On Thu, Feb 9, 2017 at 4:27 PM, Mark Greer <mgr...@animalcreek.com> wrote: > > On Thu, Feb 09, 2017 at 10:54:08AM -0500, Geoff Lansberry wrote: > > On Wed, Feb 8, 2017 at 5:56 PM, Mark Greer <mgr...@animalcreek.com> wrote: > > > > > > I just realized that the linux-nfc is not CC'd so adding it. > > > > > > On Wed, Feb 08, 2017 at 03:53:39PM -0700, Mark Greer wrote: > > > > On Sun, Dec 18, 2016 at 08:07:33PM -0700, Mark Greer wrote: > > > > > On Sat, Dec 17, 2016 at 04:19:04PM -0500, Geoff Lansberry wrote: > > > > > > Mark, from our consultant: > > > > > > > > > > > > It isn't important whether the flood script is successful in writing > > > > > > or not. The point of it is to force a segfault by making many > > > > > > requests. It needs to run for several hundred iterations (successful > > > > > > or not) in order to generate the segfault. > > > > > > > > > > So neard crashes even when the write fails?? Okay, I'll let it run > > > > > for > > > > > a while tomorrow (Monday). > > > > > > > > [Okay, so not exactly "tomorrow" but I did get back to this.] > > > > > > > > Geoff, a few things: > > > > > > > > 1) Any update on these issues? > > > > Yes - I've discovered the primary trigger of my problems. I'm using > > some very small tags from Murata, and they don't have a strong signal. > > Sometimes when writing, the tags can get corrupted data written to > > them. When the tag is verified by re-reading it, neard bails out > > because it finds non-utf8 characters. So far I have found it > > impossible to rewrite a tag once it has been mal-formed. > > Okay, that's interesting. neard still shouldn't bail out so we'll > have to look at that.
This is the most critical thing for us right now. Can you make a malformed tag and try it to see if you have the same issue? What I typically see is 0xFE character too early in the string, or similarly the length being too long for the string. If you can do that, then try to write or read it via neard. If you can confirm the quitting behavior, that will help a lot. I can live with the corruption, because I can detect it and re-write if it happens. > > > > > 2) Do you have all of the NFC-related patches from the nfc-next master > > > > branch? In particular, do you have all of Thierry's patches to > > > > net/nfc/digial_*.c dated around June-July 2016? Without those patches, > > > > I see a panic; with them, I don't. > > > > No, we don't. the last patch I've got in our kernel version is yours > > from 7-21-2014. I'm inquiring with more experienced people to see how > > to address that. > > Oh... Then you really need to update. Look at the master branch in the > nfc-next repo > (git://git.kernel.org/pub/scm/linux/kernel/git/sameo/nfc-next.git) Those commits are in the latest ti build, https://git.ti.com/ti-linux-kernel/ti-linux-kernel/commits/ti-linux-4.9.y, so I've built that up. Unfortunately bricked the board and don't have the dev tools to fix it with me. Bummer, should be back into it tomorrow. > > Which kernel version are you using? If its old, you should update to the > most recent version that you can. If you *absolutely* can't use a newer > kernel, let me know as I may be able to help. > > Also, be sure to use the latest version of neard (currently 0.16). > > > > > 3) Assuming you have all of those patches, please revert the one with > > > > the > > > > summary line of, "NFC: digital: Set the command pending flag", and tell > > > > me > > > > if that stops the "Bogus state" messages. I don't know which > > > > repo/branch > > > > you're using so I can't provide a commit id. > > > > No to the patches, but I found the commit you are talking about and > > will look at it. Maybe possible to cherry pick it. > > No, don't cherry-pick it. I was asking you to revert that commit. > Ignore this request for now because the commit is fine, it just seemed > to cover up the issue so I wanted to see if it did for you too. I will > dig into it more once I finish setting up a new SDCard for my test system > (and get some time to look at it). > > Mark > --
