Skip to contentSkip to conversation

Build failures due to commit 416161db (btrfs: offline dedupe)

DISCUSSION Guenter Roeck 4 participants 10 messages Sep 13, 2013 – Sep 18, 2013 Btrfs file system

lore
Share & tools
lore
#0 lore raw

fs/btrfs/ioctl.c: In function 'btrfs_ioctl_file_extent_same':
fs/btrfs/ioctl.c:2802:3: error: implicit declaration of function '__put_user_unaligned' [-Werror=implicit-function-declaration]
cc1: some warnings being treated as errors
make[2]: *** [fs/btrfs/ioctl.o] Error 1
make[2]: *** Waiting for unfinished jobs....

Seen with alpha:allmodconfig, arm:allmodconfig, m68k:allmodconfig, and xtensa:allmodconfig.

Guenter

#1 lore raw
Quoted 8 lines · Guenter Roeck

fs/btrfs/ioctl.c: In function 'btrfs_ioctl_file_extent_same':
fs/btrfs/ioctl.c:2802:3: error: implicit declaration of function '__put_user_unaligned' [-Werror=implicit-function-declaration]
cc1: some warnings being treated as errors
make[2]: *** [fs/btrfs/ioctl.o] Error 1
make[2]: *** Waiting for unfinished jobs....

Seen with alpha:allmodconfig, arm:allmodconfig, m68k:allmodconfig, and xtensa:allmodconfig.

Known issue, cfr. my early warning 10 days ago:

"Btrfs is the first user of __put_user_unaligned() outside the compat code, hence now all 32-bit architectures should make sure to implement this, too."

http://marc.info/?l=linux-arch&m=137820065929216&w=2

and today's thread https://lkml.org/lkml/2013/9/12/814

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]

In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds

#2 lore raw
Quoted 10 lines · Geert Uytterhoeven
Quoted 8 lines · Guenter Roeck

fs/btrfs/ioctl.c: In function 'btrfs_ioctl_file_extent_same':
fs/btrfs/ioctl.c:2802:3: error: implicit declaration of function '__put_user_unaligned' [-Werror=implicit-function-declaration]
cc1: some warnings being treated as errors
make[2]: *** [fs/btrfs/ioctl.o] Error 1
make[2]: *** Waiting for unfinished jobs....

Seen with alpha:allmodconfig, arm:allmodconfig, m68k:allmodconfig, and xtensa:allmodconfig.

Quoted 9 lines

Known issue, cfr. my early warning 10 days ago:

"Btrfs is the first user of __put_user_unaligned() outside the compat code, hence now all 32-bit architectures should make sure to implement this, too."

http://marc.info/?l=linux-arch&m=137820065929216&w=2

and today's thread https://lkml.org/lkml/2013/9/12/814

It doesn't seem right that a patch breaks the build for several platforms, and the problem is then blamed on the platform code instead of the code that is introducing the problem.

Maybe we should add BROKEN to the btrfs dependencies for the affected platforms. After all, it _is_ broken.

Guenter

#3 lore raw

Quoting Guenter Roeck (2013-09-13 12:35:35)

Quoted 21 lines · Geert Uytterhoeven
Quoted 10 lines · Guenter Roeck
Quoted 8 lines

fs/btrfs/ioctl.c: In function 'btrfs_ioctl_file_extent_same':
fs/btrfs/ioctl.c:2802:3: error: implicit declaration of function '__put_user_unaligned' [-Werror=implicit-function-declaration]
cc1: some warnings being treated as errors
make[2]: *** [fs/btrfs/ioctl.o] Error 1
make[2]: *** Waiting for unfinished jobs....

Seen with alpha:allmodconfig, arm:allmodconfig, m68k:allmodconfig, and xtensa:allmodconfig.

Quoted 9 lines

Known issue, cfr. my early warning 10 days ago:

"Btrfs is the first user of __put_user_unaligned() outside the compat code, hence now all 32-bit architectures should make sure to implement this, too."

http://marc.info/?l=linux-arch&m=137820065929216&w=2

and today's thread https://lkml.org/lkml/2013/9/12/814

Quoted 6 lines

It doesn't seem right that a patch breaks the build for several platforms, and the problem is then blamed on the platform code instead of the code that is introducing the problem.

