Skip to contentSkip to conversation

Properly cast to avoid compiler warnings, fixes FTBFS on alpha and ia64.

PATCH In review Dimitri John Ledkov 3 participants 9 messages Sep 21, 2014 – Sep 22, 2014

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/4 Properly cast to avoid compiler warnings, fixes FTBFS on alpha and ia64.

Bug-Debian: http://bugs.debian.org/539433
Bug-Debian: http://bugs.debian.org/583768
Authors:
Luca Bruno <[email protected]>
Alexander Kurtz <[email protected]> Daniel Baumann <[email protected]>

---
 btrfs-convert.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
btrfs-convert.c +1 -1
index 71b7bd6..3673050 100644
@@ -2441,7 +2441,7 @@ static int do_rollback(const char *devname)
 	ext2_root = btrfs_read_fs_root(root->fs_info, &key);
 	if (!ext2_root || IS_ERR(ext2_root)) {
 		fprintf(stderr, "unable to open subvol %llu\n",
-			key.objectid);
+			(unsigned long long) key.objectid);
 		goto fail;
 	}
 
Signature · 1 line
2.1.0.rc1
#1 lore raw

PATCH 2/4 Fixes FTBFS with --no-add-needed.

From: Luk Claes <[email protected]>

Bug-Debian: http://bugs.debian.org/554059

---
 Makefile | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
Makefile +1 -1
index e721e99..441e925 100644
@@ -26,7 +26,7 @@ TESTS = fsck-tests.sh convert-tests.sh
 INSTALL = install
 prefix ?= /usr/local
 bindir = $(prefix)/bin
-lib_LIBS = -luuid -lblkid -lm -lz -llzo2 -L.
+lib_LIBS = -luuid -lblkid -lm -lz -llzo2 -lcom_err -L.
 libdir ?= $(prefix)/lib
 incdir = $(prefix)/include/btrfs
 LIBS = $(lib_LIBS) $(libs_static)
Signature · 1 line
2.1.0.rc1
#2 lore raw

Re: [PATCH 2/4] Fixes FTBFS with --no-add-needed.

Makefile TESTS = fsck-tests.sh convert-tests.sh · line 29

Quoted 8 lines · Dimitri John Ledkov
Makefile +1 -1
@@ -26,7 +26,7 @@ TESTS = fsck-tests.sh convert-tests.sh
 INSTALL = install
 prefix ?= /usr/local
 bindir = $(prefix)/bin
-lib_LIBS = -luuid -lblkid -lm -lz -llzo2 -L.
+lib_LIBS = -luuid -lblkid -lm -lz -llzo2 -lcom_err -L.

This will add the com_err library to all binaries, while it's actually used only by the btrfs-convert utility and this dependency has been properly added in ef85e7e285daf (released in 3.12).

One can add a per-target libraries like

btrfs_convert_libs = -lext2fs -lcom_err

Also please be more descriptive what the actual error is in the changelog itself so I don't have to look it up in the referenced bug.

#3 lore raw

PATCH 3/4 Fixing unaligned memory accesses.

In reply to Dimitri John Ledkov

From: Shawn Landen <[email protected]>

Bug-Debian: http://bugs.debian.org/656955

---
 ctree.h   | 18 ++++++++++++++----
 volumes.c |  5 +++--
 2 files changed, 17 insertions(+), 6 deletions(-)
ctree.h +14 -4
index fa73c4a..92c6ad3 100644
@@ -19,6 +19,8 @@
 #ifndef __BTRFS__
 #define __BTRFS__
 
+#include <stdint.h>
+
 #if BTRFS_FLAT_INCLUDES
 #include "list.h"
 #include "kerncompat.h"
@@ -1191,13 +1193,17 @@ struct btrfs_root {
 static inline u##bits btrfs_##name(const struct extent_buffer *eb)	\
 {									\
 	const struct btrfs_header *h = (struct btrfs_header *)eb->data;	\
-	return le##bits##_to_cpu(h->member);				\
+	uint##bits##_t t;						\
+	memcpy(&t, &h->member, sizeof(h->member));			\
+	return le##bits##_to_cpu(t);					\
 }									\
 static inline void btrfs_set_##name(struct extent_buffer *eb,		\
 				    u##bits val)			\
 {									\
 	struct btrfs_header *h = (struct btrfs_header *)eb->data;	\
-	h->member = cpu_to_le##bits(val);				\
+	uint##bits##_t t;						\
+	t = cpu_to_le##bits(val);					\
+	memcpy(&h->member, &t, sizeof(h->member));			\
 }
 
