Message ID | 1422284164-16867-2-git-send-email-maxime.ripard@free-electrons.com |
---|---|
State | Superseded |
Headers | show |
Hi Brian, On Mon, Jan 26, 2015 at 03:56:03PM +0100, Maxime Ripard wrote: > The NDDB register holds the data that are needed by the read and write > commands. > > However, during a read PIO access, the datasheet specifies that after each 32 > bits read in that register, when BCH is enabled, we have to make sure that the > RDDREQ bit is set in the NDSR register. > > This fixes an issue that was seen on the Armada 385, and presumably other mvebu > SoCs, when a read on a newly erased page would end up in the driver reporting a > timeout from the NAND. > > Cc: <stable@vger.kernel.org> # v3.14 > Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com> Any chance for this fix to come in 3.19? Thanks, Maxime
Hi Maxime, On Mon, 26 Jan 2015 15:56:03 +0100 Maxime Ripard <maxime.ripard@free-electrons.com> wrote: > The NDDB register holds the data that are needed by the read and write > commands. > > However, during a read PIO access, the datasheet specifies that after each 32 > bits read in that register, when BCH is enabled, we have to make sure that the > RDDREQ bit is set in the NDSR register. > > This fixes an issue that was seen on the Armada 385, and presumably other mvebu > SoCs, when a read on a newly erased page would end up in the driver reporting a > timeout from the NAND. > > Cc: <stable@vger.kernel.org> # v3.14 > Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com> > --- > drivers/mtd/nand/pxa3xx_nand.c | 45 ++++++++++++++++++++++++++++++++++++------ > 1 file changed, 39 insertions(+), 6 deletions(-) > > diff --git a/drivers/mtd/nand/pxa3xx_nand.c b/drivers/mtd/nand/pxa3xx_nand.c > index 96b0b1d27df1..e6918befb951 100644 > --- a/drivers/mtd/nand/pxa3xx_nand.c > +++ b/drivers/mtd/nand/pxa3xx_nand.c > @@ -23,6 +23,7 @@ > #include <linux/mtd/partitions.h> > #include <linux/io.h> > #include <linux/irq.h> > +#include <linux/jiffies.h> > #include <linux/slab.h> > #include <linux/of.h> > #include <linux/of_device.h> > @@ -480,6 +481,38 @@ static void disable_int(struct pxa3xx_nand_info *info, uint32_t int_mask) > nand_writel(info, NDCR, ndcr | int_mask); > } > > +static void drain_fifo(struct pxa3xx_nand_info *info, void *data, int len) > +{ > + u32 *dst = (u32 *)data; > + > + if (info->ecc_bch) { > + while (len--) { > + u32 timeout; > + > + *dst++ = nand_readl(info, NDDB); > + > + /* > + * According to the datasheet, when reading > + * from NDDB with BCH enabled, after each 32 > + * bits reads, we have to make sure that the > + * NDSR.RDDREQ bit is set > + */ I know the datasheet says this bit should be checked after each transfer, but I wonder if we shouldn't check it before reading the data. What happens if you drain all the data available in the FIFO ? Is the controller still setting the RDDREQ bit ? Moreover, the datasheet says this RDDREQ bit should be checked after each 32 bytes (not 32 bits) transfer. Testing it after each readl call shouldn't hurt though. Best Regards, Boris
+ Rob This patch has conflicts with an ARM64-preparation from Rob. I'd like to get this patch in first, as it's a bugfix. But I'd like to settle Boris's comments first. (Regarding the request to get this into 3.19: not likely. Judging by the age of the "bug", it's not massively critical, and we have no time. It can get out through -stable once it's gotten proper review and testing.) On Wed, Feb 04, 2015 at 11:10:28AM +0100, Boris Brezillon wrote: > On Mon, 26 Jan 2015 15:56:03 +0100 > Maxime Ripard <maxime.ripard@free-electrons.com> wrote: > > > The NDDB register holds the data that are needed by the read and write > > commands. > > > > However, during a read PIO access, the datasheet specifies that after each 32 > > bits read in that register, when BCH is enabled, we have to make sure that the > > RDDREQ bit is set in the NDSR register. > > > > This fixes an issue that was seen on the Armada 385, and presumably