Maybe we should add BROKEN to the btrfs dependencies for the affected platforms. After all, it _is_ broken.

I'm happy to fix this with a bigger put of the info struct, just let me know the preferred arch-happy solution.

-chris

#4 lore raw
Quoted 26 lines · Chris Mason

Quoting Guenter Roeck (2013-09-13 12:35:35)

Quoted 25 lines · Geert Uytterhoeven
Quoted 10 lines · Guenter Roeck
Quoted 8 lines

fs/btrfs/ioctl.c: In function 'btrfs_ioctl_file_extent_same':
fs/btrfs/ioctl.c:2802:3: error: implicit declaration of function '__put_user_unaligned' [-Werror=implicit-function-declaration]
cc1: some warnings being treated as errors
make[2]: *** [fs/btrfs/ioctl.o] Error 1
make[2]: *** Waiting for unfinished jobs....

Seen with alpha:allmodconfig, arm:allmodconfig, m68k:allmodconfig, and xtensa:allmodconfig.

Quoted 9 lines

Known issue, cfr. my early warning 10 days ago:

"Btrfs is the first user of __put_user_unaligned() outside the compat code, hence now all 32-bit architectures should make sure to implement this, too."

http://marc.info/?l=linux-arch&m=137820065929216&w=2

and today's thread https://lkml.org/lkml/2013/9/12/814

It doesn't seem right that a patch breaks the build for several platforms, and the problem is then blamed on the platform code instead of the code that is introducing the problem.

Quoted 6 lines
Quoted 2 lines

Maybe we should add BROKEN to the btrfs dependencies for the affected platforms. After all, it _is_ broken.

I'm happy to fix this with a bigger put of the info struct, just let me know the preferred arch-happy solution.

Me not either. The only requirement I would have is that it should not break a build. Of course, it would be even better if it would actually work ;-).

Guenter

#5 lore raw
In reply to Chris Mason
Quoted 26 lines · Chris Mason

Quoting Guenter Roeck (2013-09-13 12:35:35)

Quoted 25 lines · Geert Uytterhoeven
Quoted 10 lines · Guenter Roeck
Quoted 8 lines

fs/btrfs/ioctl.c: In function 'btrfs_ioctl_file_extent_same':
fs/btrfs/ioctl.c:2802:3: error: implicit declaration of function '__put_user_unaligned' [-Werror=implicit-function-declaration]
cc1: some warnings being treated as errors
make[2]: *** [fs/btrfs/ioctl.o] Error 1
make[2]: *** Waiting for unfinished jobs....

Seen with alpha:allmodconfig, arm:allmodconfig, m68k:allmodconfig, and xtensa:allmodconfig.

Quoted 9 lines

Known issue, cfr. my early warning 10 days ago:

"Btrfs is the first user of __put_user_unaligned() outside the compat code, hence now all 32-bit architectures should make sure to implement this, too."

http://marc.info/?l=linux-arch&m=137820065929216&w=2

and today's thread https://lkml.org/lkml/2013/9/12/814

It doesn't seem right that a patch breaks the build for several platforms, and the problem is then blamed on the platform code instead of the code that is introducing the problem.

Quoted 5 lines
Quoted 2 lines

Maybe we should add BROKEN to the btrfs dependencies for the affected platforms. After all, it _is_ broken.

I'm happy to fix this with a bigger put of the info struct, just let me know the preferred arch-happy solution.

In fact old versions of the patch were putting the whole struct but during review I was asked to change it. This should be very straight forward to fix so long as we all stay calm ;)

	--Mark

--
Mark Fasheh

#6 lore raw

Quoting Mark Fasheh (2013-09-13 13:58:01)

Quoted 9 lines · Chris Mason
Quoted 3 lines

Quoting Guenter Roeck (2013-09-13 12:35:35)
I'm happy to fix this with a bigger put of the info struct, just let me know the preferred arch-happy solution.

