Hi Jason & Co, Thanks for picking this up. On Mon, 2017-12-18 at 16:14 -0700, Jason Gunthorpe wrote: > On Tue, Nov 28, 2017 at 11:58:07PM +0530, Bharat Potnuri wrote: > > isert commands that failed during isert_rdma_rw_ctx_post() are queued to > > Queue-Full(QF) queue and are scheduled to be reposted during queue-full > > queue processing. During this reposting, the rdma contexts are initialised > > again in isert_rdma_rw_ctx_post(), which is leaking significant memory. > > > > unreferenced object 0xffff8830201d9640 (size 64): > > comm "kworker/0:2", pid 195, jiffies 4295374851 (age 4528.436s) > > hex dump (first 32 bytes): > > 00 60 8b cb 2e 00 00 00 00 10 00 00 00 00 00 00 .`.............. > > 00 90 e3 cb 2e 00 00 00 00 10 00 00 00 00 00 00 ................ > > backtrace: > > [<ffffffff8170711e>] kmemleak_alloc+0x4e/0xb0 > > [<ffffffff811f8ba5>] __kmalloc+0x125/0x2b0 > > [<ffffffffa046b24f>] rdma_rw_ctx_init+0x15f/0x6f0 [ib_core] > > [<ffffffffa07ab644>] isert_rdma_rw_ctx_post+0xc4/0x3c0 [ib_isert] > > [<ffffffffa07ad972>] isert_put_datain+0x112/0x1c0 [ib_isert] > > [<ffffffffa07dddce>] lio_queue_data_in+0x2e/0x30 [iscsi_target_mod] > > [<ffffffffa076c322>] target_qf_do_work+0x2b2/0x4b0 [target_core_mod] > > [<ffffffff81080c3b>] process_one_work+0x1db/0x5d0 > > [<ffffffff8108107d>] worker_thread+0x4d/0x3e0 > > [<ffffffff81088667>] kthread+0x117/0x150 > > [<ffffffff81713fa7>] ret_from_fork+0x27/0x40 > > [<ffffffffffffffff>] 0xffffffffffffffff > > > > Here is patch to use the older rdma contexts while reposting > > the isert commands intead of reinitialising them. > > > > Signed-off-by: Potnuri Bharat Teja <bharat@xxxxxxxxxxx> > > Reviewed-by: Sagi Grimberg <sagi@xxxxxxxxxxx> > > drivers/infiniband/ulp/isert/ib_isert.c | 7 +++++++ > > drivers/infiniband/ulp/isert/ib_isert.h | 1 + > > 2 files changed, 8 insertions(+) > > Thanks, applied to RDMA for-next > > Jason Per Potnuri, it should be stable CC' to 4.11.y, and depends on: 7a56dc8 iser-target: avoid posting a recv buffer twice 555a65f iser-target: Fix queue-full response handling fa7e25c target: Fix unknown fabric callback queue-full errors -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