other mvebu > > SoCs, when a read on a newly erased page would end up in the driver reporting a > > timeout from the NAND. > > > > Cc: <stable@vger.kernel.org> # v3.14 > > Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com> > > --- > > drivers/mtd/nand/pxa3xx_nand.c | 45 ++++++++++++++++++++++++++++++++++++------ > > 1 file changed, 39 insertions(+), 6 deletions(-) > > > > diff --git a/drivers/mtd/nand/pxa3xx_nand.c b/drivers/mtd/nand/pxa3xx_nand.c > > index 96b0b1d27df1..e6918befb951 100644 > > --- a/drivers/mtd/nand/pxa3xx_nand.c > > +++ b/drivers/mtd/nand/pxa3xx_nand.c > > @@ -23,6 +23,7 @@ > > #include <linux/mtd/partitions.h> > > #include <linux/io.h> > > #include <linux/irq.h> > > +#include <linux/jiffies.h> > > #include <linux/slab.h> > > #include <linux/of.h> > > #include <linux/of_device.h> > > @@ -480,6 +481,38 @@ static void disable_int(struct pxa3xx_nand_info *info, uint32_t int_mask) > > nand_writel(info, NDCR, ndcr | int_mask); > > } > > > > +static void drain_fifo(struct pxa3xx_nand_info *info, void *data, int len) > > +{ > > + u32 *dst = (u32 *)data; > > + > > + if (info->ecc_bch) { > > + while (len--) { > > + u32 timeout; > > + > > + *dst++ = nand_readl(info, NDDB); > > + > > + /* > > + * According to the datasheet, when reading > > + * from NDDB with BCH enabled, after each 32 > > + * bits reads, we have to make sure that the > > + * NDSR.RDDREQ bit is set > > + */ > > I know the datasheet says this bit should be checked after each > transfer, but I wonder if we shouldn't check it before reading the data. > What happens if you drain all the data available in the FIFO ? Is the > controller still setting the RDDREQ bit ? > > Moreover, the datasheet says this RDDREQ bit should be checked after > each 32 bytes (not 32 bits) transfer. > Testing it after each readl call shouldn't hurt though. Seems like that could quite possibly kill performance unnecessarily, couldn't it? But then, PIO is probably not that fast in the first place... Brian
Hi Brian, On Thu, 5 Feb 2015 17:08:35 -0800 Brian Norris <computersforpeace@gmail.com> wrote: > + Rob > > This patch has conflicts with an ARM64-preparation from Rob. I'd like to > get this patch in first, as it's a bugfix. But I'd like to settle > Boris's comments first. > > (Regarding the request to get this into 3.19: not likely. Judging by the > age of the "bug", it's not massively critical, and we have no time. It > can get out through -stable once it's gotten proper review and testing.) > > On Wed, Feb 04, 2015 at 11:10:28AM +0100, Boris Brezillon wrote: > > On Mon, 26 Jan 2015 15:56:03 +0100 > > Maxime Ripard <maxime.ripard@free-electrons.com> wrote: > > > > > The NDDB register holds the data that are needed by the read and write > > > commands. > > > > > > However, during a read PIO access, the datasheet specifies that after each 32 > > > bits read in that register, when BCH is enabled, we have to make sure that the > > > RDDREQ bit is set in the NDSR register. > > > > > > This fixes an issue that was seen on the Armada 385, and presumably other mvebu > > > SoCs, when a read on a newly erased page would end up in the driver reporting a > > > timeout from the NAND. > > > > > > Cc: <stable@vger.kernel.org> # v3.14 > > > Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com> > > > --- > > > drivers/mtd/nand/pxa3xx_nand.c | 45 ++++++++++++++++++++++++++++++++++++------ > > > 1 file changed, 39 insertions(+), 6 deletions(-) > > > > > > diff --git a/drivers/mtd/nand/pxa3xx_nand.c b/drivers/mtd/nand/pxa3xx_nand.c > > > index 96b0b1d27df1..e6918befb951 100644 > > > --- a/drivers/mtd/nand/pxa3xx_nand.c > > > +++ b/drivers/mtd/nand/pxa3xx_nand.c > > > @@ -23,6 +23,7 @@ > > > #include <linux/mtd/partitions.h> > > > #include <linux/io.h> > > > #include <linux/irq.h> > > > +#include <linux/jiffies.h> > > > #include <linux/slab.h> > > > #include <linux/of.h> > > > #include <linux/of_device.h> > > > @@ -480,6 +481,38 @@ static void disable_int(struct pxa3xx_nand_info *info, uint32_t int_mask) > > > nand_writel(info, NDCR, ndcr | int_mask); > > > } > > > > > > +static void drain_fifo(struct pxa3xx_nand_info *info, void *data, int len) > > > +{ > > > + u32 *dst = (u32 *)data; > > > + > > > + if (info->ecc_bch) { > > > + while (len--) { > > > + u32 timeout; > > > + > > > + *dst++ = nand_readl(info, NDDB); > > > + > > > + /* > > > + * According to the datasheet, when reading > > > + * from NDDB with BCH enabled, after each 32 > > > + * bits reads, we have to make sure that the > > > + * NDSR.RDDREQ bit is set > > > + */ > > > > I know the datasheet says this bit should be checked after each > > transfer, but I wonder if we shouldn't check it before reading the data. > > What happens if you drain all the data available in the FIFO ? Is the > > controller still setting the RDDREQ bit ? > > > > Moreover, the datasheet says this RDDREQ bit should be checked after > > each 32 bytes (not 32 bits) transfer. > > Testing it after each readl call shouldn't hurt though. > > Seems like that could quite possibly kill performance unnecessarily, > couldn't it? But then, PIO is probably not that fast in the first > place... Absolutety, my point was, it shouldn't hurt from a functional POV, but yes it will definitely impact performances. But that's not the first thing I would rework of if you're concerned about performances: when doing PIO read/write, the page read/write operations (I mean the part reading the internal fifo) are all done in interrupt context (called from pxa3xx_nand_irq), and doing this will prevent any other interrupt from taking place while you are draining/filling the FIFO :-(. An alternative would be to move this part into the read/write_buf functions, but that's a lot of work... Best Regards, Boris
On Fri, Feb 06, 2015 at 09:13:07AM +0100, Boris Brezillon wrote: > Hi Brian, > > On Thu, 5 Feb 2015 17:08:35 -0800 > Brian Norris <computersforpeace@gmail.com> wrote: > > On Wed, Feb 04, 2015 at 11:10:28AM +0100, Boris Brezillon wrote: > > > On Mon, 26 Jan 2015 15:56:03 +0100 > > > Maxime Ripard <maxime.ripard@free-electrons.com> wrote: > > > > + /* > > > > + * According to the datasheet, when reading > > > > + * from NDDB with BCH enabled, after each 32 > > > > + * bits reads, we have to make sure that the > > > > + * NDSR.RDDREQ bit is set > > > > + */ > > > > > > I know the datasheet says this bit should be checked after each > > > transfer, but I wonder if we shouldn't check it before reading the data. > > > What happens if you drain all the data available in the FIFO ? Is the > > > controller still setting the RDDREQ bit ? > > > > > > Moreover, the datasheet says this RDDREQ bit should be checked after > > > each 32 bytes (not 32 bits) transfer. > > > Testing it after each readl call shouldn't hurt though. > > > > Seems like that could quite possibly kill performance unnecessarily, > > couldn't it? But then, PIO is probably not that fast in the first > > place... > > Absolutety, my point was, it shouldn't hurt from a functional POV, but > yes it will definitely impact performances. OK. > But that's not the first thing I would rework of if you're concerned > about performances: when doing PIO read/write, the page read/write > operations (I mean the part reading the internal fifo) are all done in > interrupt context (called from pxa3xx_nand_irq), and doing this will > prevent any other interrupt from taking place while you are > draining/filling the FIFO :-(. ...which reminds me; the jiffies-based timeout in this patch isn't going to work in interrupt context. So it needs to be replaced either with a tight udelay() loop, or it needs to be moved out of the ISR. > An alternative would be to move this part into the read/write_buf > functions, but that's a lot of work... Yeah, that probably would be preferable, but I suppose it's not urgent either. Brian
