On 05-May-20 11:59 AM, Thomas Monjalon wrote:
05/05/2020 12:28, Burakov, Anatoly:
On 03-May-20 5:26 PM, Renata Saiakhova wrote:
Delete memzones for HW rings in igb and ixgbe while freeing queues
Signed-off-by: Renata Saiakhova <renata.saiakh...@ekinops.com>
---
+Thomas
Should this perhaps be fixed in all drivers, not just ixgbe/igb? Is this
safe to do in multiprocess? I'm not too well versed in ethdev mechanics
when it comes to multiprocess, presumably the application itself is
responsible for synchronizing access to ports, so freeing the resources
should be OK?
The application is responsible of port policy.
If the application decides to close a port,
it must be safe in all threads and processes,
meaning it is application responsibility to not refer to port resources.
About fixing in all drivers, is it something missing in other drivers?
I can see several other drivers using the dma_reserve API, so presumably
they would suffer from the same issue, unless they use this API for a
different purpose.
--
Thanks,
Anatoly
