On 09-Jan-20 1:27 PM, Olivier Matz wrote:
To populate a mempool with a virtual area, the mempool code calls
rte_mempool_populate_iova() for each iova-contiguous area. It happens
(rarely) that this area is too small to store one object. In this case,
rte_mempool_populate_iova() returns an error, which is forwarded by
rte_mempool_populate_virt().
This case should not throw an error in
rte_mempool_populate_virt(). Instead, the area that is too small should
just be ignored.
To fix this issue, change the return value of
rte_mempool_populate_iova() to -ENOBUFS when no object can be populated,
so it can be ignored by the caller. As this would be an API change, add
a compat wrapper to keep the current API unchanged. The wrapper will be
removed for 20.11.
Fixes: 354788b60cfd ("mempool: allow populating with unaligned virtual area")
Cc: sta...@dpdk.org
Signed-off-by: Olivier Matz <olivier.m...@6wind.com>
---
The approach fixes the issue on my end, so
Tested-by: Anatoly Burakov <anatoly.bura...@intel.com>
Is there a simple way to ensure that we won't forget to remove the
wrapper for 20.11? Anatoly suggested me to use versioned symbols, but
it's not clear to me how.
Yes, i'd like to do better than "ah shur we won't forget pinky swear".
Can't we do this with ABI versioning? E.g.
rte_populate_iova_v20() ... returns EINVAL
rte_populate_iova_v21() ... returns ENOBUFS
I'm pretty sure, even if it doesn't break, it will still be more likely
to not be forgotten because there's almost a guarantee that someone will
grep for symbol versioning macros across the codebase around 20.11
timeframe.
--
Thanks,
Anatoly