On 02/06/2015 05:13 AM, Boris Brezillon wrote: > Hi Brian, > > On Thu, 5 Feb 2015 17:08:35 -0800 > Brian Norris <computersforpeace@gmail.com> wrote: > >> + Rob >> >> This patch has conflicts with an ARM64-preparation from Rob. I'd like to >> get this patch in first, as it's a bugfix. But I'd like to settle >> Boris's comments first. >> >> (Regarding the request to get this into 3.19: not likely. Judging by the >> age of the "bug", it's not massively critical, and we have no time. It >> can get out through -stable once it's gotten proper review and testing.) >> >> On Wed, Feb 04, 2015 at 11:10:28AM +0100, Boris Brezillon wrote: >>> On Mon, 26 Jan 2015 15:56:03 +0100 >>> Maxime Ripard <maxime.ripard@free-electrons.com> wrote: >>> >>>> The NDDB register holds the data that are needed by the read and write >>>> commands. >>>> >>>> However, during a read PIO access, the datasheet specifies that after each 32 >>>> bits read in that register, when BCH is enabled, we have to make sure that the >>>> RDDREQ bit is set in the NDSR register. >>>> >>>> This fixes an issue that was seen on the Armada 385, and presumably other mvebu >>>> SoCs, when a read on a newly erased page would end up in the driver reporting a >>>> timeout from the NAND. >>>> >>>> Cc: <stable@vger.kernel.org> # v3.14 >>>> Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com> >>>> --- >>>> drivers/mtd/nand/pxa3xx_nand.c | 45 ++++++++++++++++++++++++++++++++++++------ >>>> 1 file changed, 39 insertions(+), 6 deletions(-) >>>> >>>> diff --git a/drivers/mtd/nand/pxa3xx_nand.c b/drivers/mtd/nand/pxa3xx_nand.c >>>> index 96b0b1d27df1..e6918befb951 100644 >>>> --- a/drivers/mtd/nand/pxa3xx_nand.c >>>> +++ b/drivers/mtd/nand/pxa3xx_nand.c >>>> @@ -23,6 +23,7 @@ >>>> #include <linux/mtd/partitions.h> >>>> #include <linux/io.h> >>>> #include <linux/irq.h> >>>> +#include <linux/jiffies.h> >>>> #include <linux/slab.h> >>>> #include <linux/of.h> >>>> #include <linux/of_device.h> >>>> @@ -480,6 +481,38 @@ static void disable_int(struct pxa3xx_nand_info *info, uint32_t int_mask) >>>> nand_writel(info, NDCR, ndcr | int_mask); >>>> } >>>> >>>> +static void drain_fifo(struct pxa3xx_nand_info *info, void *data, int len) >>>> +{ >>>> + u32 *dst = (u32 *)data; >>>> + >>>> + if (info->ecc_bch) { >>>> + while (len--) { >>>> + u32 timeout; >>>> + >>>> + *dst++ = nand_readl(info, NDDB); >>>> + >>>> + /* >>>> + * According to the datasheet, when reading >>>> + * from NDDB with BCH enabled, after each 32 >>>> + * bits reads, we have to make sure that the >>>> + * NDSR.RDDREQ bit is set >>>> + */ >>> >>> I know the datasheet says this bit should be checked after each >>> transfer, but I wonder if we shouldn't check it before reading the data. >>> What happens if you drain all the data available in the FIFO ? Is the >>> controller still setting the RDDREQ bit ? >>> >>> Moreover, the datasheet says this RDDREQ bit should be checked after >>> each 32 bytes (not 32 bits) transfer. >>> Testing it after each readl call shouldn't hurt though. >> >> Seems like that could quite possibly kill performance unnecessarily, >> couldn't it? But then, PIO is probably not that fast in the first >> place... > > Absolutety, my point was, it shouldn't hurt from a functional POV, but > yes it will definitely impact performances. > But that's not the first thing I would rework of if you're concerned > about performances: when doing PIO read/write, the page read/write > operations (I mean the part reading the internal fifo) are all done in > interrupt context (called from pxa3xx_nand_irq), and doing this will > prevent any other interrupt from taking place while you are > draining/filling the FIFO :-(. But NAND operations are serialized, and there won't be any other interrupt for the controller until it's has drained the FIFO. So this doesn't really seem to hit performance to me. Or am I missing anything here? > An alternative would be to move this part into the read/write_buf > functions, but that's a lot of work... > Yeah, indeed. This also has other benefits. As we discussed on IRC, it would allow to support raw writes (i.e. ECC off).
