Messages in this thread | | | Date | Mon, 17 Sep 2018 18:16:23 +0300 | From | Alexander Fomichev <> | Subject | Re: [PATCH v2 0/8] Fix breakage caused by the NTB multi-port patchset |
| |
On Fri, Jul 20, 2018 at 12:00:26PM -0600, Logan Gunthorpe wrote:
> To reiterate the main points on the feedback: the switchtec driver, > in the cross link mode, will not be able to implement the port number > function callbacks and will have to always return 0. It's a physical > impossibility due to the symmetry. Therefore, in order fix this feature > (which worked when merged), the clients must be changed to support not > knowing the port number in the way they worked previously (ie. a legacy > mode where there will only be two ports and you know there will be two > sets of doorbells, one for each peer). The majority of this patch set > fixes these issues. > > The other point of controversy is the dma mask. I still strongly > disagree with doing it in the driver as the code is clearly common > to all and not at all driver specific. Moreover, I think as written, > it is extra dangerous seeing all impleminting drivers are operating on > the new struct device before it's initialized in ntb_register(). > Patch 2 in this series also fixes that. > > [...] > > Logan Gunthorpe (8): > NTB: ntb_tool: reading the link file should not end in a NULL byte > NTB: Setup the DMA mask globally for all drivers > NTB: Fix the default port and peer numbers for legacy drivers > NTB: ntb_pingpong: Choose doorbells based on port number > NTB: perf: Don't require one more memory window than number of peers > NTB: perf: Fix support for hardware that doesn't have port numbers > NTB: perf: Fix race condition when run with ntb_test > NTB: ntb_test: Fix bug when counting remote files
Tested successfully on i5-based desktop PC, NTB on PM8535 chip, both ends are on the local system.
Tested-by: Alexander Fomichev <fomichev.ru@gmail.com>
-- Regards, Alexander
| |