workqueue: fix enum type for gcc-13
PATCH 1/1 workqueue: fix enum type for gcc-13
From: Arnd Bergmann <[email protected]>
In gcc-13, the WORK_STRUCT_WQ_DATA_MASK constant is a signed 64-bit type on 32-bit architectures because the enum definition has both negative numbers and numbers above LONG_MAX in it:
kernel/workqueue.c: In function 'get_work_pwq':kernel/workqueue.c:709:24: error: cast to pointer from integer of different size [-Werror=int-to-pointer-cast]
709 | return (void *)(data & WORK_STRUCT_WQ_DATA_MASK);
| ^kernel/workqueue.c: In function 'get_work_pool':kernel/workqueue.c:737:25: error: cast to pointer from integer of different size [-Werror=int-to-pointer-cast]
737 | return ((struct pool_workqueue *)
| ^kernel/workqueue.c: In function 'get_work_pool_id':kernel/workqueue.c:759:25: error: cast to pointer from integer of different size [-Werror=int-to-pointer-cast]
759 | return ((struct pool_workqueue *)
| ^
Change the enum definition to ensure all values can fit into the range of 'unsigned long' on all architectures.
--- include/linux/workqueue.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
include/linux/workqueue.h +1 -1
@@ -83,7 +83,7 @@ enum {/* convenience constants */ WORK_STRUCT_FLAG_MASK = (1UL <<WORK_STRUCT_FLAG_BITS ) -1 , -WORK_STRUCT_WQ_DATA_MASK = ~WORK_STRUCT_FLAG_MASK , +WORK_STRUCT_WQ_DATA_MASK = (unsigned long )~WORK_STRUCT_FLAG_MASK , WORK_STRUCT_NO_POOL = (unsigned long)WORK_OFFQ_POOL_NONE << WORK_OFFQ_POOL_SHIFT,/* bit mask for work_busy() return values */
Signature · 1 line
2.39.0
From: Arnd Bergmann
Quoted 6 lines
Sent: 17 January 2023 16:41
In gcc-13, the
WORK_STRUCT_WQ_DATA_MASKconstant is a signed 64-bit type on 32-bit architectures because the enum definition has both negative numbers and numbers aboveLONG_MAXin it:
...
include/linux/workqueue.h enum { · line 87
Quoted 5 lines
include/linux/workqueue.h +1
-1/* convenience constants */ WORK_STRUCT_FLAG_MASK = (1UL <<WORK_STRUCT_FLAG_BITS ) -1 , -WORK_STRUCT_WQ_DATA_MASK = ~WORK_STRUCT_FLAG_MASK , +WORK_STRUCT_WQ_DATA_MASK = (unsigned long )~WORK_STRUCT_FLAG_MASK , WORK_STRUCT_NO_POOL = (unsigned long)WORK_OFFQ_POOL_NONE << WORK_OFFQ_POOL_SHIFT,
How can that make any difference?
You aren't changing the value or type (which makes no difference) of WORK_STRUCT_WQ_DATA_MASK.
Indeed you require it to have the high bit set.
So if the enum contains a -1 somewhere gcc-13 will promote everything to s64.
Either declare gcc-13 as 'BROKEN' or change the:
return (void *)(data & WORK_STRUCT_WQ_DATA_MASK); to return (void *)(data & ~WORK_STRUCT_FLAG_MASK);
or use #defines.
David
-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales)
Tejun Heo
maintainerHello,
Quoted 19 lines · David Laight
From: Arnd Bergmann
Quoted 6 lines
Sent: 17 January 2023 16:41
In gcc-13, the
WORK_STRUCT_WQ_DATA_MASKconstant is a signed 64-bit type on 32-bit architectures because the enum definition has both negative numbers and numbers aboveLONG_MAXin it:...
include/linux/workqueue.h enum { · line 87
Quoted 5 lines
include/linux/workqueue.h +1
-1/* convenience constants */ WORK_STRUCT_FLAG_MASK = (1UL <<WORK_STRUCT_FLAG_BITS ) -1 , -WORK_STRUCT_WQ_DATA_MASK = ~WORK_STRUCT_FLAG_MASK , +WORK_STRUCT_WQ_DATA_MASK = (unsigned long )~WORK_STRUCT_FLAG_MASK , WORK_STRUCT_NO_POOL = (unsigned long)WORK_OFFQ_POOL_NONE << WORK_OFFQ_POOL_SHIFT,How can that make any difference?
