| Message ID | d4ef0bd6-816e-f6fa-9385-f05f775f0ae2@sberdevices.ru (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [RFC,v1] mtd: rawnand: meson: invalidate cache on polling ECC bit | expand |
Hello, we reproduced this problem on one of our boards. It triggers very rare when 'usleep_range()' is present, but when sleeping is removed - it fires always. I suppose problem is with caching, as 'info_buf' memory is mapped by 'dma_map_single()'. Thanks, Arseniy On 13.03.2023 10:32, Arseniy Krasnov wrote: > 'info_buf' memory is cached and driver polls ECC bit in it. This bit > is set by the NAND controller. If 'usleep_range()' returns before device > sets this bit, 'info_buf' will be cached and driver won't see update of > this bit and will loop forever. > > Fixes: 8fae856c5350 ("mtd: rawnand: meson: add support for Amlogic NAND flash controller") > Signed-off-by: Arseniy Krasnov <AVKrasnov@sberdevices.ru> > --- > drivers/mtd/nand/raw/meson_nand.c | 8 +++++++- > 1 file changed, 7 insertions(+), 1 deletion(-) > > diff --git a/drivers/mtd/nand/raw/meson_nand.c b/drivers/mtd/nand/raw/meson_nand.c > index 5ee01231ac4c..2c05c08a0eaf 100644 > --- a/drivers/mtd/nand/raw/meson_nand.c > +++ b/drivers/mtd/nand/raw/meson_nand.c > @@ -176,6 +176,7 @@ struct meson_nfc { > > dma_addr_t daddr; > dma_addr_t iaddr; > + u32 info_bytes; > > unsigned long assigned_cs; > }; > @@ -503,6 +504,7 @@ static int meson_nfc_dma_buffer_setup(struct nand_chip *nand, void *databuf, > nfc->daddr, datalen, dir); > return ret; > } > + nfc->info_bytes = infolen; > cmd = GENCMDIADDRL(NFC_CMD_AIL, nfc->iaddr); > writel(cmd, nfc->reg_base + NFC_REG_CMD); > > @@ -520,8 +522,10 @@ static void meson_nfc_dma_buffer_release(struct nand_chip *nand, > struct meson_nfc *nfc = nand_get_controller_data(nand); > > dma_unmap_single(nfc->dev, nfc->daddr, datalen, dir); > - if (infolen) > + if (infolen) { > dma_unmap_single(nfc->dev, nfc->iaddr, infolen, dir); > + nfc->info_bytes = 0; > + } > } > > static int meson_nfc_read_buf(struct nand_chip *nand, u8 *buf, int len) > @@ -710,6 +714,8 @@ static void meson_nfc_check_ecc_pages_valid(struct meson_nfc *nfc, > usleep_range(10, 15); > /* info is updated by nfc dma engine*/ > smp_rmb(); > + dma_sync_single_for_cpu(nfc->dev, nfc->iaddr, nfc->info_bytes, > + DMA_FROM_DEVICE); > ret = *info & ECC_COMPLETE; > } while (!ret); > }
Hi Arseniy, avkrasnov@sberdevices.ru wrote on Mon, 13 Mar 2023 10:36:11 +0300: > Hello, > > we reproduced this problem on one of our boards. It triggers very rare > when 'usleep_range()' is present, but when sleeping is removed - it fires > always. I suppose problem is with caching, as 'info_buf' memory is mapped by > 'dma_map_single()'. The fix looks really legitimate, indeed I get that the usleep_range() might make it work most of the time but not always. Having this bit in a DMA buf area is a bit strange. Well, the fix LGTM anyway. > > Thanks, Arseniy > > On 13.03.2023 10:32, Arseniy Krasnov wrote: > > 'info_buf' memory is cached and driver polls ECC bit in it. This bit > > is set by the NAND controller. If 'usleep_range()' returns before device > > sets this bit, 'info_buf' will be cached and driver won't see update of > > this bit and will loop forever. > > > > Fixes: 8fae856c5350 ("mtd: rawnand: meson: add support for Amlogic NAND flash controller") > > Signed-off-by: Arseniy Krasnov <AVKrasnov@sberdevices.ru> > > --- > > drivers/mtd/nand/raw/meson_nand.c | 8 +++++++- > > 1 file changed, 7 insertions(+), 1 deletion(-) > > > > diff --git a/drivers/mtd/nand/raw/meson_nand.c b/drivers/mtd/nand/raw/meson_nand.c > > index 5ee01231ac4c..2c05c08a0eaf 100644 > > --- a/drivers/mtd/nand/raw/meson_nand.c > > +++ b/drivers/mtd/nand/raw/meson_nand.c > > @@ -176,6 +176,7 @@ struct meson_nfc { > > > > dma_addr_t daddr; > > dma_addr_t iaddr; > > + u32 info_bytes; > > > > unsigned long assigned_cs; > > }; > > @@ -503,6 +504,7 @@ static int meson_nfc_dma_buffer_setup(struct nand_chip *nand, void *databuf, > > nfc->daddr, datalen, dir); > > return ret; > > } > > + nfc->info_bytes = infolen; > > cmd = GENCMDIADDRL(NFC_CMD_AIL, nfc->iaddr); > > writel(cmd, nfc->reg_base + NFC_REG_CMD); > > > > @@ -520,8 +522,10 @@ static void meson_nfc_dma_buffer_release(struct nand_chip *nand, > > struct meson_nfc *nfc = nand_get_controller_data(nand); > > > > dma_unmap_single(nfc->dev, nfc->daddr, datalen, dir); > > - if (infolen) > > + if (infolen) { > > dma_unmap_single(nfc->dev, nfc->iaddr, infolen, dir); > > + nfc->info_bytes = 0; > > + } > > } > > > > static int meson_nfc_read_buf(struct nand_chip *nand, u8 *buf, int len) > > @@ -710,6 +714,8 @@ static void meson_nfc_check_ecc_pages_valid(struct meson_nfc *nfc, > > usleep_range(10, 15); > > /* info is updated by nfc dma engine*/ > > smp_rmb(); > > + dma_sync_single_for_cpu(nfc->dev, nfc->iaddr, nfc->info_bytes, > > + DMA_FROM_DEVICE); > > ret = *info & ECC_COMPLETE; > > } while (!ret); > > } Thanks, Miquèl
On 13/03/2023 12:18, Miquel Raynal wrote: > Hi Arseniy, > > avkrasnov@sberdevices.ru wrote on Mon, 13 Mar 2023 10:36:11 +0300: > >> Hello, >> >> we reproduced this problem on one of our boards. It triggers very rare >> when 'usleep_range()' is present, but when sleeping is removed - it fires >> always. I suppose problem is with caching, as 'info_buf' memory is mapped by >> 'dma_map_single()'. > > The fix looks really legitimate, indeed I get that the usleep_range() > might make it work most of the time but not always. Having this bit in > a DMA buf area is a bit strange. Well, the fix LGTM anyway. Yep it looks legitimate! LGTM Reviewed-by: Neil Armstrong <neil.armstrong@linaro.org> > >> >> Thanks, Arseniy >> >> On 13.03.2023 10:32, Arseniy Krasnov wrote: >>> 'info_buf' memory is cached and driver polls ECC bit in it. This bit >>> is set by the NAND controller. If 'usleep_range()' returns before device >>> sets this bit, 'info_buf' will be cached and driver won't see update of >>> this bit and will loop forever. >>> >>> Fixes: 8fae856c5350 ("mtd: rawnand: meson: add support for Amlogic NAND flash controller") >>> Signed-off-by: Arseniy Krasnov <AVKrasnov@sberdevices.ru> >>> --- >>> drivers/mtd/nand/raw/meson_nand.c | 8 +++++++- >>> 1 file changed, 7 insertions(+), 1 deletion(-) >>> >>> diff --git a/drivers/mtd/nand/raw/meson_nand.c b/drivers/mtd/nand/raw/meson_nand.c >>> index 5ee01231ac4c..2c05c08a0eaf 100644 >>> --- a/drivers/mtd/nand/raw/meson_nand.c >>> +++ b/drivers/mtd/nand/raw/meson_nand.c >>> @@ -176,6 +176,7 @@ struct meson_nfc { >>> >>> dma_addr_t daddr; >>> dma_addr_t iaddr; >>> + u32 info_bytes; >>> >>> unsigned long assigned_cs; >>> }; >>> @@ -503,6 +504,7 @@ static int meson_nfc_dma_buffer_setup(struct nand_chip *nand, void *databuf, >>> nfc->daddr, datalen, dir); >>> return ret; >>> } >>> + nfc->info_bytes = infolen; >>> cmd = GENCMDIADDRL(NFC_CMD_AIL, nfc->iaddr); >>> writel(cmd, nfc->reg_base + NFC_REG_CMD); >>> >>> @@ -520,8 +522,10 @@ static void meson_nfc_dma_buffer_release(struct nand_chip *nand, >>> struct meson_nfc *nfc = nand_get_controller_data(nand); >>> >>> dma_unmap_single(nfc->dev, nfc->daddr, datalen, dir); >>> - if (infolen) >>> + if (infolen) { >>> dma_unmap_single(nfc->dev, nfc->iaddr, infolen, dir); >>> + nfc->info_bytes = 0; >>> + } >>> } >>> >>> static int meson_nfc_read_buf(struct nand_chip *nand, u8 *buf, int len) >>> @@ -710,6 +714,8 @@ static void meson_nfc_check_ecc_pages_valid(struct meson_nfc *nfc, >>> usleep_range(10, 15); >>> /* info is updated by nfc dma engine*/ >>> smp_rmb(); >>> + dma_sync_single_for_cpu(nfc->dev, nfc->iaddr, nfc->info_bytes, >>> + DMA_FROM_DEVICE); >>> ret = *info & ECC_COMPLETE; >>> } while (!ret); >>> } > > > Thanks, > Miquèl
Hello guys! Was it applied to some nand 'prepare-for-merge' release branch? On Mon, Mar 13, 2023 at 12:23:12PM +0100, Neil Armstrong wrote: > On 13/03/2023 12:18, Miquel Raynal wrote: > > Hi Arseniy, > > > > avkrasnov@sberdevices.ru wrote on Mon, 13 Mar 2023 10:36:11 +0300: > > > > > Hello, > > > > > > we reproduced this problem on one of our boards. It triggers very rare > > > when 'usleep_range()' is present, but when sleeping is removed - it fires > > > always. I suppose problem is with caching, as 'info_buf' memory is mapped by > > > 'dma_map_single()'. > > > > The fix looks really legitimate, indeed I get that the usleep_range() > > might make it work most of the time but not always. Having this bit in > > a DMA buf area is a bit strange. Well, the fix LGTM anyway. > > Yep it looks legitimate! > > LGTM > > > Reviewed-by: Neil Armstrong <neil.armstrong@linaro.org> > > > > > > > > > Thanks, Arseniy > > > > > > On 13.03.2023 10:32, Arseniy Krasnov wrote: > > > > 'info_buf' memory is cached and driver polls ECC bit in it. This bit > > > > is set by the NAND controller. If 'usleep_range()' returns before device > > > > sets this bit, 'info_buf' will be cached and driver won't see update of > > > > this bit and will loop forever. > > > > > > > > Fixes: 8fae856c5350 ("mtd: rawnand: meson: add support for Amlogic NAND flash controller") > > > > Signed-off-by: Arseniy Krasnov <AVKrasnov@sberdevices.ru> > > > > --- > > > > drivers/mtd/nand/raw/meson_nand.c | 8 +++++++- > > > > 1 file changed, 7 insertions(+), 1 deletion(-) > > > > > > > > diff --git a/drivers/mtd/nand/raw/meson_nand.c b/drivers/mtd/nand/raw/meson_nand.c > > > > index 5ee01231ac4c..2c05c08a0eaf 100644 > > > > --- a/drivers/mtd/nand/raw/meson_nand.c > > > > +++ b/drivers/mtd/nand/raw/meson_nand.c > > > > @@ -176,6 +176,7 @@ struct meson_nfc { > > > > dma_addr_t daddr; > > > > dma_addr_t iaddr; > > > > + u32 info_bytes; > > > > unsigned long assigned_cs; > > > > }; > > > > @@ -503,6 +504,7 @@ static int meson_nfc_dma_buffer_setup(struct nand_chip *nand, void *databuf, > > > > nfc->daddr, datalen, dir); > > > > return ret; > > > > } > > > > + nfc->info_bytes = infolen; > > > > cmd = GENCMDIADDRL(NFC_CMD_AIL, nfc->iaddr); > > > > writel(cmd, nfc->reg_base + NFC_REG_CMD); > > > > @@ -520,8 +522,10 @@ static void meson_nfc_dma_buffer_release(struct nand_chip *nand, > > > > struct meson_nfc *nfc = nand_get_controller_data(nand); > > > > dma_unmap_single(nfc->dev, nfc->daddr, datalen, dir); > > > > - if (infolen) > > > > + if (infolen) { > > > > dma_unmap_single(nfc->dev, nfc->iaddr, infolen, dir); > > > > + nfc->info_bytes = 0; > > > > + } > > > > } > > > > static int meson_nfc_read_buf(struct nand_chip *nand, u8 *buf, int len) > > > > @@ -710,6 +714,8 @@ static void meson_nfc_check_ecc_pages_valid(struct meson_nfc *nfc, > > > > usleep_range(10, 15); > > > > /* info is updated by nfc dma engine*/ > > > > smp_rmb(); > > > > + dma_sync_single_for_cpu(nfc->dev, nfc->iaddr, nfc->info_bytes, > > > > + DMA_FROM_DEVICE); > > > > ret = *info & ECC_COMPLETE; > > > > } while (!ret); > > > > } > > > > > > Thanks, > > Miquèl >
Hi Dmitry, ddrokosov@sberdevices.ru wrote on Mon, 20 Mar 2023 13:43:14 +0300: > Hello guys! > > Was it applied to some nand 'prepare-for-merge' release branch? Not yet, I'll apply it this week in mtd/fixes, you'll be noticed. Thanks, Miquèl > On Mon, Mar 13, 2023 at 12:23:12PM +0100, Neil Armstrong wrote: > > On 13/03/2023 12:18, Miquel Raynal wrote: > > > Hi Arseniy, > > > > > > avkrasnov@sberdevices.ru wrote on Mon, 13 Mar 2023 10:36:11 +0300: > > > > > > > Hello, > > > > > > > > we reproduced this problem on one of our boards. It triggers very rare > > > > when 'usleep_range()' is present, but when sleeping is removed - it fires > > > > always. I suppose problem is with caching, as 'info_buf' memory is mapped by > > > > 'dma_map_single()'. > > > > > > The fix looks really legitimate, indeed I get that the usleep_range() > > > might make it work most of the time but not always. Having this bit in > > > a DMA buf area is a bit strange. Well, the fix LGTM anyway. > > > > Yep it looks legitimate! > > > > LGTM > > > > > > Reviewed-by: Neil Armstrong <neil.armstrong@linaro.org> > > > > > > > > > > > > > Thanks, Arseniy > > > > > > > > On 13.03.2023 10:32, Arseniy Krasnov wrote: > > > > > 'info_buf' memory is cached and driver polls ECC bit in it. This bit > > > > > is set by the NAND controller. If 'usleep_range()' returns before device > > > > > sets this bit, 'info_buf' will be cached and driver won't see update of > > > > > this bit and will loop forever. > > > > > > > > > > Fixes: 8fae856c5350 ("mtd: rawnand: meson: add support for Amlogic NAND flash controller") > > > > > Signed-off-by: Arseniy Krasnov <AVKrasnov@sberdevices.ru> > > > > > --- > > > > > drivers/mtd/nand/raw/meson_nand.c | 8 +++++++- > > > > > 1 file changed, 7 insertions(+), 1 deletion(-) > > > > > > > > > > diff --git a/drivers/mtd/nand/raw/meson_nand.c b/drivers/mtd/nand/raw/meson_nand.c > > > > > index 5ee01231ac4c..2c05c08a0eaf 100644 > > > > > --- a/drivers/mtd/nand/raw/meson_nand.c > > > > > +++ b/drivers/mtd/nand/raw/meson_nand.c > > > > > @@ -176,6 +176,7 @@ struct meson_nfc { > > > > > dma_addr_t daddr; > > > > > dma_addr_t iaddr; > > > > > + u32 info_bytes; > > > > > unsigned long assigned_cs; > > > > > }; > > > > > @@ -503,6 +504,7 @@ static int meson_nfc_dma_buffer_setup(struct nand_chip *nand, void *databuf, > > > > > nfc->daddr, datalen, dir); > > > > > return ret; > > > > > } > > > > > + nfc->info_bytes = infolen; > > > > > cmd = GENCMDIADDRL(NFC_CMD_AIL, nfc->iaddr); > > > > > writel(cmd, nfc->reg_base + NFC_REG_CMD); > > > > > @@ -520,8 +522,10 @@ static void meson_nfc_dma_buffer_release(struct nand_chip *nand, > > > > > struct meson_nfc *nfc = nand_get_controller_data(nand); > > > > > dma_unmap_single(nfc->dev, nfc->daddr, datalen, dir); > > > > > - if (infolen) > > > > > + if (infolen) { > > > > > dma_unmap_single(nfc->dev, nfc->iaddr, infolen, dir); > > > > > + nfc->info_bytes = 0; > > > > > + } > > > > > } > > > > > static int meson_nfc_read_buf(struct nand_chip *nand, u8 *buf, int len) > > > > > @@ -710,6 +714,8 @@ static void meson_nfc_check_ecc_pages_valid(struct meson_nfc *nfc, > > > > > usleep_range(10, 15); > > > > > /* info is updated by nfc dma engine*/ > > > > > smp_rmb(); > > > > > + dma_sync_single_for_cpu(nfc->dev, nfc->iaddr, nfc->info_bytes, > > > > > + DMA_FROM_DEVICE); > > > > > ret = *info & ECC_COMPLETE; > > > > > } while (!ret); > > > > > } > > > > > > > > > Thanks, > > > Miquèl > > >
Hello Miquèl, On Mon, Mar 20, 2023 at 11:55:58AM +0100, Miquel Raynal wrote: > Hi Dmitry, > > ddrokosov@sberdevices.ru wrote on Mon, 20 Mar 2023 13:43:14 +0300: > > > Hello guys! > > > > Was it applied to some nand 'prepare-for-merge' release branch? > > Not yet, I'll apply it this week in mtd/fixes, you'll be noticed. > > Thanks, > Miquèl > Thank you for clarification! [...]
On Mon, 2023-03-13 at 07:32:44 UTC, Arseniy Krasnov wrote: > 'info_buf' memory is cached and driver polls ECC bit in it. This bit > is set by the NAND controller. If 'usleep_range()' returns before device > sets this bit, 'info_buf' will be cached and driver won't see update of > this bit and will loop forever. > > Fixes: 8fae856c5350 ("mtd: rawnand: meson: add support for Amlogic NAND flash controller") > Signed-off-by: Arseniy Krasnov <AVKrasnov@sberdevices.ru> > Reviewed-by: Neil Armstrong <neil.armstrong@linaro.org> Applied to https://git.kernel.org/pub/scm/linux/kernel/git/mtd/linux.git mtd/fixes, thanks. Miquel
diff --git a/drivers/mtd/nand/raw/meson_nand.c b/drivers/mtd/nand/raw/meson_nand.c index 5ee01231ac4c..2c05c08a0eaf 100644 --- a/drivers/mtd/nand/raw/meson_nand.c +++ b/drivers/mtd/nand/raw/meson_nand.c @@ -176,6 +176,7 @@ struct meson_nfc { dma_addr_t daddr; dma_addr_t iaddr; + u32 info_bytes; unsigned long assigned_cs; }; @@ -503,6 +504,7 @@ static int meson_nfc_dma_buffer_setup(struct nand_chip *nand, void *databuf, nfc->daddr, datalen, dir); return ret; } + nfc->info_bytes = infolen; cmd = GENCMDIADDRL(NFC_CMD_AIL, nfc->iaddr); writel(cmd, nfc->reg_base + NFC_REG_CMD); @@ -520,8 +522,10 @@ static void meson_nfc_dma_buffer_release(struct nand_chip *nand, struct meson_nfc *nfc = nand_get_controller_data(nand); dma_unmap_single(nfc->dev, nfc->daddr, datalen, dir); - if (infolen) + if (infolen) { dma_unmap_single(nfc->dev, nfc->iaddr, infolen, dir); + nfc->info_bytes = 0; + } } static int meson_nfc_read_buf(struct nand_chip *nand, u8 *buf, int len) @@ -710,6 +714,8 @@ static void meson_nfc_check_ecc_pages_valid(struct meson_nfc *nfc, usleep_range(10, 15); /* info is updated by nfc dma engine*/ smp_rmb(); + dma_sync_single_for_cpu(nfc->dev, nfc->iaddr, nfc->info_bytes, + DMA_FROM_DEVICE); ret = *info & ECC_COMPLETE; } while (!ret); }
'info_buf' memory is cached and driver polls ECC bit in it. This bit is set by the NAND controller. If 'usleep_range()' returns before device sets this bit, 'info_buf' will be cached and driver won't see update of this bit and will loop forever. Fixes: 8fae856c5350 ("mtd: rawnand: meson: add support for Amlogic NAND flash controller") Signed-off-by: Arseniy Krasnov <AVKrasnov@sberdevices.ru> --- drivers/mtd/nand/raw/meson_nand.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-)