On Fri, Feb 06, 2015 at 11:17:15AM -0300, Ezequiel Garcia wrote: > On 02/06/2015 05:13 AM, Boris Brezillon wrote: > > On Thu, 5 Feb 2015 17:08:35 -0800 > > Brian Norris <computersforpeace@gmail.com> wrote: > >> On Wed, Feb 04, 2015 at 11:10:28AM +0100, Boris Brezillon wrote: > >>> I know the datasheet says this bit should be checked after each > >>> transfer, but I wonder if we shouldn't check it before reading the data. > >>> What happens if you drain all the data available in the FIFO ? Is the > >>> controller still setting the RDDREQ bit ? > >>> > >>> Moreover, the datasheet says this RDDREQ bit should be checked after > >>> each 32 bytes (not 32 bits) transfer. > >>> Testing it after each readl call shouldn't hurt though. > >> > >> Seems like that could quite possibly kill performance unnecessarily, > >> couldn't it? But then, PIO is probably not that fast in the first > >> place... > > > > Absolutety, my point was, it shouldn't hurt from a functional POV, but > > yes it will definitely impact performances. > > But that's not the first thing I would rework of if you're concerned > > about performances: when doing PIO read/write, the page read/write > > operations (I mean the part reading the internal fifo) are all done in > > interrupt context (called from pxa3xx_nand_irq), and doing this will > > prevent any other interrupt from taking place while you are > > draining/filling the FIFO :-(. > > But NAND operations are serialized, and there won't be any other > interrupt for the controller until it's has drained the FIFO. So this > doesn't really seem to hit performance to me. > > Or am I missing anything here? I think we're talking about two things: 1. The $subject patch is probably adding too many extra register reads (8 times too many?). In the grand scheme of things, this probably isn't significant. 2. The driver as already written is doing too much in its interrupt handler; this is bad practice in general and can hurt the responsiveness of the system. And particularly here, we also have the problem of a potential lockup, since the new timeout loop is waiting on jiffies, which will not be updated. > > An alternative would be to move this part into the read/write_buf > > functions, but that's a lot of work... > > > > Yeah, indeed. This also has other benefits. As we discussed on IRC, it > would allow to support raw writes (i.e. ECC off). OK. This is probably a good work item for later. But let's get back on point. I'd like to see version 2 with the following: 1. For sure the jiffies loop needs to be rewritten 2. Optionally, the RDDREQ bit should be checked less often, per the datasheet and Boris's comments Maxime, would that be OK? Then maybe I can get this into 3.20. Maybe not in my primary -rc1 pullreq, if we haven't had enough time to look at and test v2, but ASAP. Only after than am I likely to take Rob's ARM64 cleanup. Brian
