[EXT] Re: [RFC] qla2xxx: Add dev_loss_tmo kernel module options

Daniel Wagner dwagner at suse.de
Tue Apr 27 10:51:31 BST 2021


On Wed, Apr 21, 2021 at 09:56:59AM +0200, Daniel Wagner wrote:
> Ah, didn't know about nvmediscovery until very recentetly. I try to get
> it working with this approach (as this patch is not really a proper
> solution).

Finally found some time to play with this. FTR, the nvmediscovery
carries following information:

  UDEV  [65238.364677] change   /devices/virtual/fc/fc_udev_device (fc)
  ACTION=change
  DEVPATH=/devices/virtual/fc/fc_udev_device
  FC_EVENT=nvmediscovery
  NVMEFC_HOST_TRADDR=nn-0x20000024ff7fa448:pn-0x21000024ff7fa448
  NVMEFC_TRADDR=nn-0x200200a09890f5bf:pn-0x203800a09890f5bf
  SEQNUM=12357
  SUBSYSTEM=fc
  USEC_INITIALIZED=65238333374

The udev rule I came up is:

  ACTION=="change", SUBSYSTEM=="fc", ENV{FC_EVENT}=="nvmediscovery", \
      ENV{NVMEFC_TRADDR}=="*", \
      RUN+="/usr/local/sbin/qla2xxx_dev_loss_tmo.sh $env{NVMEFC_TRADDR} 4294967295"

and the script is:

  #!/bin/sh

  TRADDR=$1
  TMO=$2

  id=$(echo $TRADDR | sed -n "s/.*pn-0x\([0-9a-f]\+\)/\1/p")
  find /sys/kernel/debug/qla2xxx -name pn-$id -exec /bin/sh -c "echo $TMO > {}/dev_loss_tmo" \;

I am sure this can be done in a more elegant way. Anyway, I am testing
this right now, the first 30 minutes look good...



More information about the Linux-nvme mailing list