Skip to contentSkip to conversation

md: dm-writeback: add __noreturn to BUG-ging function

PATCH In review Randy Dunlap 4 participants 10 messages Nov 13, 2020 – Nov 20, 2020

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

PATCH 1/1 md: dm-writeback: add __noreturn to BUG-ging function

Building on arch/s390/ flags this as an error, so add the __noreturn attribute modifier to prevent the build error.

cc1: some warnings being treated as errors
../drivers/md/dm-writecache.c: In function 'persistent_memory_claim':
../drivers/md/dm-writecache.c:323:1: error: no return statement in function returning non-void [-Werror=return-type]

Fixes 48debafe4f2f ("dm: add writecache target")
Cc Mikulas Patocka <[email protected]>
Cc Alasdair Kergon <[email protected]>
Cc Mike Snitzer <[email protected]>
Cc Heiko Carstens <[email protected]>
Cc Vasily Gorbik <[email protected]>
Cc Christian Borntraeger <[email protected]>
---
 drivers/md/dm-writecache.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/md/dm-writecache.c +1 -1
@@ -317,7 +317,7 @@ err1:
 	return r;
 }
 #else
-static int persistent_memory_claim(struct dm_writecache *wc)
+static int __noreturn persistent_memory_claim(struct dm_writecache *wc)
 {
 	BUG();
 }
#1 lore raw
Quoted 6 lines · Randy Dunlap

Building on arch/s390/ flags this as an error, so add the __noreturn attribute modifier to prevent the build error.

cc1: some warnings being treated as errors
../drivers/md/dm-writecache.c: In function 'persistent_memory_claim':
../drivers/md/dm-writecache.c:323:1: error: no return statement in function returning non-void [-Werror=return-type]

ok with me, but I am asking why

the unreachable macro is not good enough. For x86 it obviously is.

form arch/s390/include/asm/bug.h

#define BUG() do {                                      \
        __EMIT_BUG(0);                                  \
        unreachable();                                  \

} while (0)

Quoted 27 lines
Fixes 48debafe4f2f ("dm: add writecache target")
Cc Mikulas Patocka <[email protected]>
Cc Alasdair Kergon <[email protected]>
Cc Mike Snitzer <[email protected]>
Cc Heiko Carstens <[email protected]>
Cc Vasily Gorbik <[email protected]>
Cc Christian Borntraeger <[email protected]>
---
 drivers/md/dm-writecache.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/md/dm-writecache.c +1 -1
@@ -317,7 +317,7 @@ err1:
 	return r;
 }
 #else
-static int persistent_memory_claim(struct dm_writecache *wc)
+static int __noreturn persistent_memory_claim(struct dm_writecache *wc)
 {
 	BUG();
 }
#2 lore raw
Quoted 12 lines · Christian Borntraeger
Quoted 6 lines · Randy Dunlap

Building on arch/s390/ flags this as an error, so add the __noreturn attribute modifier to prevent the build error.

cc1: some warnings being treated as errors
../drivers/md/dm-writecache.c: In function 'persistent_memory_claim':
../drivers/md/dm-writecache.c:323:1: error: no return statement in function returning non-void [-Werror=return-type]

ok with me, but I am asking why

Quoted 8 lines

the unreachable macro is not good enough. For x86 it obviously is.

form arch/s390/include/asm/bug.h

#define BUG() do {                                      \
        __EMIT_BUG(0);                                  \
        unreachable();                                  \

} while (0)

Hi Christian,

Good question.
I don't see any guidance about when to use one or the other etc.

I see __noreturn being used 109 times and unreachable(); being used 33 times, but only now that I look at them. That had nothing to do with why I used __noreturn in the patch.

Quoted 16 lines
Quoted 16 lines
Fixes 48debafe4f2f ("dm: add writecache target")
Cc Mikulas Patocka <[email protected]>
Cc Alasdair Kergon <[email protected]>
Cc Mike Snitzer <[email protected]>
Cc Heiko Carstens <[email protected]>
Cc Vasily Gorbik <[email protected]>
Cc Christian Borntraeger <[email protected]>
---
 drivers/md/dm-writecache.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
Quoted 12 lines

drivers/md/dm-writecache.c err1: · line 323

Quoted 12 lines
drivers/md/dm-writecache.c +1 -1
@@ -317,7 +317,7 @@ err1:
 	return r;
 }
 #else
-static int persistent_memory_claim(struct dm_writecache *wc)
+static int __noreturn persistent_memory_claim(struct dm_writecache *wc)
 {
 	BUG();
 }

thanks.