On Fri, Feb 06, 2015 at 11:38:59AM -0800, Brian Norris wrote: > > > An alternative would be to move this part into the read/write_buf > > > functions, but that's a lot of work... > > > > > > > Yeah, indeed. This also has other benefits. As we discussed on IRC, it > > would allow to support raw writes (i.e. ECC off). > > OK. This is probably a good work item for later. But let's get back on > point. > > I'd like to see version 2 with the following: > > 1. For sure the jiffies loop needs to be rewritten > > 2. Optionally, the RDDREQ bit should be checked less often, per the > datasheet and Boris's comments > > Maxime, would that be OK? Yep. I actually worked on these two issues just today. It looks like I blew my boards' bootloader in the process, so it might take so time. > Then maybe I can get this into 3.20. Maybe not in my primary -rc1 > pullreq, if we haven't had enough time to look at and test v2, but ASAP. > Only after than am I likely to take Rob's ARM64 cleanup. Sounds good, thanks! Maxime
diff --git a/drivers/mtd/nand/pxa3xx_nand.c b/drivers/mtd/nand/pxa3xx_nand.c index 96b0b1d27df1..e6918befb951 100644 --- a/drivers/mtd/nand/pxa3xx_nand.c +++ b/drivers/mtd/nand/pxa3xx_nand.c @@ -23,6 +23,7 @@ #include <linux/mtd/partitions.h> #include <linux/io.h> #include <linux/irq.h> +#include <linux/jiffies.h> #include <linux/slab.h> #include <linux/of.h> #include <linux/of_device.h> @@ -480,6 +481,38 @@ static void disable_int(struct pxa3xx_nand_info *info, uint32_t int_mask) nand_writel(info, NDCR, ndcr | int_mask); } +static void drain_fifo(struct pxa3xx_nand_info *info, void *data, int len) +{ + u32 *dst = (u32 *)data; + + if (info->ecc_bch) { + while (len--) { + u32 timeout; + + *dst++ = nand_readl(info, NDDB); + + /* + * According to the datasheet, when reading + * from NDDB with BCH enabled, after each 32 + * bits reads, we have to make sure that the + * NDSR.RDDREQ bit is set + */ + timeout = jiffies + msecs_to_jiffies(5); + while (!(nand_readl(info, NDSR) & NDSR_RDDREQ)) { + if (!time_before(jiffies, timeout)) { + dev_err(&info->pdev->dev, + "Timeout on RDDREQ while draining the FIFO\n"); + return; + } + + cpu_relax(); + } + } + } else { + __raw_readsl(info->mmio_base + NDDB, data, len); + } +} + static void handle_data_pio(struct pxa3xx_nand_info *info) { unsigned int do_bytes = min(info->data_size, info->chunk_size); @@ -496,14 +529,14 @@ static void handle_data_pio(struct pxa3xx_nand_info *info) DIV_ROUND_UP(info->oob_size, 4)); break; case STATE_PIO_READING: - __raw_readsl(info->mmio_base + NDDB, - info->data_buff + info->data_buff_pos, - DIV_ROUND_UP(do_bytes, 4)); + drain_fifo(info, + info->data_buff + info->data_buff_pos, + DIV_ROUND_UP(do_bytes, 4)); if (info->oob_size > 0) - __raw_readsl(info->mmio_base + NDDB, - info->oob_buff + info->oob_buff_pos, - DIV_ROUND_UP(info->oob_size, 4)); + drain_fifo(info, + info->oob_buff + info->oob_buff_pos, + DIV_ROUND_UP(info->oob_size, 4)); break; default: dev_err(&info->pdev->dev, "%s: invalid state %d\n", __func__,
The NDDB register holds the data that are needed by the read and write commands. However, during a read PIO access, the datasheet specifies that after each 32 bits read in that register, when BCH is enabled, we have to make sure that the RDDREQ bit is set in the NDSR register. This fixes an issue that was seen on the Armada 385, and presumably other mvebu SoCs, when a read on a newly erased page would end up in the driver reporting a timeout from the NAND. Cc: <stable@vger.kernel.org> # v3.14 Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com> --- drivers/mtd/nand/pxa3xx_nand.c | 45 ++++++++++++++++++++++++++++++++++++------ 1 file changed, 39 insertions(+), 6 deletions(-)