 #define BTRFS_SETGET_FUNCS(name, type, member, bits)			\
@@ -1219,11 +1225,15 @@ static inline void btrfs_set_##name(struct extent_buffer *eb,		\
 #define BTRFS_SETGET_STACK_FUNCS(name, type, member, bits)		\
 static inline u##bits btrfs_##name(const type *s)			\
 {									\
-	return le##bits##_to_cpu(s->member);				\
+	uint##bits##_t t;						\
+	memcpy(&t, &s->member, sizeof(s->member));			\
+	return le##bits##_to_cpu(t);					\
 }									\
 static inline void btrfs_set_##name(type *s, u##bits val)		\
 {									\
-	s->member = cpu_to_le##bits(val);				\
+	uint##bits##_t t;						\
+	t = cpu_to_le##bits(val);					\
+	memcpy(&s->member, &t, sizeof(s->member));			\
 }
 
 BTRFS_SETGET_FUNCS(device_type, struct btrfs_dev_item, type, 64);
volumes.c +3 -2
index 388c94e..102380b 100644
@@ -472,10 +472,11 @@ static int find_next_chunk(struct btrfs_root *root, u64 objectid, u64 *offset)
 		if (found_key.objectid != objectid)
 			*offset = 0;
 		else {
+			u64 t;
 			chunk = btrfs_item_ptr(path->nodes[0], path->slots[0],
 					       struct btrfs_chunk);
-			*offset = found_key.offset +
-				btrfs_chunk_length(path->nodes[0], chunk);
+			t = found_key.offset + btrfs_chunk_length(path->nodes[0], chunk);
+			memcpy(offset, &t, sizeof(found_key.offset));
 		}
 	}
 	ret = 0;
Signature · 1 line
2.1.0.rc1
#4 lore raw

Re: [PATCH 3/4] Fixing unaligned memory accesses.

Quoted 3 lines · Dimitri John Ledkov

From: Shawn Landen <[email protected]>

Bug-Debian: http://bugs.debian.org/656955

The bug seems old (2012) and agains 0.19. We've fixed a few unaligned access bugs in the meantime. Can you please retest with 3.16?

Quoted 12 lines
ctree.h +2 -0
@@ -19,6 +19,8 @@
 #ifndef __BTRFS__
 #define __BTRFS__
 
+#include <stdint.h>
+
 #if BTRFS_FLAT_INCLUDES
 #include "list.h"
 #include "kerncompat.h"

ctree.h struct btrfs_root { · line 1198

Quoted 7 lines
ctree.h +3 -1
@@ -1191,13 +1193,17 @@ struct btrfs_root {
 static inline u##bits btrfs_##name(const struct extent_buffer *eb)	\
 {									\
 	const struct btrfs_header *h = (struct btrfs_header *)eb->data;	\
-	return le##bits##_to_cpu(h->member);				\
+	uint##bits##_t t;						\
+	memcpy(&t, &h->member, sizeof(h->member));			\
+	return le##bits##_to_cpu(t);					\

The change to memcpy is safe, the compiler is smart enough to not emit any memcpy call for x86_64 and there's no change to the leXX_to_cpu macros.

However, I'd like to check first if this is really necessary due to the old version in the bugreport. I'd prefer using the u8/.../u64 types instead of the stdint.h ones, for sake of consistency with the rest of the codebase.

volumes.c static int find_next_chunk(struct btrfs_root *root, u64 objectid, u64 *offset) · line 479

Quoted 13 lines
volumes.c +3 -2
@@ -472,10 +472,11 @@ static int find_next_chunk(struct btrfs_root *root, u64 objectid, u64 *offset)
 		if (found_key.objectid != objectid)
 			*offset = 0;
 		else {
+			u64 t;
 			chunk = btrfs_item_ptr(path->nodes[0], path->slots[0],
 					       struct btrfs_chunk);
-			*offset = found_key.offset +
-				btrfs_chunk_length(path->nodes[0], chunk);
+			t = found_key.offset + btrfs_chunk_length(path->nodes[0], chunk);
+			memcpy(offset, &t, sizeof(found_key.offset));

That's not enough, there are more direct assignments to *offset in that function. The preferred way is to add 'put_unaligned' helper into kerncompat.h and use it.

