| Message ID | 20170525081747.4c86164c@bbrezillon |
|---|---|
| State | New |
| Headers | show |
On 25/05/17 18:19, Boris Brezillon wrote: > Le Wed, 24 May 2017 22:58:53 +0000, > Chris Packham <Chris.Packham@alliedtelesis.co.nz> a écrit : > >> On 25/05/17 10:36, Boris Brezillon wrote: >>> Le Wed, 24 May 2017 22:03:52 +0000, >>> Chris Packham <Chris.Packham@alliedtelesis.co.nz> a écrit : >>> >>>> On 24/05/17 23:25, Boris Brezillon wrote: >>>>> On Wed, 24 May 2017 13:23:01 +0200 >>>>> Boris Brezillon <boris.brezillon@free-electrons.com> wrote: >>>>> >>>>>> Hi Chris, >>>>>> >>>>>> On Wed, 24 May 2017 09:36:56 +0000 >>>>>> Chris Packham <Chris.Packham@alliedtelesis.co.nz> wrote: >>>>>> >>>>>>> On 23/05/17 17:27, Chris Packham wrote: >>>>>>>> Hi, >>>>>>>> >>>>>>>> I'm doing some testing on linux-next and I'm finding that my nand flash >>>>>>>> has disappeared. >>>>>>>> >>>>>>>> pxa3xx-nand f10d0000.flash: This platform can't do DMA on this device >>>>>>>> pxa3xx-nand f10d0000.flash: non-supported command ef >>>>>>>> pxa3xx-nand f10d0000.flash: non-supported command ee >>>>>>>> pxa3xx-nand f10d0000.flash: non-supported command ef >>>>>>>> pxa3xx-nand f10d0000.flash: non-supported command ee >>>>>>>> On-die ECC forcefully enabled, not supported >>>>>>>> nand: No NAND device found >>>>>>>> pxa3xx-nand f10d0000.flash: failed to scan nand at cs 0 >>>>>>>> >>>>>>>> This was working around 4.11. I'll try to do some more digging tomorrow >>>>>>>> to narrow down a failure point but I thought I'd send this out now just >>>>>>>> in case it rings any bells. >>>>>>>> >>>>>>>> The board I'm using (DB-88F6820-AMC) is unfortunately out-of tree but it >>>>>>>> should be pretty close to the armada-388-db. I can make my dts available >>>>>>>> if it's helpful. >>>>>>> >>>>>>> Still works on 4.12-rc2. Fails on next-20170524. >>>>>>> >>>>>>> This appears to be due to commit b566d9c055de ("mtd: nand: add support >>>>>>> for Micron on-die ECC"). Which based on the description seems intentional. >>>>>>> >>>>>>> Since I have access to a hardware platform that has a micron flash with >>>>>>> ECC forcefully enabled how can I help to get this implemented. >>>>>> >>>>>> Can you try with this patch applied [1]? >>>>> >>>>> Sorry, wrong patch. Can you try this one [1] instead? >>>>> >>>>> [1]http://code.bulix.org/pkfhmi-135875 >>>>> >>>> >>>> With the patch above the chip is detected but ubifs is unhappy >>> >>> Hm, weird. And if you revert my both Thomas patch and mine, what do you >>> get? >> >> Seems to work just fine. > > Okay. Let's try with something simpler then. Can you test the following > patch? > > --->8--- > From 0a0eaa39ea9b19191ae0c31ff3625c224a8a55f2 Mon Sep 17 00:00:00 2001 > From: Boris Brezillon <boris.brezillon@free-electrons.com> > Date: Thu, 25 May 2017 08:15:15 +0200 > Subject: [PATCH] mtd: nand: pxa3xx: Implement failing > ->onfi_{set,get}_features() stubs > > Implement stubs returning -ENOTSUPP for ->onfi_{set,get}_features() so > that we don't let the core think we are supporting the GET/SET FEATURES > operation. > > Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com> > --- Yep that seems to work.