In fact old versions of the patch were putting the whole struct but during review I was asked to change it. This should be very straight forward to fix so long as we all stay calm ;)

        --Mark

Mark, could you please send a patch for the whole-struct option until the unaligned put is upstreamed?

-chris

#7 lore raw
Quoted 4 lines · Chris Mason

Mark, could you please send a patch for the whole-struct option until the unaligned put is upstreamed?

-chris

Here you go. It's been lightly tested and needs review.

Thanks,

	--Mark

--
Mark Fasheh

From: Mark Fasheh <[email protected]>

[PATCH] btrfs: change extent-same to copy entire argument struct

btrfs_ioctl_file_extent_same() uses __put_user_unaligned() to copy some data back to it's argument struct. Unfortunately, not all architectures provide __put_user_unaligned(), so compiles break on them if btrfs is selected.

Instead, just copy the whole struct in / out at the start and end of operations, respectively.

---
 fs/btrfs/ioctl.c | 76 +++++++++++++++++++++++++++++++++-----------------------
 1 file changed, 45 insertions(+), 31 deletions(-)
fs/btrfs/ioctl.c +45 -31
index 1a5b946..25d6920 100644
@@ -2696,9 +2696,9 @@ out_unlock:
 static long btrfs_ioctl_file_extent_same(struct file *file,
 					 void __user *argp)
 {
-	struct btrfs_ioctl_same_args *args = argp;
-	struct btrfs_ioctl_same_args same;
-	struct btrfs_ioctl_same_extent_info info;
+	struct btrfs_ioctl_same_args tmp;
+	struct btrfs_ioctl_same_args *same;
+	struct btrfs_ioctl_same_extent_info *info;
 	struct inode *src = file->f_dentry->d_inode;
 	struct file *dst_file = NULL;
 	struct inode *dst;
@@ -2706,6 +2706,7 @@ static long btrfs_ioctl_file_extent_same(struct file *file,
 	u64 len;
 	int i;
 	int ret;
+	unsigned long size;
 	u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
 	bool is_admin = capable(CAP_SYS_ADMIN);
 
@@ -2716,15 +2717,30 @@ static long btrfs_ioctl_file_extent_same(struct file *file,
 	if (ret)
 		return ret;
 
-	if (copy_from_user(&same,
+	if (copy_from_user(&tmp,
 			   (struct btrfs_ioctl_same_args __user *)argp,
-			   sizeof(same))) {
+			   sizeof(tmp))) {
 		ret = -EFAULT;
 		goto out;
 	}
 
-	off = same.logical_offset;
-	len = same.length;
+	size = sizeof(tmp) +
+		tmp.dest_count * sizeof(struct btrfs_ioctl_same_extent_info);
+
+	same = kmalloc(size, GFP_NOFS);
+	if (!same) {
+		ret = -EFAULT;
+		goto out;
+	}
+
+	if (copy_from_user(same,
+			   (struct btrfs_ioctl_same_args __user *)argp, size)) {
+		ret = -EFAULT;
+		goto out;
+	}
+
+	off = same->logical_offset;
+	len = same->length;
 
 	/*
 	 * Limit the total length we will dedupe for each operation.
@@ -2752,27 +2768,28 @@ static long btrfs_ioctl_file_extent_same(struct file *file,
 	if (!S_ISREG(src->i_mode))
 		goto out;
 
-	ret = 0;
-	for (i = 0; i < same.dest_count; i++) {
-		if (copy_from_user(&info, &args->info[i], sizeof(info))) {
-			ret = -EFAULT;
-			goto out;
-		}
+	/* pre-format output fields to sane values */
+	for (i = 0; i < same->dest_count; i++) {
+		same->info[i].bytes_deduped = 0ULL;
+		same->info[i].status = 0;
+	}
 
-		info.bytes_deduped = 0;
+	ret = 0;
+	for (i = 0; i < same->dest_count; i++) {
+		info = &same->info[i];
 