You aren't changing the value or type (which makes no difference) ofWORK_STRUCT_WQ_DATA_MASK.
Indeed you require it to have the high bit set.
Quoted 4 lines
So if the enum contains a -1 somewhere gcc-13 will promote everything to s64.
Either declare gcc-13 as 'BROKEN' or change the:
I have a hard time understanding why gcc would change its behavior so that there's no way to compile the same code in a consistent manner across two adjacent compiler versions. The new behavior is fine but it makes no sense to introduce it like this. If at all possible, marking gcc13 broken sounds about right to me.
Thanks.
Signature · 1 line
tejun
Quoted 8 lines · Tejun Heo
Quoted 8 lines
From: Arnd Bergmann
Quoted 6 lines
Sent: 17 January 2023 16:41
In gcc-13, the
WORK_STRUCT_WQ_DATA_MASKconstant is a signed 64-bit type on 32-bit architectures because the enum definition has both negative numbers and numbers aboveLONG_MAXin it:...
Quoted 11 lines
Quoted 5 lines
include/linux/workqueue.h enum { · line 87
Quoted 5 lines
include/linux/workqueue.h +1
-1/* convenience constants */ WORK_STRUCT_FLAG_MASK = (1UL <<WORK_STRUCT_FLAG_BITS ) -1 , -WORK_STRUCT_WQ_DATA_MASK = ~WORK_STRUCT_FLAG_MASK , +WORK_STRUCT_WQ_DATA_MASK = (unsigned long )~WORK_STRUCT_FLAG_MASK , WORK_STRUCT_NO_POOL = (unsigned long)WORK_OFFQ_POOL_NONE << WORK_OFFQ_POOL_SHIFT,I have a hard time understanding why gcc would change its behavior so that there's no way to compile the same code in a consistent manner across two adjacent compiler versions. The new behavior is fine but it makes no sense to introduce it like this. If at all possible, marking gcc13 broken sounds about right to me.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107405 has some more information on the change. In short, the old behavior was a gcc extension that was somewhat surprising and not well documented, while the new behavior is consistent with C23 and C++ as well as easier to understand: Any constant that is defined as part of an enum now has the same type as the enum itself, even if it fits within a shorter type. In a definition like
enum e {
A = -1,
B = -1u,
};
the enum type has to be compatible with 'long long' because anything shorter would not fit both -1 and -1u (UINT_MAX). A and B were both signed types to match the signedness of the enum type, but A was actually a 32-bit integer since that is sufficient, while B was also a 64-bit type since it exceeds INT_MAX. Now they are both the same type.
I don't think there is a chance they will revert to the old behavior, though we could try asking for an command line flag to warn about cases where this changes code generation.
Arnd
Tejun Heo
maintainerQuoted 6 lines · Arnd Bergmann
the enum type has to be compatible with 'long long' because anything shorter would not fit both -1 and -1u (
UINT_MAX). A and B were both signed types to match the signedness of the enum type, but A was actually a 32-bit integer since that is sufficient, while B was also a 64-bit type since it exceedsINT_MAX. Now they are both the same type.
Yeah, the new behavior makes total sense.
Quoted 3 lines
I don't think there is a chance they will revert to the old behavior, though we could try asking for an command line flag to warn about cases where this changes code generation.
but the way that it's transitioning doesn't really make sense to me. We do phase out support for old compilers, so it isn't that difficult to manage this transition smoothly - add a compat option in the new versions, switch code later once old compilers are phased out and drop the compat flag.
The way it's currently done causes situations which can't be handled logically - we end up having to regroup enums based on their value range rather than the logical type they need to be because there's no way to override the enum type on older compilers.
Thanks.
Signature · 1 line
tejun
From: Tejun Heo
Quoted 21 lines
Sent: 19 January 2023 01:42
Quoted 6 lines · Arnd Bergmann
the enum type has to be compatible with 'long long' because anything shorter would not fit both -1 and -1u (
UINT_MAX). A and B were both signed types to match the signedness of the enum type, but A was actually a 32-bit integer since that is sufficient, while B was also a 64-bit type since it exceedsINT_MAX. Now they are both the same type.Yeah, the new behavior makes total sense.