Le Thu, 25 May 2017 22:12:12 +0000, Chris Packham <Chris.Packham@alliedtelesis.co.nz> a écrit : > On 25/05/17 18:19, Boris Brezillon wrote: > > Le Wed, 24 May 2017 22:58:53 +0000, > > Chris Packham <Chris.Packham@alliedtelesis.co.nz> a écrit : > > > >> On 25/05/17 10:36, Boris Brezillon wrote: > >>> Le Wed, 24 May 2017 22:03:52 +0000, > >>> Chris Packham <Chris.Packham@alliedtelesis.co.nz> a écrit : > >>> > >>>> On 24/05/17 23:25, Boris Brezillon wrote: > >>>>> On Wed, 24 May 2017 13:23:01 +0200 > >>>>> Boris Brezillon <boris.brezillon@free-electrons.com> wrote: > >>>>> > >>>>>> Hi Chris, > >>>>>> > >>>>>> On Wed, 24 May 2017 09:36:56 +0000 > >>>>>> Chris Packham <Chris.Packham@alliedtelesis.co.nz> wrote: > >>>>>> > >>>>>>> On 23/05/17 17:27, Chris Packham wrote: > >>>>>>>> Hi, > >>>>>>>> > >>>>>>>> I'm doing some testing on linux-next and I'm finding that my nand flash > >>>>>>>> has disappeared. > >>>>>>>> > >>>>>>>> pxa3xx-nand f10d0000.flash: This platform can't do DMA on this device > >>>>>>>> pxa3xx-nand f10d0000.flash: non-supported command ef > >>>>>>>> pxa3xx-nand f10d0000.flash: non-supported command ee > >>>>>>>> pxa3xx-nand f10d0000.flash: non-supported command ef > >>>>>>>> pxa3xx-nand f10d0000.flash: non-supported command ee > >>>>>>>> On-die ECC forcefully enabled, not supported > >>>>>>>> nand: No NAND device found > >>>>>>>> pxa3xx-nand f10d0000.flash: failed to scan nand at cs 0 > >>>>>>>> > >>>>>>>> This was working around 4.11. I'll try to do some more digging tomorrow > >>>>>>>> to narrow down a failure point but I thought I'd send this out now just > >>>>>>>> in case it rings any bells. > >>>>>>>> > >>>>>>>> The board I'm using (DB-88F6820-AMC) is unfortunately out-of tree but it > >>>>>>>> should be pretty close to the armada-388-db. I can make my dts available > >>>>>>>> if it's helpful. > >>>>>>> > >>>>>>> Still works on 4.12-rc2. Fails on next-20170524. > >>>>>>> > >>>>>>> This appears to be due to commit b566d9c055de ("mtd: nand: add support > >>>>>>> for Micron on-die ECC"). Which based on the description seems intentional. > >>>>>>> > >>>>>>> Since I have access to a hardware platform that has a micron flash with > >>>>>>> ECC forcefully enabled how can I help to get this implemented. > >>>>>> > >>>>>> Can you try with this patch applied [1]? > >>>>> > >>>>> Sorry, wrong patch. Can you try this one [1] instead? > >>>>> > >>>>> [1]http://code.bulix.org/pkfhmi-135875 > >>>>> > >>>> > >>>> With the patch above the chip is detected but ubifs is unhappy > >>> > >>> Hm, weird. And if you revert my both Thomas patch and mine, what do you > >>> get? > >> > >> Seems to work just fine. > > > > Okay. Let's try with something simpler then. Can you test the following > > patch? > > > > --->8--- > > From 0a0eaa39ea9b19191ae0c31ff3625c224a8a55f2 Mon Sep 17 00:00:00 2001 > > From: Boris Brezillon <boris.brezillon@free-electrons.com> > > Date: Thu, 25 May 2017 08:15:15 +0200 > > Subject: [PATCH] mtd: nand: pxa3xx: Implement failing > > ->onfi_{set,get}_features() stubs > > > > Implement stubs returning -ENOTSUPP for ->onfi_{set,get}_features() so > > that we don't let the core think we are supporting the GET/SET FEATURES > > operation. > > > > Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com> > > --- > > Yep that seems to work. Arrggg!! I hate this situation. Now I have to check all drivers implementing their own ->cmdfunc() to make sure they support NAND_CMD_SET/GET_FEATURES, and if they don't, do what I did for pxa3xx. Anyway, thanks for testing. I'll try to re-arrange my nand/next branch to avoid breaking bisectibility (put this patch before the on-die ECC one). Regards, Boris
diff --git a/drivers/mtd/nand/pxa3xx_nand.c b/drivers/mtd/nand/pxa3xx_nand.c index 649ba8200832..341a229046f5 100644 --- a/drivers/mtd/nand/pxa3xx_nand.c +++ b/drivers/mtd/nand/pxa3xx_nand.c @@ -1649,6 +1649,13 @@ static int pxa_ecc_init(struct pxa3xx_nand_info *info, return 0; } +static int pxa3xx_nand_get_set_features(struct mtd_info *mtd, + struct nand_chip *chip, + int feature_addr, u8 *subfeature_para) +{ + return -ENOTSUPP; +} + static int pxa3xx_nand_scan(struct mtd_info *mtd) { struct nand_chip *chip = mtd_to_nand(mtd); @@ -1812,6 +1819,8 @@ static int alloc_nand_resource(struct platform_device *pdev) chip->write_buf = pxa3xx_nand_write_buf; chip->options |= NAND_NO_SUBPAGE_WRITE; chip->cmdfunc = nand_cmdfunc; + chip->onfi_set_features = pxa3xx_nand_get_set_features; + chip->onfi_get_features = pxa3xx_nand_get_set_features; } nand_hw_control_init(chip->controller);