The reason why it fails is because we're no longer supporting unsegmented packet pools. I could change the patch to fall back to unsegmented support if the requested buf_size > ODP_CONFIG_PACKET_BUF_LEN_MAX. I'll post a v2 that does that so we don't unnecessarily break tests in 0.6.0.
It would be helpful to have Petri's views on this. The real solution, I believe, is to allow the caller of odp_buffer_pool_create() to specify whether they need unsegmented buffers, as we had originally. On Thu, Dec 18, 2014 at 9:35 AM, Taras Kondratiuk < [email protected]> wrote: > > On 12/18/2014 04:31 PM, Bill Fischofer wrote: > >> It's one patch broken into two parts [1/2] [2/2]. Is this not correct? >> I'm trying to break things up since folks don't want to see single >> patches. >> > > Split looks correct. First patch prepares pktio code for segmented > buffers, and the second patch enables them. I'd keep them as a separate > patches. > > Just found one issue with the patch. Crypto test tries to allocate a > buffer pool with 32k buffers. Allocation fails because it is more than a > limit ODP_CONFIG_PACKET_BUF_LEN_MAX. >
_______________________________________________ lng-odp mailing list [email protected] http://lists.linaro.org/mailman/listinfo/lng-odp