#5 lore raw

PATCH 4/4 Default to acting like fsck.

In reply to Dimitri John Ledkov

Inspect arguments, if we are not called as btrfs, then assume we are called to act like fsck.

Bug-Debian: http://bugs.debian.org/712078

---
 btrfs.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
btrfs.c +1 -1
index e83349c..e8a87ac 100644
@@ -222,7 +222,7 @@ int main(int argc, char **argv)
 	else
 		bname = argv[0];
 
-	if (!strcmp(bname, "btrfsck")) {
+	if (strcmp(bname, "btrfs") != 0) {
 		argv[0] = "check";
 	} else {
 		argc--;
Signature · 1 line
2.1.0.rc1
#6 lore raw

Re: [PATCH 4/4] Default to acting like fsck.

Quoted 3 lines · Dimitri John Ledkov

Inspect arguments, if we are not called as btrfs, then assume we are called to act like fsck.

[...]

btrfs.c int main(int argc, char **argv) · line 225

Quoted 2 lines
btrfs.c +1 -1
-       if (!strcmp(bname, "btrfsck")) {
+       if (strcmp(bname, "btrfs") != 0) {

That's assuming a lot.

Silently (!) breaking people's btrfs-3.15_patched-DontRandomlyPanicV2 is a recipe for needless hair-pulling. Is there a reason for not using something less like strstr(bname, "fsck") that I am missing?

Regards,

T G-R

#7 lore raw

Re: [PATCH 4/4] Default to acting like fsck.

Quoted 5 lines · Tobias Geerinckx-Rice
Quoted 3 lines · Dimitri John Ledkov

Inspect arguments, if we are not called as btrfs, then assume we are called to act like fsck.

[...]

Quoted 9 lines

btrfs.c int main(int argc, char **argv) · line 225

Quoted 2 lines
btrfs.c +1 -1
-       if (!strcmp(bname, "btrfsck")) {
+       if (strcmp(bname, "btrfs") != 0) {

That's assuming a lot.

Silently (!) breaking people's btrfs-3.15_patched-DontRandomlyPanicV2 is a recipe for needless hair-pulling. Is there a reason for not using something less like strstr(bname, "fsck") that I am missing?

Quite. This is verbatim patch as I have currently applied in Debian packaging, and it was a fast fix to prevent breakage we had at one point.

Indeed using "strstr(bname, "fsck")" would be better and sufficient to resolve the problem we encountered (specifically fsck.btrfs -> btrfs not acting like btrfs). Also using strstr, would fix btrfsck.my-build to act like fsck tool.

I'll update this one patch.

Signature · 3 lines
Regards,

Dimitri.
#8 lore raw

Re: [PATCH 4/4] Default to acting like fsck.

Quoted 17 lines · Dimitri John Ledkov
Quoted 5 lines · Tobias Geerinckx-Rice
Quoted 3 lines · Dimitri John Ledkov

Inspect arguments, if we are not called as btrfs, then assume we are called to act like fsck.

[...]

Quoted 9 lines

btrfs.c int main(int argc, char **argv) · line 225

Quoted 2 lines
btrfs.c +1 -1
-       if (!strcmp(bname, "btrfsck")) {
+       if (strcmp(bname, "btrfs") != 0) {

That's assuming a lot.

Silently (!) breaking people's btrfs-3.15_patched-DontRandomlyPanicV2 is a recipe for needless hair-pulling. Is there a reason for not using something less like strstr(bname, "fsck") that I am missing?

Quoted 8 lines

Quite. This is verbatim patch as I have currently applied in Debian packaging, and it was a fast fix to prevent breakage we had at one point.

Indeed using "strstr(bname, "fsck")" would be better and sufficient to resolve the problem we encountered (specifically fsck.btrfs -> btrfs not acting like btrfs). Also using strstr, would fix btrfsck.my-build to act like fsck tool.

The intention was to provide backward compatibility shortcut for 'btrfsck' -> 'btrfs check' and nothing else. The referenced bug is again for 0.19 but there's an upstream-shipped stub fsck.btrfs (since 3.14) that should avoid any packaging tricks.