-		dst_file = fget(info.fd);
+		dst_file = fget(info->fd);
 		if (!dst_file) {
-			info.status = -EBADF;
+			info->status = -EBADF;
 			goto next;
 		}
 
 		if (!(is_admin || (dst_file->f_mode & FMODE_WRITE))) {
-			info.status = -EINVAL;
+			info->status = -EINVAL;
 			goto next;
 		}
 
-		info.status = -EXDEV;
+		info->status = -EXDEV;
 		if (file->f_path.mnt != dst_file->f_path.mnt)
 			goto next;
 
@@ -2781,32 +2798,29 @@ static long btrfs_ioctl_file_extent_same(struct file *file,
 			goto next;
 
 		if (S_ISDIR(dst->i_mode)) {
-			info.status = -EISDIR;
+			info->status = -EISDIR;
 			goto next;
 		}
 
 		if (!S_ISREG(dst->i_mode)) {
-			info.status = -EACCES;
+			info->status = -EACCES;
 			goto next;
 		}
 
-		info.status = btrfs_extent_same(src, off, len, dst,
-						info.logical_offset);
-		if (info.status == 0)
-			info.bytes_deduped += len;
+		info->status = btrfs_extent_same(src, off, len, dst,
+						info->logical_offset);
+		if (info->status == 0)
+			info->bytes_deduped += len;
 
 next:
 		if (dst_file)
 			fput(dst_file);
-
-		if (__put_user_unaligned(info.status, &args->info[i].status) ||
-		    __put_user_unaligned(info.bytes_deduped,
-					 &args->info[i].bytes_deduped)) {
-			ret = -EFAULT;
-			goto out;
-		}                                                               
 	}
 
+	ret = copy_to_user(argp, same, size);
+	if (ret)
+		ret = -EFAULT;
+
 out:
 	mnt_drop_write_file(file);
 	return ret;
Signature · 1 line
1.8.1.4
#8 lore raw
Quoted 8 lines · Mark Fasheh
Quoted 4 lines · Chris Mason

Mark, could you please send a patch for the whole-struct option until the unaligned put is upstreamed?

-chris

Here you go. It's been lightly tested and needs review.

At the very least it does fix the build error on the affected platforms.

Guenter

Quoted 169 lines

Thanks,

	--Mark

--
Mark Fasheh

From: Mark Fasheh <[email protected]>

[PATCH] btrfs: change extent-same to copy entire argument struct

btrfs_ioctl_file_extent_same() uses __put_user_unaligned() to copy some data back to it's argument struct. Unfortunately, not all architectures provide __put_user_unaligned(), so compiles break on them if btrfs is selected.

Instead, just copy the whole struct in / out at the start and end of operations, respectively.

---
 fs/btrfs/ioctl.c | 76 +++++++++++++++++++++++++++++++++-----------------------
 1 file changed, 45 insertions(+), 31 deletions(-)
fs/btrfs/ioctl.c +45 -31
index 1a5b946..25d6920 100644
@@ -2696,9 +2696,9 @@ out_unlock:
 static long btrfs_ioctl_file_extent_same(struct file *file,
 					 void __user *argp)
 {
-	struct btrfs_ioctl_same_args *args = argp;
-	struct btrfs_ioctl_same_args same;
-	struct btrfs_ioctl_same_extent_info info;
+	struct btrfs_ioctl_same_args tmp;
+	struct btrfs_ioctl_same_args *same;
+	struct btrfs_ioctl_same_extent_info *info;
 	struct inode *src = file->f_dentry->d_inode;
 	struct file *dst_file = NULL;
 	struct inode *dst;
@@ -2706,6 +2706,7 @@ static long btrfs_ioctl_file_extent_same(struct file *file,
 	u64 len;
 	int i;
 	int ret;
+	unsigned long size;
 	u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
 	bool is_admin = capable(CAP_SYS_ADMIN);
 
@@ -2716,15 +2717,30 @@ static long btrfs_ioctl_file_extent_same(struct file *file,
 	if (ret)
 		return ret;
 
