- Aug 24, 2015
-
-
When the switch is loaded, some PTP outgoing frames get discarded, so we never get back the timestamp interrupt. By round-robin'ing the slots, we get back to the "busy" one, which holds an skb structure that must be released. Without this freeing, the skb remains in the output queue of the socket, and after a while we get EAGAIN in send() because the buffer is full (400 frames, more or less). And the ptp port is stuck until we reopen the socket. This diagnosis exposed the fact that it's not an overflow of pending descriptors, so I change the message from descriptor overflow: tx timestamp pending to discarding tx frame that got no timestamp Finally, this is pr_warn, not pr_err as we now know it's expected. Signed-off-by:Alessandro Rubini <rubini@gnudd.com>
-
Signed-off-by:
-
- Jul 17, 2015
-
-
Alessandro Rubini authored
Signed-off-by:
-
- Jul 09, 2015
-
-
- May 04, 2015
-
-
Adam Wujek authored
Add reporting via ifconfig: --rx errors (counter RX PCS Errors from pstats) --carrier (counter RX Sync Lost from pstats) Signed-off-by:Adam Wujek <adam.wujek@cern.ch>
-
- Dec 11, 2014
-
-
Alessandro Rubini authored
Otherwise, the port files are empty, even if pstats/info is correct wrs-192.168.16.9#wc -c /proc/sys/pstats/* 529 /proc/sys/pstats/description 7 /proc/sys/pstats/info 0 /proc/sys/pstats/port0 0 /proc/sys/pstats/port1 0 /proc/sys/pstats/port10 0 /proc/sys/pstats/port11 0 /proc/sys/pstats/port12 0 /proc/sys/pstats/port13 0 /proc/sys/pstats/port14 0 /proc/sys/pstats/port15 0 /proc/sys/pstats/port16 0 /proc/sys/pstats/port17 0 /proc/sys/pstats/port2 0 /proc/sys/pstats/port3 0 /proc/sys/pstats/port4 0 /proc/sys/pstats/port5 0 /proc/sys/pstats/port6 0 /proc/sys/pstats/port7 0 /proc/sys/pstats/port8 0 /proc/sys/pstats/port9 536 total Signed-off-by:
-
- Dec 10, 2014
-
-
Adam Wujek authored
get rid of constants: PSTATS_CNT_PP and PSTATS_ADR_PP Use number of ports read from FPGA instead of PSTATS_NUM_OF_COUNTERS Signed-off-by: -
Adam Wujek authored
Read is done at module init. Signed-off-by: -
Adam Wujek authored
Masking is not needed since uint32_t << 32 zero all lower bits. Reported-by:
-
Adam Wujek authored
Signed-off-by: -
Adam Wujek authored
Signed-off-by: -
Adam Wujek authored
Changes also in wr_pstats. Wr_nic now calls callback function in wr_pstats to get counters Signed-off-by: -
By writing anything to a port file, you zero the counters for that port. By writing to "info" or "description" you zero counters for all ports. Zeroing is performed by copying the current values to a new array, and always subytracting that array to the current counter values. No zeroing is performed for network statistics, that still reads the "real" counters, featuring an overflow at 4G but no zeroing ever. To simplify a little the code, I turned the one-dimensional array with index arithmetics into a two-dimensional array, so the compiler makes the math for us, transparently. NOTE: the spinlock must be audited, we are not safe yet, not only because of this patch. Signed-off-by: -
Signed-off-by: -
Adam Wujek authored
Signed-off-by: -
Signed-off-by: -
-
- Dec 05, 2014
-
-
Alessandro Rubini authored
Signed-off-by: -
Alessandro Rubini authored
Signed-off-by:
-
- Dec 04, 2014
-
-
Signed-off-by: -
Signed-off-by: -
Add posibility to setup MAC address at load time for SPEC. Signed-off-by: -
In kernel this function returns 1/true on success! Signed-off-by: -
- Remove only for write in #72557df - NetworkManger (Unbuntu LTS 12.04) insists on accessing to the phy when it is used from LM32 and flood the dmesg so we remove the warning.
-
NetworkManager tries to access the phy, so this WARN_ON is pretty common if the daemon is running. Now we know who is misbehaving, and the driver can keep silent (disclaimer: this opinionated message is by Alessandro :)
-
Alessandro Rubini authored
Signed-off-by: -
Signed-off-by: -
Signed-off-by: -
commit inspired by commit from spec-sw repo: kernel/wr_nic: fix rx timestamps c246c3ce36ac908ee3aada7db1f293bd12e42346 (commit edited for checkpatch by AR) Signed-off-by: -
The calibration procedures are defined but never called. They were used in wrs-V2 hardware, but callers disappeared in January 2012. Better remove them than let them rust forever. Signed-off-by: -
patch taken from spec-sw repo: kernel/wr_nic: major cleanup of tx timestamping ab3f2f8d18e4f6ed744438314e5a78b8a509de74
-
(commit amended by alessandro to avoid #ifdef in the functions) Signed-off-by: -
This change adds warnings, but they will disappear after move to 3.14, where %pa is supported. Signed-off-by: -
Alessandro Rubini authored
This fixes an oops when programs access ethtool methods, especially wrn_get_drvinfo. It can be reproduced by a simple run of "irqbalance", which I never did in my development machine. Reported-by:
Grzegorz Daniluk <grzegorz.daniluk@gmail.com> Signed-off-by:
-
Alessandro Rubini authored
Signed-off-by: -
-
-
Alessandro Rubini authored
Signed-off-by: -
Alessandro Rubini authored
Signed-off-by: -
Alessandro Rubini authored
Signed-off-by:
-
