| Message ID | 1515443808-31105-1-git-send-email-kdasu.kdev@gmail.com |
|---|---|
| State | Accepted |
| Delegated to: | Boris Brezillon |
| Headers | show |
| Series | [v3] mtd: nand: brcmnand: Disable prefetch by default | expand |
On 01/08/2018 12:36 PM, Kamal Dasu wrote: > Brcm nand controller prefetch feature needs to be disabled > by default. Enabling affects performance on random reads as > well as dma reads. > > Signed-off-by: Kamal Dasu <kdasu.kdev@gmail.com> > Fixes: 27c5b17cd1b1 ("mtd: nand: add NAND driver "library" for Broadcom STB NAND controller") > Cc: stable@vger.kernel.org Acked-by: Florian Fainelli <f.fainelli@gmail.com>
On 01/08/2018 12:36 PM, Kamal Dasu wrote: > Brcm nand controller prefetch feature needs to be disabled > by default. Enabling affects performance on random reads as > well as dma reads. > > Signed-off-by: Kamal Dasu <kdasu.kdev@gmail.com> > Fixes: 27c5b17cd1b1 ("mtd: nand: add NAND driver "library" for Broadcom STB NAND controller") > Cc: stable@vger.kernel.org Boris, is there a reason this is not in your v4.16 pull request? If not, could you take that fix? Thank you
On Sat, 20 Jan 2018 15:13:35 -0800 Florian Fainelli <f.fainelli@gmail.com> wrote: > On 01/08/2018 12:36 PM, Kamal Dasu wrote: > > Brcm nand controller prefetch feature needs to be disabled > > by default. Enabling affects performance on random reads as > > well as dma reads. > > > > Signed-off-by: Kamal Dasu <kdasu.kdev@gmail.com> > > Fixes: 27c5b17cd1b1 ("mtd: nand: add NAND driver "library" for Broadcom STB NAND controller") > > Cc: stable@vger.kernel.org > > Boris, is there a reason this is not in your v4.16 pull request? Look carefully, it is there ;-). " Kamal Dasu (1): mtd: nand: brcmnand: Disable prefetch by default " > If not, could you take that fix? > > Thank you
On 01/20/2018 03:53 PM, Boris Brezillon wrote: > On Sat, 20 Jan 2018 15:13:35 -0800 > Florian Fainelli <f.fainelli@gmail.com> wrote: > >> On 01/08/2018 12:36 PM, Kamal Dasu wrote: >>> Brcm nand controller prefetch feature needs to be disabled >>> by default. Enabling affects performance on random reads as >>> well as dma reads. >>> >>> Signed-off-by: Kamal Dasu <kdasu.kdev@gmail.com> >>> Fixes: 27c5b17cd1b1 ("mtd: nand: add NAND driver "library" for Broadcom STB NAND controller") >>> Cc: stable@vger.kernel.org >> >> Boris, is there a reason this is not in your v4.16 pull request? > > Look carefully, it is there ;-). Indeed, I can't read, my bad.
diff --git a/drivers/mtd/nand/brcmnand/brcmnand.c b/drivers/mtd/nand/brcmnand/brcmnand.c index dd56a67..2a978d9 100644 --- a/drivers/mtd/nand/brcmnand/brcmnand.c +++ b/drivers/mtd/nand/brcmnand/brcmnand.c @@ -2193,16 +2193,9 @@ static int brcmnand_setup_dev(struct brcmnand_host *host) if (ctrl->nand_version >= 0x0702) tmp |= ACC_CONTROL_RD_ERASED; tmp &= ~ACC_CONTROL_FAST_PGM_RDIN; - if (ctrl->features & BRCMNAND_HAS_PREFETCH) { - /* - * FIXME: Flash DMA + prefetch may see spurious erased-page ECC - * errors - */ - if (has_flash_dma(ctrl)) - tmp &= ~ACC_CONTROL_PREFETCH; - else - tmp |= ACC_CONTROL_PREFETCH; - } + if (ctrl->features & BRCMNAND_HAS_PREFETCH) + tmp &= ~ACC_CONTROL_PREFETCH; + nand_writereg(ctrl, offs, tmp); return 0;
Brcm nand controller prefetch feature needs to be disabled by default. Enabling affects performance on random reads as well as dma reads. Signed-off-by: Kamal Dasu <kdasu.kdev@gmail.com> Fixes: 27c5b17cd1b1 ("mtd: nand: add NAND driver "library" for Broadcom STB NAND controller") Cc: stable@vger.kernel.org --- drivers/mtd/nand/brcmnand/brcmnand.c | 13 +++---------- 1 file changed, 3 insertions(+), 10 deletions(-)