-	if (copy_from_user(&same,
+	if (copy_from_user(&tmp,
 			   (struct btrfs_ioctl_same_args __user *)argp,
-			   sizeof(same))) {
+			   sizeof(tmp))) {
 		ret = -EFAULT;
 		goto out;
 	}
 
-	off = same.logical_offset;
-	len = same.length;
+	size = sizeof(tmp) +
+		tmp.dest_count * sizeof(struct btrfs_ioctl_same_extent_info);
+
+	same = kmalloc(size, GFP_NOFS);
+	if (!same) {
+		ret = -EFAULT;
+		goto out;
+	}
+
+	if (copy_from_user(same,
+			   (struct btrfs_ioctl_same_args __user *)argp, size)) {
+		ret = -EFAULT;
+		goto out;
+	}
+
+	off = same->logical_offset;
+	len = same->length;
 
 	/*
 	 * Limit the total length we will dedupe for each operation.
@@ -2752,27 +2768,28 @@ static long btrfs_ioctl_file_extent_same(struct file *file,
 	if (!S_ISREG(src->i_mode))
 		goto out;
 
-	ret = 0;
-	for (i = 0; i < same.dest_count; i++) {
-		if (copy_from_user(&info, &args->info[i], sizeof(info))) {
-			ret = -EFAULT;
-			goto out;
-		}
+	/* pre-format output fields to sane values */
+	for (i = 0; i < same->dest_count; i++) {
+		same->info[i].bytes_deduped = 0ULL;
+		same->info[i].status = 0;
+	}
 
-		info.bytes_deduped = 0;
+	ret = 0;
+	for (i = 0; i < same->dest_count; i++) {
+		info = &same->info[i];
 
-		dst_file = fget(info.fd);
+		dst_file = fget(info->fd);
 		if (!dst_file) {
-			info.status = -EBADF;
+			info->status = -EBADF;
 			goto next;
 		}
 
 		if (!(is_admin || (dst_file->f_mode & FMODE_WRITE))) {
-			info.status = -EINVAL;
+			info->status = -EINVAL;
 			goto next;
 		}
 
-		info.status = -EXDEV;
+		info->status = -EXDEV;
 		if (file->f_path.mnt != dst_file->f_path.mnt)
 			goto next;
 
@@ -2781,32 +2798,29 @@ static long btrfs_ioctl_file_extent_same(struct file *file,
 			goto next;
 
 		if (S_ISDIR(dst->i_mode)) {
-			info.status = -EISDIR;
+			info->status = -EISDIR;
 			goto next;
 		}
 
 		if (!S_ISREG(dst->i_mode)) {
-			info.status = -EACCES;
+			info->status = -EACCES;
 			goto next;
 		}
 
-		info.status = btrfs_extent_same(src, off, len, dst,
-						info.logical_offset);
-		if (info.status == 0)
-			info.bytes_deduped += len;
+		info->status = btrfs_extent_same(src, off, len, dst,
+						info->logical_offset);
+		if (info->status == 0)
+			info->bytes_deduped += len;
 
 next:
 		if (dst_file)
 			fput(dst_file);
-
-		if (__put_user_unaligned(info.status, &args->info[i].status) ||
-		    __put_user_unaligned(info.bytes_deduped,
-					 &args->info[i].bytes_deduped)) {
-			ret = -EFAULT;
-			goto out;
-		}                                                               
 	}
 
+	ret = copy_to_user(argp, same, size);
+	if (ret)
+		ret = -EFAULT;
+
 out:
 	mnt_drop_write_file(file);
 	return ret;
Signature · 1 line
1.8.1.4
#9 lore raw
Quoted 10 lines · Guenter Roeck
Quoted 8 lines · Mark Fasheh
Quoted 4 lines · Chris Mason

Mark, could you please send a patch for the whole-struct option until the unaligned put is upstreamed?

-chris

Here you go. It's been lightly tested and needs review.

At the very least it does fix the build error on the affected platforms.

Thanks for verifying that Guenter.

	--Mark

--
Mark Fasheh