Skip to contentSkip to conversation

btrfs: avoid opencoded 64-bit div/mod operation

PATCH In review Arnd Bergmann 3 participants 3 messages Dec 16, 2024 Btrfs file system

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 btrfs: avoid opencoded 64-bit div/mod operation

From: Arnd Bergmann <[email protected]>

Dividing 64-bit numbers causes a link failure on 32-bit builds:

arm-linux-gnueabi-ld: fs/btrfs/sysfs.o: in function `btrfs_read_policy_store':
sysfs.c:(.text+0x3ce0): undefined reference to `__aeabi_ldivmod'

Use an explicit call to div_u64_rem() here to work around this. It would be possible to optimize this further, but this is not a performance critical operation.

Fixes 185fa5c7ac5a ("btrfs: introduce RAID1 round-robin read balancing")
---
 fs/btrfs/sysfs.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
fs/btrfs/sysfs.c +3 -1
index 50bc4b6cb821..67bc8fa4d6ab 100644
@@ -1433,7 +1433,9 @@ static ssize_t btrfs_read_policy_store(struct kobject *kobj,
 #ifdef CONFIG_BTRFS_EXPERIMENTAL
 	if (index == BTRFS_READ_POLICY_RR) {
 		if (value != -1) {
-			if ((value % fs_devices->fs_info->sectorsize) != 0) {
+			u32 rem;
+			div_u64_rem(value, fs_devices->fs_info->sectorsize, &rem);
+			if (rem) {
 				btrfs_err(fs_devices->fs_info,
 "read_policy: min_contiguous_read %lld should be multiples of the sectorsize %u",
 					  value, fs_devices->fs_info->sectorsize);
Signature · 1 line
2.39.5
#1 lore raw
Quoted 22 lines · Arnd Bergmann

From: Arnd Bergmann <[email protected]>

Dividing 64-bit numbers causes a link failure on 32-bit builds:

arm-linux-gnueabi-ld: fs/btrfs/sysfs.o: in function `btrfs_read_policy_store':
sysfs.c:(.text+0x3ce0): undefined reference to `__aeabi_ldivmod'

Use an explicit call to div_u64_rem() here to work around this. It would be possible to optimize this further, but this is not a performance critical operation.

Fixes 185fa5c7ac5a ("btrfs: introduce RAID1 round-robin read balancing")
---
  fs/btrfs/sysfs.c | 4 +++-
  1 file changed, 3 insertions(+), 1 deletion(-)
fs/btrfs/sysfs.c +0 -0
index 50bc4b6cb821..67bc8fa4d6ab 100644

fs/btrfs/sysfs.c static ssize_t btrfs_read_policy_store(struct kobject *kobj, · line 1437

Quoted 6 lines
fs/btrfs/sysfs.c +2 -1
@@ -1433,7 +1433,9 @@ static ssize_t btrfs_read_policy_store(struct kobject *kobj,
  #ifdef CONFIG_BTRFS_EXPERIMENTAL
  	if (index == BTRFS_READ_POLICY_RR) {
  		if (value != -1) {
-			if ((value % fs_devices->fs_info->sectorsize) != 0) {
+			u32 rem;
+			div_u64_rem(value, fs_devices->fs_info->sectorsize, &rem);

The original check is already bad, it's just a IS_ALIGNED().

So a much simpler solution is:

+ if (!IS_ALIGNED(value, fs_info->sectorsize)) {

Thanks,
Qu

Quoted 4 lines
fs/btrfs/sysfs.c +1 -0
+			if (rem) {
  				btrfs_err(fs_devices->fs_info,
  "read_policy: min_contiguous_read %lld should be multiples of the sectorsize %u",
  					  value, fs_devices->fs_info->sectorsize);
#2 lore raw
Quoted 14 lines

fs/btrfs/sysfs.c static ssize_t btrfs_read_policy_store(struct · line 1438

Quoted 11 lines
fs/btrfs/sysfs.c +2 -1
@@ -1433,7 +1433,9 @@ static ssize_t btrfs_read_policy_store(struct 
 kobject *kobj,
   #ifdef CONFIG_BTRFS_EXPERIMENTAL
       if (index == BTRFS_READ_POLICY_RR) {
           if (value != -1) {
-            if ((value % fs_devices->fs_info->sectorsize) != 0) {
+            u32 rem;
+            div_u64_rem(value, fs_devices->fs_info->sectorsize, &rem);

The original check is already bad, it's just a IS_ALIGNED().

Quoted 3 lines

So a much simpler solution is:

+            if (!IS_ALIGNED(value, fs_info->sectorsize)) {

Right, this was also in the review comment by David, and the change is already in the local work-space. I will be sending v4 out. Currently dealing with some rebase issues.

Thanks, Anand