RDMA/siw: Always consume all skbuf data in sk_data_ready() upcall.
authorBernard Metzler <bmt@zurich.ibm.com>
Tue, 20 Sep 2022 08:12:02 +0000 (10:12 +0200)
committerLeon Romanovsky <leon@kernel.org>
Tue, 20 Sep 2022 18:21:18 +0000 (21:21 +0300)
commit754209850df8367c954ac1de7671c7430b1f342c
treed464c87a6dbe50bac753de8151e7f8cafadbc95f
parent0227f4d0d15433c34f5dca68817c0d12ca244feb
RDMA/siw: Always consume all skbuf data in sk_data_ready() upcall.

For header and trailer/padding processing, siw did not consume new
skb data until minimum amount present to fill current header or trailer
structure, including potential payload padding. Not consuming any
data during upcall may cause a receive stall, since tcp_read_sock()
is not upcalling again if no new data arrive.
A NFSoRDMA client got stuck at RDMA Write reception of unaligned
payload, if the current skb did contain only the expected 3 padding
bytes, but not the 4 bytes CRC trailer. Expecting 4 more bytes already
arrived in another skb, and not consuming those 3 bytes in the current
upcall left the Write incomplete, waiting for the CRC forever.

Fixes: 8b6a361b8c48 ("rdma/siw: receive path")
Reported-by: Olga Kornievskaia <kolga@netapp.com>
Tested-by: Olga Kornievskaia <kolga@netapp.com>
Signed-off-by: Bernard Metzler <bmt@zurich.ibm.com>
Link: https://lore.kernel.org/r/20220920081202.223629-1-bmt@zurich.ibm.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
drivers/infiniband/sw/siw/siw_qp_rx.c