Douglas Pollock <douglas.pollock@xxxxxxxx> :
[...]
> Let me know if you would like me to test something in particular.
Yes.
I assume it is a bit premature to try to run r8169 with HT SMP, acpi and
nvidia binary module at the same time.
So I would suggest to build a kernel with both SMP and acpi disabled.
Then boot in single user mode or anything similar which avoids the nvidia
module to be loaded at all before the test*. I'd welcome two things:
- behavior with a simple 'ping -c 144 remote_host'
No load, it just walks twice the Rx/Tx rings.
- behavior under load. Use the tool of your choice. No need to reboot
after the aforementioned test if the system is still sane.
If you can spend some time on it, a report between each patch would be nice:
1) r8169-tx-index-overflow.patch
2) r8169-tx-index-overflow.patch
+ r8169-dma-api-tx.patch
3) r8169-tx-index-overflow.patch
+ r8169-dma-api-tx.patch
+ r8169-dma-api-rx-buffers.patch
etc, until:
- the 16 rounds of patches/tests are exhausted ;
- you are bored :o)
The 4 most recent patches (-getstats/-addr-high/-ethtool/-static) can wait.
[*] Why ?
1) there is a ~200x ratio in size from r8169 to nvidia module
r8169 9728 0
nvidia 1701612 8
2) because.
--
Ueimor
|