Signature · 1 line
~Randy
#3 lore raw
Quoted 22 lines · Randy Dunlap
Quoted 12 lines · Christian Borntraeger
Quoted 6 lines · Randy Dunlap

Building on arch/s390/ flags this as an error, so add the __noreturn attribute modifier to prevent the build error.

cc1: some warnings being treated as errors
../drivers/md/dm-writecache.c: In function 'persistent_memory_claim':
../drivers/md/dm-writecache.c:323:1: error: no return statement in function returning non-void [-Werror=return-type]

ok with me, but I am asking why

Quoted 8 lines

the unreachable macro is not good enough. For x86 it obviously is.

form arch/s390/include/asm/bug.h

#define BUG() do {                                      \
        __EMIT_BUG(0);                                  \
        unreachable();                                  \

} while (0)

Quoted 8 lines

Hi Christian,

Good question.
I don't see any guidance about when to use one or the other etc.

I see __noreturn being used 109 times and unreachable(); being used 33 times, but only now that I look at them. That had nothing to do with why I used __noreturn in the patch.

But doesn't that speak to the proper fix being needed in unreachable()? Or at a minimum the fix is needed to arch/s390/include/asm/bug.h's BUG.

I really don't think we should be papering over that by sprinkling __noreturn around the kernel's BUG() callers.

Maybe switch arch/s390/include/asm/bug.h's BUG to be like arch/mips/include/asm/bug.h? It itself uses __noreturn with a 'static inline' function definition rather than #define.

Does that fix the issue?

Thanks,
Mike

p.s. you modified dm-writecache.c (not dm-writeback, wich doesn't exist).

#4 lore raw
Quoted 44 lines · Mike Snitzer
Quoted 22 lines · Randy Dunlap
Quoted 12 lines · Christian Borntraeger
Quoted 6 lines · Randy Dunlap

Building on arch/s390/ flags this as an error, so add the __noreturn attribute modifier to prevent the build error.

cc1: some warnings being treated as errors
../drivers/md/dm-writecache.c: In function 'persistent_memory_claim':
../drivers/md/dm-writecache.c:323:1: error: no return statement in function returning non-void [-Werror=return-type]

ok with me, but I am asking why

Quoted 8 lines

the unreachable macro is not good enough. For x86 it obviously is.

form arch/s390/include/asm/bug.h

#define BUG() do {                                      \
        __EMIT_BUG(0);                                  \
        unreachable();                                  \

} while (0)

Quoted 8 lines

Hi Christian,

Good question.
I don't see any guidance about when to use one or the other etc.

I see __noreturn being used 109 times and unreachable(); being used 33 times, but only now that I look at them. That had nothing to do with why I used __noreturn in the patch.

But doesn't that speak to the proper fix being needed in unreachable()? Or at a minimum the fix is needed to arch/s390/include/asm/bug.h's BUG.

I really don't think we should be papering over that by sprinkling __noreturn around the kernel's BUG() callers.

Maybe switch arch/s390/include/asm/bug.h's BUG to be like arch/mips/include/asm/bug.h? It itself uses __noreturn with a 'static inline' function definition rather than #define.

Quoted 7 lines

Does that fix the issue?

Thanks,
Mike

p.s. you modified dm-writecache.c (not dm-writeback, wich doesn't exist).

I don't think my suggestion will help.. given it'd still leave persistent_memory_claim() without a return statement.

Think it worthwhile to just add a dummy 'return 0;' after the BUG().

Mike

#6 lore raw
Quoted 11 lines · Mike Snitzer
Quoted 4 lines · Mike Snitzer

I don't think my suggestion will help.. given it'd still leave persistent_memory_claim() without a return statement.

Think it worthwhile to just add a dummy 'return 0;' after the BUG().

