diff options
| author | Trond Myklebust <trond.myklebust@primarydata.com> | 2015-10-05 17:53:49 +0300 |
|---|---|---|
| committer | Trond Myklebust <trond.myklebust@primarydata.com> | 2015-10-08 15:27:04 +0300 |
| commit | edc1b01cd3b20a5fff049e98f82a2b0d24a34c89 (patch) | |
| tree | 1f6a52705eeec5ef5e450963cac1d7e38905f446 /include/linux/sunrpc | |
| parent | 66d7a56a6254389587d0999dcaab1d2634cd4e24 (diff) | |
| download | linux-edc1b01cd3b20a5fff049e98f82a2b0d24a34c89.tar.xz | |
SUNRPC: Move TCP receive data path into a workqueue context
Stream protocols such as TCP can often build up a backlog of data to be
read due to ordering. Combine this with the fact that some workloads such
as NFS read()-intensive workloads need to receive a lot of data per RPC
call, and it turns out that receiving the data from inside a softirq
context can cause starvation.
The following patch moves the TCP data receive into a workqueue context.
We still end up calling tcp_read_sock(), but we do so from a process
context, meaning that softirqs are enabled for most of the time.
With this patch, I see a doubling of read bandwidth when running a
multi-threaded iozone workload between a virtual client and server setup.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Diffstat (limited to 'include/linux/sunrpc')
| -rw-r--r-- | include/linux/sunrpc/xprtsock.h | 2 |
1 files changed, 2 insertions, 0 deletions
diff --git a/include/linux/sunrpc/xprtsock.h b/include/linux/sunrpc/xprtsock.h index 357e44c1a46b..0ece4ba06f06 100644 --- a/include/linux/sunrpc/xprtsock.h +++ b/include/linux/sunrpc/xprtsock.h @@ -44,6 +44,8 @@ struct sock_xprt { */ unsigned long sock_state; struct delayed_work connect_worker; + struct work_struct recv_worker; + struct mutex recv_mutex; struct sockaddr_storage srcaddr; unsigned short srcport; |