Quoted 3 lines
I don't think there is a chance they will revert to the old behavior, though we could try asking for an command line flag to warn about cases where this changes code generation.
but the way that it's transitioning doesn't really make sense to me. We do phase out support for old compilers, so it isn't that difficult to manage this transition smoothly - add a compat option in the new versions, switch code later once old compilers are phased out and drop the compat flag.
Quoted 4 lines
The way it's currently done causes situations which can't be handled logically - we end up having to regroup enums based on their value range rather than the logical type they need to be because there's no way to override the enum type on older compilers.
Didn't someone say that either C++ or C23 defines a syntax to explicitly set the type of the enum.
IIRC gcc-13 doesn't support that.
If gcc-13 supported explicit setting of the type and also generated a warn/error enum with 'large' values when the type wasn't specified (ie where the gcc extension had been used) then at least there would be no obscure surprises.
Getting an old gcc to ignore the type is easy.
David
-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales)
Quoted 20 lines · Arnd Bergmann
From: Arnd Bergmann <[email protected]>
In gcc-13, the
WORK_STRUCT_WQ_DATA_MASKconstant is a signed 64-bit type on 32-bit architectures because the enum definition has both negative numbers and numbers aboveLONG_MAXin it:
kernel/workqueue.c: In function 'get_work_pwq':kernel/workqueue.c:709:24: error: cast to pointer from integer of different size [-Werror=int-to-pointer-cast]
709 | return (void *)(data &WORK_STRUCT_WQ_DATA_MASK);
| ^kernel/workqueue.c: In function 'get_work_pool':kernel/workqueue.c:737:25: error: cast to pointer from integer of different size [-Werror=int-to-pointer-cast]
737 | return ((structpool_workqueue*)
| ^kernel/workqueue.c: In function 'get_work_pool_id':kernel/workqueue.c:759:25: error: cast to pointer from integer of different size [-Werror=int-to-pointer-cast]
759 | return ((structpool_workqueue*)
| ^
Quoted 7 lines
Change the enum definition to ensure all values can fit into the range of 'unsigned long' on all architectures.
Signed-off-by Arnd Bergmann <[email protected]>--- include/linux/workqueue.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
Quoted 33 lines · Arnd Bergmann
From: Arnd Bergmann <[email protected]>
In gcc-13, the
WORK_STRUCT_WQ_DATA_MASKconstant is a signed 64-bit type on 32-bit architectures because the enum definition has both negative numbers and numbers aboveLONG_MAXin it:
kernel/workqueue.c: In function 'get_work_pwq':kernel/workqueue.c:709:24: error: cast to pointer from integer of different size [-Werror=int-to-pointer-cast]
709 | return (void *)(data &WORK_STRUCT_WQ_DATA_MASK);
| ^kernel/workqueue.c: In function 'get_work_pool':kernel/workqueue.c:737:25: error: cast to pointer from integer of different size [-Werror=int-to-pointer-cast]
737 | return ((structpool_workqueue*)
| ^kernel/workqueue.c: In function 'get_work_pool_id':kernel/workqueue.c:759:25: error: cast to pointer from integer of different size [-Werror=int-to-pointer-cast]
759 | return ((structpool_workqueue*)
| ^Change the enum definition to ensure all values can fit into the range of 'unsigned long' on all architectures.
Signed-off-by Arnd Bergmann <[email protected]>--- include/linux/workqueue.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)include/linux/workqueue.h +0
-0
include/linux/workqueue.h enum { · line 87
Quoted 6 lines
include/linux/workqueue.h +1
-1@@ -83,7 +83,7 @@ enum {/* convenience constants */ WORK_STRUCT_FLAG_MASK = (1UL <<WORK_STRUCT_FLAG_BITS ) -1 , -WORK_STRUCT_WQ_DATA_MASK = ~WORK_STRUCT_FLAG_MASK , +WORK_STRUCT_WQ_DATA_MASK = (unsigned long )~WORK_STRUCT_FLAG_MASK , WORK_STRUCT_NO_POOL = (unsigned long)WORK_OFFQ_POOL_NONE << WORK_OFFQ_POOL_SHIFT,
Tested with "old" gcc: 7.3.1
But I personally prefer to redefine them as non-enum.
Quoted 5 lines
include/linux/workqueue.h +0
-1/* bit mask for work_busy() return values */ --2.39.0
Applied to wq/for-6.5.
Thanks.
Signature · 1 line
tejun