Re: [PATCH v5 09/12] mtd: rawnand: ingenic: Make use of ecc-engine property
Hi Paul, Paul Cercueil wrote on Mon, 18 Mar 2019 21:37:09 +0100: > Hi, > > Le ven. 15 mars 2019 à 15:37, Paul Cercueil a écrit : > > Hi, > > > > Le ven. 15 mars 2019 à 9:40, Miquel Raynal > a > > écrit : > >> Hi Paul, > >> >> Paul Cercueil wrote on Wed, 13 Mar 2019 >> > >> >> 23:22:56 > >> +0100: > >> >>> Use the 'ecc-engine' standard property instead of the custom > >>> 'ingenic,bch-controller' custom property, which is now deprecated. > >>> >>> Signed-off-by: Paul Cercueil > >>> --- > >>> >>> Notes: > >>> v5: New patch > >>> >>> drivers/mtd/nand/raw/ingenic/ingenic_ecc.c | 13 ++--- > >>> 1 file changed, 10 insertions(+), 3 deletions(-) > >>> >>> diff --git a/drivers/mtd/nand/raw/ingenic/ingenic_ecc.c >>> > >>> >>> b/drivers/mtd/nand/raw/ingenic/ingenic_ecc.c > >>> index d7f3a8c3abea..30436ca6628a 100644 > >>> --- a/drivers/mtd/nand/raw/ingenic/ingenic_ecc.c > >>> +++ b/drivers/mtd/nand/raw/ingenic/ingenic_ecc.c > >>> @@ -82,9 +82,9 @@ static struct ingenic_ecc >>> *ingenic_ecc_get(struct > >>> device_node *np) > >>> >>> /** > >>>* of_ingenic_ecc_get() - get the ECC controller from a DT node > >>> - * @of_node: the node that contains a bch-controller property. > >>> + * @of_node: the node that contains a ecc-engine property. > >> >> Would "contains an ecc-engine property" be better English? > >> >> I am not sure what is the rule when it comes to plain English with > >> variable names. However if you agree, no need to re-send the series, >> I > >> can fix it when applying. > > > > Yes, that's better. > > > >> BTW, I added hw ECC engines support to my generic ECC engine > >> implementation, but migrating the whole raw NAND subsystem (using I/O > >> requests like in the SPI-NAND core, adding prepare/finish_io_req >> hooks) > >> is going to be much more invasive than initially expected, so I am >> not > >> sure I will finish the migration any time soon. > > > > Ok, I will follow the development then. > > > >> Thanks, > >> Miquèl > > > > One thing I notice with my patchset: it works perfectly on top of > 4.20, > > but on top of 5.0 I am unable to erase any eraseblock with > flash_erase. > > I get -EIO every time. I'm trying to debug it but didn't go very far, > > it looks like nand_status_op() gives me a status of 0xff. Do you know > > what could have changed between 4.20 and 5.0 that could trigger this > bug? > > > > Nevermind. It works now. Great! > > > Second thing, everytime I reboot it fails to find the BBT. That's > because > > the BBT marker is overwritten by the ECC data as they occupy the same > area > > in the OOB space. Is there a way to move the BBT marker? Or should I > use > > NAND_BBT_NO_OOB then? Since the eraseblocks where the BBTs are located > > is used in my system partition, won't that conflict with the data? > > Response to myself: It's possible to move the BBT marker. But in my case I > have to deal with three possible layouts, so it's simpler to just use > NAND_BBT_NO_OOB then. The BBT pages are marked so that they're not used > for data in the partitions. Maybe you should check Frieder's series: mtd: rawnand: Support bad block markers in first, second or last page Thanks, Miquèl
Re: [PATCH v5 09/12] mtd: rawnand: ingenic: Make use of ecc-engine property
