ceph: Delete features that are not used in the kernel
RFC PATCH 1/1 ceph: Delete features that are not used in the kernel
From: Leon Romanovsky <[email protected]>
The ceph_features.h has declaration of features that are not in-use in kernel code. This causes to seeing such compilation warnings in almost every kernel compilation.
./include/linux/ceph/ceph_features.h:14:24: warning: 'CEPH_FEATURE_UID' defined but not used [-Wunused-const-variable=]
14 | static const uint64_t CEPH_FEATURE_##name = (1ULL<<bit); \
| ^~~~~~~~~~~~~
./include/linux/ceph/ceph_features.h:75:1: note: in expansion of macro 'DEFINE_CEPH_FEATURE'
75 | DEFINE_CEPH_FEATURE( 0, 1, UID)
| ^~~~~~~~~~~~~~~~~~~The upstream kernel indeed doesn't have any use of them, so delete it.
---
I'm sending this as RFC because probably the patch is wrong, but I would like to bring your attention to the existing problem and asking for an acceptable solution.
--- include/linux/ceph/ceph_features.h | 9 +-------- 1 file changed, 1 insertion(+), 8 deletions(-)
include/linux/ceph/ceph_features.h +1 -8
@@ -72,13 +72,9 @@* the bit ever speak to each other. */ -DEFINE_CEPH_FEATURE( 0, 1, UID) -DEFINE_CEPH_FEATURE( 1, 1, NOSRCADDR) DEFINE_CEPH_FEATURE_RETIRED( 2, 1, MONCLOCKCHECK, JEWEL, LUMINOUS) -DEFINE_CEPH_FEATURE( 3, 1, FLOCK) DEFINE_CEPH_FEATURE( 4, 1, SUBSCRIBE2) -DEFINE_CEPH_FEATURE( 5, 1, MONNAMES) DEFINE_CEPH_FEATURE( 6, 1, RECONNECT_SEQ) DEFINE_CEPH_FEATURE( 7, 1, DIRLAYOUTHASH)DEFINE_CEPH_FEATURE (8 ,1 ,OBJECTLOCATOR )
@@ -176,10 +172,7 @@ DEFINE_CEPH_FEATURE_DEPRECATED(63, 1, RESERVED_BROKEN, LUMINOUS) // client-facin* Features supported. */ #define CEPH_FEATURES_SUPPORTED_DEFAULT \ - (CEPH_FEATURE_NOSRCADDR | \ - CEPH_FEATURE_FLOCK | \ - CEPH_FEATURE_SUBSCRIBE2 | \ - CEPH_FEATURE_RECONNECT_SEQ | \ + (CEPH_FEATURE_RECONNECT_SEQ | \ CEPH_FEATURE_DIRLAYOUTHASH | \ CEPH_FEATURE_PGID64 | \ CEPH_FEATURE_PGPOOL3 | \
--
2.26.2
Ilya Dryomov
maintainerQuoted 14 lines · Leon Romanovsky
From: Leon Romanovsky <[email protected]>
The
ceph_features.hhas declaration of features that are not in-use in kernel code. This causes to seeing such compilation warnings in almost every kernel compilation.
./include/linux/ceph/ceph_features.h:14:24: warning: 'CEPH_FEATURE_UID' defined but not used [-Wunused-const-variable=]14 | static const uint64_t CEPH_FEATURE_##name = (1ULL<<bit); \ | ^~~~~~~~~~~~~ ./include/linux/ceph/ceph_features.h:75:1: note: in expansion of macro 'DEFINE_CEPH_FEATURE' 75 | DEFINE_CEPH_FEATURE( 0, 1, UID) | ^~~~~~~~~~~~~~~~~~~
Quoted 7 lines
The upstream kernel indeed doesn't have any use of them, so delete it.
Signed-off-by Leon Romanovsky <[email protected]>---
I'm sending this as RFC because probably the patch is wrong, but I would like to bring your attention to the existing problem and asking for an acceptable solution.
Hi Leon,
Yes, removing unused feature definitions is wrong. Annotating them as potentially unused would be much better -- I'll send a patch.
I don't think any of us builds with W=1, so these things don't get noticed.
Thanks,
Ilya
Quoted 23 lines · Ilya Dryomov
Quoted 14 lines · Leon Romanovsky
From: Leon Romanovsky <[email protected]>
The
ceph_features.hhas declaration of features that are not in-use in kernel code. This causes to seeing such compilation warnings in almost every kernel compilation.
./include/linux/ceph/ceph_features.h:14:24: warning: 'CEPH_FEATURE_UID' defined but not used [-Wunused-const-variable=]14 | static const uint64_t CEPH_FEATURE_##name = (1ULL<<bit); \ | ^~~~~~~~~~~~~ ./include/linux/ceph/ceph_features.h:75:1: note: in expansion of macro 'DEFINE_CEPH_FEATURE' 75 | DEFINE_CEPH_FEATURE( 0, 1, UID) | ^~~~~~~~~~~~~~~~~~~Quoted 7 lines
The upstream kernel indeed doesn't have any use of them, so delete it.
Signed-off-by Leon Romanovsky <[email protected]>---
I'm sending this as RFC because probably the patch is wrong, but I would like to bring your attention to the existing problem and asking for an acceptable solution.
Quoted 7 lines
Hi Leon,
Yes, removing unused feature definitions is wrong. Annotating them as potentially unused would be much better -- I'll send a patch.
I don't think any of us builds with W=1, so these things don't get noticed.
Thanks, W=1 is our default compilation level for Mellanox RDMA submissions.
Quoted 4 lines
Thanks,
Ilya