| Message ID | 20210908052928.17375-1-rpearsonhpe@gmail.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | RDMA/rxe: Various bug fixes. | expand |
On 9/7/21 22:29, Bob Pearson wrote: > The first patch is a repeat of an earlier patch after rebasing to > for-next version 5.14.0-rc6+. Hi Bob, What is for-next version 5.14.0-rc6+? I tried to apply this series on v5.14, v5.14-rc6 and Jason's for-rc branch but 'git am' refused to apply this series on all these kernel versions ... Thanks, Bart.
On 9/8/21 10:52 PM, Bart Van Assche wrote: > On 9/7/21 22:29, Bob Pearson wrote: >> The first patch is a repeat of an earlier patch after rebasing to >> for-next version 5.14.0-rc6+. > > Hi Bob, > > What is for-next version 5.14.0-rc6+? I tried to apply this series on v5.14, v5.14-rc6 and Jason's for-rc branch but 'git am' refused to apply this series on all these kernel versions ... > > Thanks, > > Bart. Bart, I think I fixed the git am issue and resubmitted to the correct queue. You were on copy. Please try to see if this gets you to joy for blktest. The one case you mentioned srp/002 is working OK for me now. Bob
Currently some blktest test cases fail to execute correctly. This series of patches allows these test cases to pass and implements some other related changes. The first patch is a repeat of an earlier patch after rebasing to for-next version 5.14.0-rc6+. It adds memory barriers to kernel to kernel queues. The logic for this is the same as an earlier patch that only treated user to kernel queues. Without this patch kernel to kernel queues are expected to intermittently fail at low frequency. The second patch is also a repeat after rebasing to for-next version 5.14.0-rc6+. It fixes a multicast bug. The third patch separates the keys in rxe_mr and ib_mr. This allows the sequence do { ib_post_send( IB_WR_LOCAL_INV ) ib_update_fast_reg_key() ib_map_mr_sg() ib_post_send( IB_WR_REG_MR ) } while ( !done ) to operate correctly since now the changes to the local copy of the keys only happens at the final post_send. The same fix is applied to MWs since they need to synchronize with IB_WR_BIND_MW work requests. The fourth patch creates duplicate mapping tables for fast MRs. This prevents rkeys referencing fast MRs from accessing data from an updated map after the call to ib_map_mr_sg() call by keeping the new and old mappings separate. The final patch cleans up the state and type enums used by MRs. Bob Pearson (5): RDMA/rxe: Add memory barriers to kernel queues RDMA/rxe: Fix memory allocation while locked RDMA/rxe: Separate HW and SW l/rkeys RDMA/rxe: Create duplicate mapping tables for FMRs RDMA/rxe: Cleanup MR status and type enums drivers/infiniband/sw/rxe/rxe_comp.c | 10 +- drivers/infiniband/sw/rxe/rxe_cq.c | 25 +-- drivers/infiniband/sw/rxe/rxe_loc.h | 2 + drivers/infiniband/sw/rxe/rxe_mcast.c | 2 +- drivers/infiniband/sw/rxe/rxe_mr.c | 269 +++++++++++++++++++------- drivers/infiniband/sw/rxe/rxe_mw.c | 36 ++-- drivers/infiniband/sw/rxe/rxe_qp.c | 10 +- drivers/infiniband/sw/rxe/rxe_queue.h | 73 ++----- drivers/infiniband/sw/rxe/rxe_req.c | 35 +--- drivers/infiniband/sw/rxe/rxe_resp.c | 38 +--- drivers/infiniband/sw/rxe/rxe_srq.c | 2 +- drivers/infiniband/sw/rxe/rxe_verbs.c | 92 +++------ drivers/infiniband/sw/rxe/rxe_verbs.h | 48 ++--- 13 files changed, 307 insertions(+), 335 deletions(-)