Hi, Le ven. 15 mars 2019 à 15:37, Paul Cercueil a écrit : Hi, Le ven. 15 mars 2019 à 9:40, Miquel Raynal a écrit : Hi Paul, Paul Cercueil wrote on Wed, 13 Mar 2019 23:22:56 +0100: Use the 'ecc-engine' standard property instead of the custom 'ingenic,bch-controller' custom property, which is now deprecated. Signed-off-by: Paul Cercueil --- Notes: v5: New patch drivers/mtd/nand/raw/ingenic/ingenic_ecc.c | 13 ++--- 1 file changed, 10 insertions(+), 3 deletions(-) diff --git a/drivers/mtd/nand/raw/ingenic/ingenic_ecc.c b/drivers/mtd/nand/raw/ingenic/ingenic_ecc.c index d7f3a8c3abea..30436ca6628a 100644 --- a/drivers/mtd/nand/raw/ingenic/ingenic_ecc.c +++ b/drivers/mtd/nand/raw/ingenic/ingenic_ecc.c @@ -82,9 +82,9 @@ static struct ingenic_ecc *ingenic_ecc_get(struct device_node *np) /** * of_ingenic_ecc_get() - get the ECC controller from a DT node - * @of_node: the node that contains a bch-controller property. + * @of_node: the node that contains a ecc-engine property. Would "contains an ecc-engine property" be better English? I am not sure what is the rule when it comes to plain English with variable names. However if you agree, no need to re-send the series, I can fix it when applying. Yes, that's better. BTW, I added hw ECC engines support to my generic ECC engine implementation, but migrating the whole raw NAND subsystem (using I/O requests like in the SPI-NAND core, adding prepare/finish_io_req hooks) is going to be much more invasive than initially expected, so I am not sure I will finish the migration any time soon. Ok, I will follow the development then. Thanks, Miquèl One thing I notice with my patchset: it works perfectly on top of 4.20, but on top of 5.0 I am unable to erase any eraseblock with flash_erase. I get -EIO every time. I'm trying to debug it but didn't go very far, it looks like nand_status_op() gives me a status of 0xff. Do you know what could have changed between 4.20 and 5.0 that could trigger this bug? Nevermind. It works now. Second thing, everytime I reboot it fails to find the BBT. That's because the BBT marker is overwritten by the ECC data as they occupy the same area in the OOB space. Is there a way to move the BBT marker? Or should I use NAND_BBT_NO_OOB then? Since the eraseblocks where the BBTs are located is used in my system partition, won't that conflict with the data? Response to myself: It's possible to move the BBT marker. But in my case I have to deal with three possible layouts, so it's simpler to just use NAND_BBT_NO_OOB then. The BBT pages are marked so that they're not used for data in the partitions. I'll send a V6 then. Thanks, -Paul
Re: [PATCH v5 09/12] mtd: rawnand: ingenic: Make use of ecc-engine property
Hi, Le ven. 15 mars 2019 à 9:40, Miquel Raynal a écrit : Hi Paul, Paul Cercueil wrote on Wed, 13 Mar 2019 23:22:56 +0100: Use the 'ecc-engine' standard property instead of the custom 'ingenic,bch-controller' custom property, which is now deprecated. Signed-off-by: Paul Cercueil --- Notes: v5: New patch drivers/mtd/nand/raw/ingenic/ingenic_ecc.c | 13 ++--- 1 file changed, 10 insertions(+), 3 deletions(-) diff --git a/drivers/mtd/nand/raw/ingenic/ingenic_ecc.c b/drivers/mtd/nand/raw/ingenic/ingenic_ecc.c index d7f3a8c3abea..30436ca6628a 100644 --- a/drivers/mtd/nand/raw/ingenic/ingenic_ecc.c +++ b/drivers/mtd/nand/raw/ingenic/ingenic_ecc.c @@ -82,9 +82,9 @@ static struct ingenic_ecc *ingenic_ecc_get(struct device_node *np) /** * of_ingenic_ecc_get() - get the ECC controller from a DT node - * @of_node: the node that contains a bch-controller property. + * @of_node: the node that contains a ecc-engine property. Would "contains an ecc-engine property" be better English? I am not sure what is the rule when it comes to plain English with variable names. However if you agree, no need to re-send the series, I can fix it when applying. Yes, that's better. BTW, I added hw ECC engines support to my generic ECC engine implementation, but migrating the whole raw NAND subsystem (using I/O requests like in the SPI-NAND core, adding prepare/finish_io_req hooks) is going to be much more invasive than initially expected, so I am not sure I will finish the migration any time soon. Ok, I will follow the development then. Thanks, Miquèl One thing I notice with my patchset: it works perfectly on top of 4.20, but on top