> =
> > BUG: KASAN: stack-out-of-bounds in rxe_post_send+0x77d/0x9b0 [rdma_rxe]
> > Read of size 8 at addr ffff880061aef860 by task 01/1080
> >
> > CPU: 2 PID: 1080 Comm: 01 Not tainted 4.16.0-rc3-dbg+ #2
> > Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.0.0-
> prebuilt.qemu-project.org 04/01/2014
> > Call Trace:
> > dump_stack+0x85/0xc7
> > print_address_description+0x65/0x270
> > kasan_report+0x231/0x350
> > rxe_post_send+0x77d/0x9b0 [rdma_rxe]
> > __ib_drain_sq+0x1ad/0x250 [ib_core]
> > ib_drain_qp+0x9/0x30 [ib_core]
> > srp_destroy_qp+0x51/0x70 [ib_srp]
> > srp_free_ch_ib+0xfc/0x380 [ib_srp]
> > srp_create_target+0x1071/0x19e0 [ib_srp]
> > kernfs_fop_write+0x180/0x210
> > __vfs_write+0xb1/0x2e0
> > vfs_write+0xf6/0x250
> > SyS_write+0x99/0x110
> > do_syscall_64+0xee/0x2b0
> > entry_SYSCALL_64_after_hwframe+0x42/0xb7
> >
> > The buggy address belongs to the page:
> > page:ffffea000186bbc0 count:0 mapcount:0 mapping:0000000000000000
> index:0x0
> > flags: 0x4000000000000000()
> > raw: 4000000000000000 0000000000000000 0000000000000000
> 00000000ffffffff
> > raw: 0000000000000000 ffffea000186bbe0 0000000000000000
> 0000000000000000
> > page dumped because: kasan: bad access detected
> >
> > Memory state around the buggy address:
> > ffff880061aef700: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
> > ffff880061aef780: 00 00 00 00 00 00 00 00 00 00 00 f1 f1 f1 f1 00
> > >ffff880061aef800: f2 f2 f2 f2 f2 f2 f2 00 00 00 00 00 f2 f2 f2 f2
> > ^
> > ffff880061aef880: f2 f2 f2 00 00 00 00 00 00 00 00 00 00 00 f2 f2
> > ffff880061aef900: f2 f2 f2 00 00 00 00 00 00 00 00 00 00 00 00 00
> >
> =================================================================
> =
> >
> > Fixes: 765d67748bcf ("IB: new common API for draining queues")
> > Signed-off-by: Bart Van Assche <bart.vanassche@xxxxxxx>
> > Cc: Steve Wise <swise@xxxxxxxxxxxxxxxxxxxxx>
> > Cc: Sagi Grimberg <sagi@xxxxxxxxxxx>
> > Cc: stable@xxxxxxxxxxxxxxx
> > drivers/infiniband/core/verbs.c | 11 ++++++++---
> > 1 file changed, 8 insertions(+), 3 deletions(-)
> >
> > diff --git a/drivers/infiniband/core/verbs.c
b/drivers/infiniband/core/verbs.c
> > index 2c7b0ceb46e6..4e2b231b03f7 100644
> > +++ b/drivers/infiniband/core/verbs.c
> > @@ -2194,7 +2194,13 @@ static void __ib_drain_sq(struct ib_qp *qp)
> > struct ib_cq *cq = qp->send_cq;
> > struct ib_qp_attr attr = { .qp_state = IB_QPS_ERR };
> > struct ib_drain_cqe sdrain;
> > - struct ib_send_wr swr = {}, *bad_swr;
> > + struct ib_send_wr *bad_swr;
> > + struct ib_rdma_wr swr = {
> > + .wr = {
> > + .opcode = IB_WR_RDMA_WRITE,
> > + .wr_cqe = &sdrain.cqe,
> > + },
> > + };
>
> I don't get it..
>
> Since when did ib_post_send() start requiring a ib_rdma_wr?
>
> IB_WR_RDMA_WRITE == 0, so even missing that is 'OK' but ugly.
>
> What is the actual bug here?
>
The WRs are now split up, so struct ib_send_wr doesn't encompass the full
size of all the possible WRs. See ib_rdma_wr, for example, which includes
ib_send_wr. So the bug is the drain code is posting a WRITE wr, but not
including the entire struct ib_rdma_wr.
Steve