On Mon, Jul 22, 2013 at 2:38 PM, Bjorn Helgaas <bhelgaas@xxxxxxxxxx> wrote: > [+cc Alex, Yinghai, linux-pci] > > Isn't this for the same issue Alex and others have been working on? yes. > > It doesn't feel like we have consensus on how this should be fixed. > You're proposing a change to work_on_cpu(), Alex proposed a change to > pci_call_probe() [1], Yinghai proposed some changes to the PCI core > SR-IOV code and several drivers [2]. > > [1] https://lkml.kernel.org/r/20130624195942.40795.27292.stgit@xxxxxxxxxxxxxxxxxxxxxxxx > [2] https://lkml.kernel.org/r/1368498506-25857-7-git-send-email-yinghai@xxxxxxxxxx Actually i used Alex's patch and attached -v2 of sriov_attatch in my local testing for a while. -v2: remove the device_schedule_callback hacking. Also, someone comment that numa_node_id() in alex's should not be used, as it could be preempted. Thanks Yinghai
Attachment:
fix_racing_removing_6_6.patch
Description: Binary data
