Skip to contentSkip to conversation

-tip: block, fuse: Fix build error in fs/fuse/dev.c

PATCH Ingo Molnar 2 participants 4 messages Jul 11, 2009

Subsystems (5)
lore
Share & tools
lore
#0 lore raw

Today's upstream tree build (x86, 64-bit, allnoconfig+CONFIG_FUSE) failed with this build error:

fs/fuse/dev.c: In function ‘request_end’:
fs/fuse/dev.c:290: error: ‘BLK_RW_SYNC’ undeclared (first use in this function) fs/fuse/dev.c:290: error: (Each undeclared identifier is reported only once fs/fuse/dev.c:290: error: for each function it appears in.) fs/fuse/dev.c:291: error: ‘BLK_RW_ASYNC’ undeclared (first use in this function)

Due to commit:

8aa7e84: Fix congestion_wait() sync/async vs read/write confusion

Creating a dependency of BDI callbacks on the BLK_RW_* constants, while those constants are only defined if CONFIG_BLOCK is enabled.

Fix it the simplest way for now, by moving the definitions early in blkdev.h (this cannot break anything), but the real fix would be to split up blkdev.h into blkdev-types.h and blkdev-api.h and make only the API definitions/declarations dependent on CONFIG_BLOCK.

Cc Jens Axboe <[email protected]>
Cc Miklos Szeredi <[email protected]>
---
 include/linux/blkdev.h |   13 ++++++++-----
 1 files changed, 8 insertions(+), 5 deletions(-)
include/linux/blkdev.h +8 -5
index 49ae079..a3cdc6d 100644
@@ -1,6 +1,14 @@
 #ifndef _LINUX_BLKDEV_H
 #define _LINUX_BLKDEV_H
 
+/*
+ * Definitions used on the !CONFIG_BLOCK case too:
+ */
+enum {
+	BLK_RW_ASYNC	= 0,
+	BLK_RW_SYNC	= 1,
+};
+
 #ifdef CONFIG_BLOCK
 
 #include <linux/sched.h>
@@ -70,11 +78,6 @@ enum rq_cmd_type_bits {
 	REQ_TYPE_ATA_PC,
 };
 
-enum {
-	BLK_RW_ASYNC	= 0,
-	BLK_RW_SYNC	= 1,
-};
-
 /*
  * For request of type REQ_TYPE_LINUX_BLOCK, rq->cmd[0] is the opcode being
  * sent down (similar to how REQ_TYPE_BLOCK_PC means that ->cmd[] holds a
#1 lore raw

Hi Ingo,

Quoted 50 lines · Ingo Molnar

Today's upstream tree build (x86, 64-bit, allnoconfig+CONFIG_FUSE) failed with this build error:

fs/fuse/dev.c: In function ‘request_end’:
fs/fuse/dev.c:290: error: ‘BLK_RW_SYNC’ undeclared (first use in this function) fs/fuse/dev.c:290: error: (Each undeclared identifier is reported only once fs/fuse/dev.c:290: error: for each function it appears in.) fs/fuse/dev.c:291: error: ‘BLK_RW_ASYNC’ undeclared (first use in this function)

Due to commit:

8aa7e84: Fix congestion_wait() sync/async vs read/write confusion

Creating a dependency of BDI callbacks on the BLK_RW_* constants, while those constants are only defined if CONFIG_BLOCK is enabled.

Fix it the simplest way for now, by moving the definitions early in blkdev.h (this cannot break anything), but the real fix would be to split up blkdev.h into blkdev-types.h and blkdev-api.h and make only the API definitions/declarations dependent on CONFIG_BLOCK.

Cc Jens Axboe <[email protected]>
Cc Miklos Szeredi <[email protected]>
---
 include/linux/blkdev.h |   13 ++++++++-----
 1 files changed, 8 insertions(+), 5 deletions(-)
include/linux/blkdev.h +8 -0
index 49ae079..a3cdc6d 100644
@@ -1,6 +1,14 @@
 #ifndef _LINUX_BLKDEV_H
 #define _LINUX_BLKDEV_H
 
+/*
+ * Definitions used on the !CONFIG_BLOCK case too:
+ */
+enum {
+	BLK_RW_ASYNC	= 0,
+	BLK_RW_SYNC	= 1,
+};
+
 #ifdef CONFIG_BLOCK
 
 #include <linux/sched.h>
@@ -70,11 +78,6 @@ enum rq_cmd_type_bits {
 	REQ_TYPE_ATA_PC,

include/linux/blkdev.h enum rq_cmd_type_bits { · line 83

Quoted 10 lines
include/linux/blkdev.h +0 -5
 };
 
-enum {
-	BLK_RW_ASYNC	= 0,
-	BLK_RW_SYNC	= 1,
-};
-
 /*
  * For request of type REQ_TYPE_LINUX_BLOCK, rq->cmd[0] is the opcode being
  * sent down (similar to how REQ_TYPE_BLOCK_PC means that ->cmd[] holds a

I have hit this build issue as well, but with CONFIG_BLOCK=y, so the patch above does NOT fix it. The problem I see is that fs/fuse/dev.c does not include <linux/blkdev.h> so it doesn't get the definitions regardless of the configuration. Same problem in fs/nfs/write.c. The patch below fixes it for me:

* * * * *

From: Jean Delvare <[email protected]>
Subject: Fix BLK_RW_* build failures

Fix build failures caused by using BLK_RW_* constants without including <linux/blkdev.h>.