of 5.0 I am unable to erase any eraseblock with flash_erase. I get -EIO every time. I'm trying to debug it but didn't go very far, it looks like nand_status_op() gives me a status of 0xff. Do you know what could have changed between 4.20 and 5.0 that could trigger this bug? Second thing, everytime I reboot it fails to find the BBT. That's because the BBT marker is overwritten by the ECC data as they occupy the same area in the OOB space. Is there a way to move the BBT marker? Or should I use NAND_BBT_NO_OOB then? Since the eraseblocks where the BBTs are located is used in my system partition, won't that conflict with the data? Thanks, -Paul
Re: [PATCH v5 09/12] mtd: rawnand: ingenic: Make use of ecc-engine property
Hi Paul, Paul Cercueil wrote on Wed, 13 Mar 2019 23:22:56 +0100: > Use the 'ecc-engine' standard property instead of the custom > 'ingenic,bch-controller' custom property, which is now deprecated. > > Signed-off-by: Paul Cercueil > --- > > Notes: > v5: New patch > > drivers/mtd/nand/raw/ingenic/ingenic_ecc.c | 13 ++--- > 1 file changed, 10 insertions(+), 3 deletions(-) > > diff --git a/drivers/mtd/nand/raw/ingenic/ingenic_ecc.c > b/drivers/mtd/nand/raw/ingenic/ingenic_ecc.c > index d7f3a8c3abea..30436ca6628a 100644 > --- a/drivers/mtd/nand/raw/ingenic/ingenic_ecc.c > +++ b/drivers/mtd/nand/raw/ingenic/ingenic_ecc.c > @@ -82,9 +82,9 @@ static struct ingenic_ecc *ingenic_ecc_get(struct > device_node *np) > > /** > * of_ingenic_ecc_get() - get the ECC controller from a DT node > - * @of_node: the node that contains a bch-controller property. > + * @of_node: the node that contains a ecc-engine property. Would "contains an ecc-engine property" be better English? I am not sure what is the rule when it comes to plain English with variable names. However if you agree, no need to re-send the series, I can fix it when applying. BTW, I added hw ECC engines support to my generic ECC engine implementation, but migrating the whole raw NAND subsystem (using I/O requests like in the SPI-NAND core, adding prepare/finish_io_req hooks) is going to be much more invasive than initially expected, so I am not sure I will finish the migration any time soon. Thanks, Miquèl
[PATCH v5 09/12] mtd: rawnand: ingenic: Make use of ecc-engine property
Use the 'ecc-engine' standard property instead of the custom 'ingenic,bch-controller' custom property, which is now deprecated. Signed-off-by: Paul Cercueil --- Notes: v5: New patch drivers/mtd/nand/raw/ingenic/ingenic_ecc.c | 13 ++--- 1 file changed, 10 insertions(+), 3 deletions(-) diff --git a/drivers/mtd/nand/raw/ingenic/ingenic_ecc.c b/drivers/mtd/nand/raw/ingenic/ingenic_ecc.c index d7f3a8c3abea..30436ca6628a 100644 --- a/drivers/mtd/nand/raw/ingenic/ingenic_ecc.c +++ b/drivers/mtd/nand/raw/ingenic/ingenic_ecc.c @@ -82,9 +82,9 @@ static struct ingenic_ecc *ingenic_ecc_get(struct device_node *np) /** * of_ingenic_ecc_get() - get the ECC controller from a DT node - * @of_node: the node that contains a bch-controller property. + * @of_node: the node that contains a ecc-engine property. * - * Get the bch-controller property from the given device tree + * Get the ecc-engine property from the given device tree * node and pass it to ingenic_ecc_get to do the work. * * Return: a pointer to ingenic_ecc, errors are encoded into the pointer. @@ -95,7 +95,14 @@ struct ingenic_ecc *of_ingenic_ecc_get(struct device_node *of_node) struct ingenic_ecc *ecc = NULL; struct device_node *np; - np = of_parse_phandle(of_node, "ingenic,bch-controller", 0); + np = of_parse_phandle(of_node, "ecc-engine", 0); + + /* +* If the ecc-engine property is not found, check for the deprecated +* ingenic,bch-controller property +*/ + if (!np) + np = of_parse_phandle(of_node, "ingenic,bch-controller", 0); if (np) { ecc = ingenic_ecc_get(np); -- 2.11.0