From mboxrd@z Thu Jan 1 00:00:00 1970 From: Geert Uytterhoeven Date: Wed, 15 Apr 2020 12:36:29 +0000 Subject: Re: [PATCH] serial: sh-sci: Make sure status register SCxSR is read in correct sequence Message-Id: List-Id: References: <1585333048-31828-1-git-send-email-kazuhiro.fujita.jg@renesas.com> In-Reply-To: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: "Lad, Prabhakar" Cc: Prabhakar Mahadev Lad , Kazuhiro Fujita , Greg Kroah-Hartman , Jiri Slaby , "open list:SERIAL DRIVERS" , Linux-Renesas , Linux Kernel Mailing List , Hao Bui , KAZUMI HARADA , Sasha Levin , Chris Brandt , Magnus Damm , Linux-sh list , Rob Landley , John Paul Adrian Glaubitz On Thu, Apr 2, 2020 at 5:24 PM Geert Uytterhoeven wrote: > On Thu, Apr 2, 2020 at 1:28 PM Lad, Prabhakar > wrote: > > On Wed, Apr 1, 2020 at 1:43 PM Geert Uytterhoeven wrote: > > > On Tue, Mar 31, 2020 at 5:58 PM Prabhakar Mahadev Lad > > > wrote: > > > > > -----Original Message----- > > > > > From: Geert Uytterhoeven > > > > > On Fri, Mar 27, 2020 at 7:17 PM Kazuhiro Fujita > > > > > wrote: > > > > > > For SCIF and HSCIF interfaces the SCxSR register holds the status of > > > > > > data that is to be read next from SCxRDR register, But where as for > > > > > > SCIFA and SCIFB interfaces SCxSR register holds status of data that is > > > > > > previously read from SCxRDR register. > > > > > > > > > > > > This patch makes sure the status register is read depending on the port > > > > > > types so that errors are caught accordingly. > > > > > > > > > > > > Cc: > > > > > > Signed-off-by: Kazuhiro Fujita > > > > > > Signed-off-by: Hao Bui > > > > > > Signed-off-by: KAZUMI HARADA > > > > > > Signed-off-by: Lad Prabhakar > > > > > > > > Nevertheless, this patch will need some testing on various hardware. > > > > > Do you have a test case to verify the broken/fixed behavior? > > > > > > > > > Agreed, its been tested on RZ/G2x & RZ/G1x by doing a loopback test, configure one interface as CS8 mode(8-bits data, No parity) and other as CS7 mode (7-bits data, 1-bit Parity) and parity errors should be detected. > > > > > > This can easily be tested on the console. Basic testing can even be > > > done with an unmodified kernel, as there is already a "parity error" > > > notice message in the driver. > > > > > > Enable even parity on the console: > > > > > > $ stty evenp > > > > > > (use "oddp" for odd parity, and invert all below) > > > > > > Typing e.g. a single "p" should trigger a parity error. > > > Typing "o" shouldn't. > > > Without this patch, no parity error is detected on SCIF. > > > > > > Likewise, pasting a sequence of "p" characters should trigger a lot of > > > parity errors, "o" shouldn't. > > > Without this patch, parity errors are detected on SCIF, except for the > > > first character. > > > > > > For more advanced testing, make the following change to the driver: > > > > > > - dev_notice(port->dev, "parity error\n"); > > > + dev_notice(port->dev, "parity error for char 0x%02x hweight %u\n", > > > c, hweight8(c)); > > > > > > Pasting an alternating sequence of "p" and "o" characters should trigger > > > parity errors for the "p" characters. > > > Without this patch, they are triggered for the "o" characters instead. > > > > > Thank you that makes life easier. > > > > > With this patch, the issues above are fixed on SCIF. > > > This has been verified on: > > > 1. SCIF on R-Car Gen 2, > > > 2. SCIF on R-Car Gen3 > > > 3. SCIF on RZ/A1H, > > > 4. SCIF on RZ/A2M. > If I disable DMA for HSCIF1 in the .dtsi, parity errors are detected > as expected. So HSCIF on R-Car Gen3 is affected, and fixed by this patch. > Hence the driver does not support parity checking when DMA is enabled. > I also think it's not easy to add support for that, if possible at all. > > > > I haven't tested it yet on: > > > 1. SCIFB on SH/R-Mobile (needs wiring up), SCIFB on R-Mobile A1 is not affected, and works before/after, as expected. > > > 2. SCIFA, SCIFB, and HSCIF on R-Car Gen2 (needs wiring up), Same for SCIFA, SCIFB on R-Car M2-W. HSCIF on R-Car M2-W is affected, and fixed by this patch. This means the modern platforms we care about are handled fine, so Tested-by: Geert Uytterhoeven That leaves us with testing on a few legacy platforms... > > > 3. (H)SCIF on R-Car Gen1 (remote boards unaccessible at the moment), > > > 4. SuperH (only remote Migo-R available, but unaccessible). > > > > > > I can test 1 and 2 (and perhaps 3 and 4) later, if needed. > > probably testing this on SuperH is gonna be a pain due to lack of > > hardware availability, > > (it needs to be tested on 19 platforms) > > how about #ifdef CONFIG_ARCH_RENESAS || CONFIG_H8300 and the fix ? > > I had a look at a few SuperH docs w.r.t. framing/parity error behavior: > - SCIF(A) on SH7724: similar to R-Car Gen2, > - H(SCIF) on SH7734: same as on R-Car Gen2, > - SCIF on SH7751: conflict between status register ("to be read next") > and flowchart ("read from"). > > Let's wait a bit, we're in the middle of the merge window anyway. > Probably we can get it tested on SuperH during the coming weeks. Anyone with a real (not qemu) SuperH system who can do the basic "stty evenp" tests above, and report back to us? Thanks a lot! Gr{oetje,eeting}s, Geert -- Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that. -- Linus Torvalds