Cc Jens Axboe <[email protected]>
Cc Miklos Szeredi <[email protected]>
Cc Ingo Molnar <[email protected]>

---
I have included pktcdvd and memcontrol in the patch, they don't build for my config so I didn't see them fail, but I suspect they would, so fix them too.

 drivers/block/pktcdvd.c |    1 +
 fs/fuse/dev.c           |    1 +
 fs/nfs/write.c          |    1 +
 mm/memcontrol.c         |    1 +
 4 files changed, 4 insertions(+)
drivers/block/pktcdvd.c +1 -0
@@ -62,6 +62,7 @@
 #include <scsi/scsi.h>
 #include <linux/debugfs.h>
 #include <linux/device.h>
+#include <linux/blkdev.h>
 
 #include <asm/uaccess.h>
 
fs/fuse/dev.c +1 -0
@@ -16,6 +16,7 @@
 #include <linux/pagemap.h>
 #include <linux/file.h>
 #include <linux/slab.h>
+#include <linux/blkdev.h>
 
 MODULE_ALIAS_MISCDEV(FUSE_MINOR);
 
fs/nfs/write.c +1 -0
@@ -19,6 +19,7 @@
 #include <linux/nfs_mount.h>
 #include <linux/nfs_page.h>
 #include <linux/backing-dev.h>
+#include <linux/blkdev.h>
 
 #include <asm/uaccess.h>
 
mm/memcontrol.c +1 -0
@@ -37,6 +37,7 @@
 #include <linux/vmalloc.h>
 #include <linux/mm_inline.h>
 #include <linux/page_cgroup.h>
+#include <linux/blkdev.h>
 #include "internal.h"
 
 #include <asm/uaccess.h>
Signature · 1 line
Jean Delvare
#2 lore raw
Quoted 55 lines · Jean Delvare

Hi Ingo,

Quoted 52 lines · Ingo Molnar

Today's upstream tree build (x86, 64-bit, allnoconfig+CONFIG_FUSE) failed with this build error:

fs/fuse/dev.c: In function ‘request_end’:
fs/fuse/dev.c:290: error: ‘BLK_RW_SYNC’ undeclared (first use in this function) fs/fuse/dev.c:290: error: (Each undeclared identifier is reported only once fs/fuse/dev.c:290: error: for each function it appears in.) fs/fuse/dev.c:291: error: ‘BLK_RW_ASYNC’ undeclared (first use in this function)

Due to commit:

8aa7e84: Fix congestion_wait() sync/async vs read/write confusion

Creating a dependency of BDI callbacks on the BLK_RW_* constants, while those constants are only defined if CONFIG_BLOCK is enabled.

Fix it the simplest way for now, by moving the definitions early in blkdev.h (this cannot break anything), but the real fix would be to split up blkdev.h into blkdev-types.h and blkdev-api.h and make only the API definitions/declarations dependent on CONFIG_BLOCK.

Cc Jens Axboe <[email protected]>
Cc Miklos Szeredi <[email protected]>
---
 include/linux/blkdev.h |   13 ++++++++-----
 1 files changed, 8 insertions(+), 5 deletions(-)
include/linux/blkdev.h +8 -0
index 49ae079..a3cdc6d 100644
@@ -1,6 +1,14 @@
 #ifndef _LINUX_BLKDEV_H
 #define _LINUX_BLKDEV_H
 
+/*
+ * Definitions used on the !CONFIG_BLOCK case too:
+ */
+enum {
+	BLK_RW_ASYNC	= 0,
+	BLK_RW_SYNC	= 1,
+};
+
 #ifdef CONFIG_BLOCK
 
 #include <linux/sched.h>
@@ -70,11 +78,6 @@ enum rq_cmd_type_bits {
 	REQ_TYPE_ATA_PC,
 };
 
Quoted 14 lines

include/linux/blkdev.h enum rq_cmd_type_bits { · line 83

Quoted 8 lines
include/linux/blkdev.h +0 -5
-enum {
-	BLK_RW_ASYNC	= 0,
-	BLK_RW_SYNC	= 1,
-};
-
 /*
  * For request of type REQ_TYPE_LINUX_BLOCK, rq->cmd[0] is the opcode being
  * sent down (similar to how REQ_TYPE_BLOCK_PC means that ->cmd[] holds a

I have hit this build issue as well, but with CONFIG_BLOCK=y, so the patch above does NOT fix it. The problem I see is that fs/fuse/dev.c does not include <linux/blkdev.h> so it doesn't get the definitions regardless of the configuration. Same problem in fs/nfs/write.c. The patch below fixes it for me:

Yes, that's a different build bug already fixed in latest upstream:

097041e: fuse: Fix build error

Ingo

#3 lore raw
Quoted 11 lines · Ingo Molnar
Quoted 5 lines · Jean Delvare

I have hit this build issue as well, but with CONFIG_BLOCK=y, so the patch above does NOT fix it. The problem I see is that fs/fuse/dev.c does not include <linux/blkdev.h> so it doesn't get the definitions regardless of the configuration. Same problem in fs/nfs/write.c. The patch below fixes it for me:

Yes, that's a different build bug already fixed in latest upstream:

097041e: fuse: Fix build error

Ah, I had missed that one because it has not been snapshot yet. Thanks for pointing it out.

Signature · 1 line
Jean Delvare