Remove pci vendor check, check only pci class #621
Conversation
pkg/fpga/utils.go
Outdated
|
|
||
| if len(os.Getenv("VID")) == 0 && pci.Vendor != vendorIntel { | ||
| return errors.Errorf("unsupported PCI device %s VID=%s PID=%s Class=%s", pci.BDF, pci.Vendor, pci.Device, pci.Class) | ||
| } else if pci.Vendor != os.Getenv("VID") && pci.Vendor != vendorIntel { |
There was a problem hiding this comment.
@rmr-silicom are you looking to import pkg/fpga in your project?
There was a problem hiding this comment.
Yes. I am adding support for the FPGA n5010 card. It uses the kernel drivers from here https://github.com/OPAE/linux-dfl-backport, but it seems the sysfs entries are mostly the same.
There was a problem hiding this comment.
@rmr-silicom can you share output of lspci for this card, to see how it is reported in terms IDs/class/etc?
Also, it would be great to see how kernel DFL drivers sysfs reports about that device.
There was a problem hiding this comment.
@mythi It has a vendor id of 0x1c2c and product ids of 0x1001, 0x1000, and the 0x12000 fpga class, of course
There was a problem hiding this comment.
Sorry, didn't catch your sysfs question... I added a few, there are a lot more...
[core@worker1 ~]$ find /sys/class/fpga_*
/sys/class/fpga_bridge
/sys/class/fpga_bridge/br0
/sys/class/fpga_manager
/sys/class/fpga_manager/fpga0
/sys/class/fpga_region
/sys/class/fpga_region/region0
/sys/class/fpga_region/region1
/sys/class/fpga_sec_mgr
/sys/class/fpga_sec_mgr/fpga_sec0
FME.0
/sys/class/fpga_region/region0/dfl-fme.0/uevent
/sys/class/fpga_region/region0/dfl-fme.0/bitstream_id
/sys/class/fpga_region/region0/dfl-fme.0/bitstream_metadata
PORT.0
/sys/class/fpga_region/region0/dfl-port.0
/sys/class/fpga_region/region0/dfl-port.0/uevent
/sys/class/fpga_region/region0/dfl-port.0/afu_id
/sys/class/fpga_region/region0/dfl-port.0/power_state
/sys/class/fpga_region/region0/dfl-port.0/ap2_event
/sys/class/fpga_region/region0/dfl-port.0/ltr
There was a problem hiding this comment.
removing vendor check by overall, but make sure that DFL APIs are compatible still.
@rmr-silicom can you update this PR to drop the vendor check completely and rename the function.
@kad would the existing reg expressions suffice or did you have something else in your mind?
There was a problem hiding this comment.
@rmr-silicom you can squash commits and update PR title. beside that, lgtm.
@mythi the current patch is ok from me for now. However just existing regexp for device node names might not be enough. I'd like later to implement validation of kernel driver behind the PCI device.
There was a problem hiding this comment.
@kad isn't the merge squash thing on your end? If I squash the commits, the branch changes and this request is lost.
There was a problem hiding this comment.
@rmr-silicom we don't squash commits during merge but ask the developers to do that. the ask is you do git rebase -i HEAD~2, combine the two commits in our branch to one and force push to remote
There was a problem hiding this comment.
Thanks for the tip!
rmr-silicom
changed the title
rmr-silicom
force-pushed
the
runtime-vid
branch
from
84a9e1c
to
5a37914
Compare
rmr-silicom
force-pushed
the
runtime-vid
branch
from
5a37914
to
25ef460
Compare
No description provided.