

64k")

Quoted 2 lines

now supports block sizes larger than PAGE_SIZE, with a new upper limit of BLK_MAX_BLOCK_SIZE (64K). But ioctl_loop06 still assumes that PAGE_SIZE

Quoted 8 lines
Quoted 3 lines

is the

Quoted 1 line

maximum allowed block size, causing failures on newer kernels(>= 6.14):

Well, this sounds like we need to go back to patch v1:
https://lists.linux.it/pipermail/ltp/2025-March/042599.html

Note: LTP has a function tst_kvercmp() for comparing kernel versions.

Close.

BLK_MAX_BLOCK_SIZE may change in the future so the only real way to test for an invalid block size is having a check for linux/blkdev.h.

Quoted 8 lines

I see commit 7c84fa710f75 ("ioprio: use ioprio.h kernel header if it exists") is an example of using a kernel header file. I think similar thing can be done to use linux/blkdev.h and if you don't have HAVE_LINUX_BLKDEV_H then you can define BLK_MAX_BLOCK_SIZE to 64k for for older kernels.

A more appropriate invalid value would be BLK_MAX_BLOCK_SIZE * 2 as non power of 2 values will be outright disallowed.

While at it, be sure to scale tst_fill_file("test.img", 0, 1024, 1024) (I think that is bs=1024, count=1024, so 1 MiB file size) so that you take into consideration the BLK_MAX_BLOCK_SIZE so that the file size will be at least maybe 10 * BLK_MAX_BLOCK_SIZE or something sensible.

Because as soon as we can support a BLK_MAX_BLOCK_SIZE of 2 MiB this test will also fail again.

Luis

Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp
author
#10 lore raw
Quoted 39 lines · Luis Chamberlain
Quoted 4 lines · Luis Chamberlain
Quoted 3 lines · Li Wang

Hi Luis,

On Fri, Mar 14, 2025 at 6:59 AM Luis Chamberlain <[email protected]>

Quoted 3 lines · wrote:
Quoted 3 lines
Quoted 3 lines · Li Wang
Quoted 1 line

The kernel commit 47dd6753 ("block/bdev: lift block size

restrictions to

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

64k")

Quoted 1 line

now supports block sizes larger than PAGE_SIZE, with a new upper

limit of

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

BLK_MAX_BLOCK_SIZE (64K). But ioctl_loop06 still assumes that

PAGE_SIZE

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

is the

Quoted 1 line

maximum allowed block size, causing failures on newer kernels(>=

6.14):

Quoted 12 lines
Quoted 6 lines
Quoted 1 line

Well, this sounds like we need to go back to patch v1:
https://lists.linux.it/pipermail/ltp/2025-March/042599.html

Note: LTP has a function tst_kvercmp() for comparing kernel versions.

Close.

BLK_MAX_BLOCK_SIZE may change in the future so the only real way to test for an invalid block size is having a check for linux/blkdev.h.

Quoted 8 lines

I see commit 7c84fa710f75 ("ioprio: use ioprio.h kernel header if it exists") is an example of using a kernel header file. I think similar thing can be done to use linux/blkdev.h and if you don't have HAVE_LINUX_BLKDEV_H then you can define BLK_MAX_BLOCK_SIZE to 64k for for older kernels.

A more appropriate invalid value would be BLK_MAX_BLOCK_SIZE * 2 as non power of 2 values will be outright disallowed.

Quoted 8 lines

While at it, be sure to scale tst_fill_file("test.img", 0, 1024, 1024) (I think that is bs=1024, count=1024, so 1 MiB file size) so that you take into consideration the BLK_MAX_BLOCK_SIZE so that the file size will be at least maybe 10 * BLK_MAX_BLOCK_SIZE or something sensible.

Because as soon as we can support a BLK_MAX_BLOCK_SIZE of 2 MiB this test will also fail again.

Make sense!! I will improve this in the next version, too.

Signature · 5 lines
Regards,
Li Wang

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
author
#11 lore raw
In reply to Luis Chamberlain via ltp
Quoted 22 lines · Luis Chamberlain
Quoted 3 lines · Li Wang

Hi Luis,

On Fri, Mar 14, 2025 at 6:59 AM Luis Chamberlain <[email protected]>

Quoted 3 lines · wrote:
Quoted 3 lines · Li Wang
Quoted 1 line

The kernel commit 47dd6753 ("block/bdev: lift block size

restrictions to

Quoted 2 lines
Quoted 2 lines

64k")

Quoted 1 line

now supports block sizes larger than PAGE_SIZE, with a new upper

limit of

Quoted 1 line
Quoted 1 line
Quoted 1 line

BLK_MAX_BLOCK_SIZE (64K). But ioctl_loop06 still assumes that

PAGE_SIZE

Quoted 2 lines
Quoted 2 lines

is the

Quoted 1 line

maximum allowed block size, causing failures on newer kernels(>=

6.14):

Quoted 5 lines
Quoted 1 line

Well, this sounds like we need to go back to patch v1:
https://lists.linux.it/pipermail/ltp/2025-March/042599.html

Quoted 7 lines
Quoted 1 line

Note: LTP has a function tst_kvercmp() for comparing kernel versions.

Close.

BLK_MAX_BLOCK_SIZE may change in the future so the only real way to test for an invalid block size is having a check for linux/blkdev.h.

Good point.

Quoted 6 lines

I see commit 7c84fa710f75 ("ioprio: use ioprio.h kernel header if it exists") is an example of using a kernel header file. I think similar thing can be done to use linux/blkdev.h and if you don't have HAVE_LINUX_BLKDEV_H then you can define BLK_MAX_BLOCK_SIZE to 64k for for older kernels.

To precise, if BLK_MAX_BLOCK_SIZE is undefined (in blkdev.h) in older kernel (e.g. kernel-6.12) then we define it to 64k.

Quoted 4 lines

A more appropriate invalid value would be BLK_MAX_BLOCK_SIZE * 2 as non power of 2 values will be outright disallowed.

Agreed. Thanks for reviewing.

Signature · 5 lines
Regards,
Li Wang

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp