Hi folks, patch 1 fixes the bug reported by Yegor Yefremov <yegorslists@xxxxxxxxxxxxxx>. patches 2 - 4 improve IRQ handling a little bit. Tested with BeagleBone Black and AM437x SK. The bug fix has been tested for almost 3 days non-stop while the following patches have been tested for a couple of hours with iperf and nuttcp and NFS root. Note that we also have a slight throughput improvement after patch 3 when running with AM437x SK (10.0.1.13 below): ===== pre-patch ===== $ iperf -c 10.0.1.13 ------------------------------------------------------------ Client connecting to 10.0.1.13, TCP port 5001 TCP window size: 85.0 KByte (default) ------------------------------------------------------------ [ 3] local 10.0.1.2 port 38430 connected with 10.0.1.13 port 5001 [ ID] Interval Transfer Bandwidth [ 3] 0.0-10.0 sec 240 MBytes 201 Mbits/sec $ iperf -c 10.0.1.101 ------------------------------------------------------------ Client connecting to 10.0.1.101, TCP port 5001 TCP window size: 85.0 KByte (default) ------------------------------------------------------------ [ 3] local 10.0.1.2 port 35143 connected with 10.0.1.101 port 5001 [ ID] Interval Transfer Bandwidth [ 3] 0.0-10.0 sec 113 MBytes 94.7 Mbits/sec ===== post-patch ===== $ iperf -c 10.0.1.13 ------------------------------------------------------------ Client connecting to 10.0.1.13, TCP port 5001 TCP window size: 85.0 KByte (default) ------------------------------------------------------------ [ 3] local 10.0.1.2 port 38447 connected with 10.0.1.13 port 5001 [ ID] Interval Transfer Bandwidth [ 3] 0.0-10.0 sec 283 MBytes 237 Mbits/sec $ iperf -c 10.0.1.101 ------------------------------------------------------------ Client connecting to 10.0.1.101, TCP port 5001 TCP window size: 85.0 KByte (default) ------------------------------------------------------------ [ 3] local 10.0.1.2 port 35157 connected with 10.0.1.101 port 5001 [ ID] Interval Transfer Bandwidth [ 3] 0.0-10.0 sec 113 MBytes 94.4 Mbits/sec That minor decrease on beagleboneblack (10.0.1.101) is so small that's likely into the error margin, but I'll run some further profiling as I have a feeling INTC's IRQ handling can be improved. In any case, patch 1 should go in during the -rc an get backported all the way back to v3.9, while the other patches can (should) be delayed for v3.20 merge window. Felipe Balbi (4): net: ethernet: cpsw: fix hangs with interrupts net: ethernet: cpsw: unroll IRQ request loop net: ethernet: cpsw: split out IRQ handler net: ethernet: cpsw: don't requests IRQs we don't use drivers/net/ethernet/ti/cpsw.c | 81 +++++++++++++++++++++++++++--------------- 1 file changed, 52 insertions(+), 29 deletions(-) -- 2.2.0 -- To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