Decided to go with this, now staged for 5.11:
https://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm.git/commit/?h=dm-5.11&id=a1e4865b4dda7071f3707f7e551289ead66e38b1

Looks like a sane solution. Thank you for following up.

#7 lore raw
Quoted 9 lines · Christian Borntraeger
Quoted 6 lines · Mike Snitzer
Quoted 3 lines · Mike Snitzer

I don't think my suggestion will help.. given it'd still leave persistent_memory_claim() without a return statement.

Quoted 7 lines
Quoted 5 lines
Quoted 1 line

Think it worthwhile to just add a dummy 'return 0;' after the BUG().

Decided to go with this, now staged for 5.11:
https://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm.git/commit/?h=dm-5.11&id=a1e4865b4dda7071f3707f7e551289ead66e38b1

Looks like a sane solution. Thank you for following up.

Yes, thanks again.

Signature · 1 line
~Randy
maintainer
#8 lore raw
In reply to Mike Snitzer
Quoted 10 lines · Mike Snitzer
Quoted 4 lines · Mike Snitzer

I don't think my suggestion will help.. given it'd still leave persistent_memory_claim() without a return statement.

Think it worthwhile to just add a dummy 'return 0;' after the BUG().

Decided to go with this, now staged for 5.11:
https://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm.git/commit/?h=dm-5.11&id=a1e4865b4dda7071f3707f7e551289ead66e38b1

Hi

I would just use "return -EOPNOTSUPP;" and drop the "#ifdef DM_WRITECACHE_HAS_PMEM" that you added.

That BUG/return -EOPNOTSUPP code can't happen at all - if DM_WRITECACHE_HAS_PMEM is not defined, WC_MODE_PMEM(wc) always returns false - so persistent_memory_claim and BUG() can't ever be called. And if it can't be called, you don't need to add a code that prints an error in that case.

If we don't have DM_WRITECACHE_HAS_PMEM, the compiler optimizer will remove all the code guarded with if (WC_MODE_PMEM(wc)) as unreachable.

Mikulas

From: Mikulas Patocka <[email protected]>
Subject: [PATCH] dm writecache: remove BUG() and fail gracefully insteadfor-nextdm-5.11

Building on arch/s390/ results in this build error:

cc1: some warnings being treated as errors
../drivers/md/dm-writecache.c: In function 'persistent_memory_claim':
../drivers/md/dm-writecache.c:323:1: error: no return statement in function returning non-void [-Werror=return-type]

Fix this by replacing the BUG() with a -EOPNOTSUPP return.

Fixes 48debafe4f2f ("dm: add writecache target")
Cc [email protected] # v4.18+
drivers/md/dm-writecache.c +1 -1
Index: linux-2.6/drivers/md/dm-writecache.c
===================================================================
@@ -319,7 +319,7 @@ err1:
 #else
 static int persistent_memory_claim(struct dm_writecache *wc)
 {
-	BUG();
+	return -EOPNOTSUPP;
 }
 #endif
 
#9 lore raw
Quoted 26 lines · Mikulas Patocka
Quoted 10 lines · Mike Snitzer
Quoted 4 lines · Mike Snitzer

I don't think my suggestion will help.. given it'd still leave persistent_memory_claim() without a return statement.

Think it worthwhile to just add a dummy 'return 0;' after the BUG().

Decided to go with this, now staged for 5.11:
https://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm.git/commit/?h=dm-5.11&id=a1e4865b4dda7071f3707f7e551289ead66e38b1

Hi

I would just use "return -EOPNOTSUPP;" and drop the "#ifdef DM_WRITECACHE_HAS_PMEM" that you added.

That BUG/return -EOPNOTSUPP code can't happen at all - if DM_WRITECACHE_HAS_PMEM is not defined, WC_MODE_PMEM(wc) always returns false - so persistent_memory_claim and BUG() can't ever be called. And if it can't be called, you don't need to add a code that prints an error in that case.

Quoted 4 lines

If we don't have DM_WRITECACHE_HAS_PMEM, the compiler optimizer will remove all the code guarded with if (WC_MODE_PMEM(wc)) as unreachable.

Mikulas

Fair enough.