| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Date: Mon, 25 Feb 2019 16:45:49 +0000
From: Avri Altman <Avri.Altman@....com>
To: Ulf Hansson <ulf.hansson@...aro.org>
CC: "linux-mmc@...r.kernel.org" <linux-mmc@...r.kernel.org>,
Wolfram Sang <wsa+renesas@...g-engineering.com>,
Adrian Hunter <adrian.hunter@...el.com>,
Jaehoon Chung <jh80.chung@...sung.com>,
Shawn Lin <shawn.lin@...k-chips.com>,
Avi Shchislowski <Avi.Shchislowski@....com>,
Alex Lemberg <Alex.Lemberg@....com>,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: RE: [PATCH v2 3/3] mmc: core: Add discard support to sd
Ulf hi,
Thanks a lot for your comments.
>
> On Wed, 6 Feb 2019 at 12:29, Avri Altman <avri.altman@....com> wrote:
> >
> > SD spec v5.1 adds discard support. The flows and commands are similar to
> > mmc, so just set the discard arg in CMD38.
>
> So this means that we from now on, if the SD card supports discard, we
> are going to use it in favor of erase. This is consistent with how we
> treat eMMC, so I think it makes sense. Could you perhaps fold in some
> information about this in the changelog, to make this clear on what
> this change means.
Done.
>
> You may also want to explain a little bit what the difference is from
> the SD card storage point of view. Like after an erase, it either 1 or
> 0s on the media, while after a discard it can be whatever.
Done.
>
> >
> > Actually, there is no need to check for the spec version like we are
> > doing, as it is assured that the reserved bits in earlier versions are
> > null. Do that anyway to document the spec version that introduce it.
>
> The check is needed for other purposes, so please just drop this statement.
Done.
>
> >
> > Signed-off-by: Avri Altman <avri.altman@....com>
> > ---
> > drivers/mmc/core/core.c | 6 +++++-
> > drivers/mmc/core/sd.c | 10 +++++++++-
> > include/linux/mmc/sd.h | 1 +
> > 3 files changed, 15 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/mmc/core/core.c b/drivers/mmc/core/core.c
> > index de0f1a1..4d62f28 100644
> > --- a/drivers/mmc/core/core.c
> > +++ b/drivers/mmc/core/core.c
> > @@ -2164,7 +2164,7 @@ static unsigned int mmc_align_erase_size(struct
> mmc_card *card,
> > * @card: card to erase
> > * @from: first sector to erase
> > * @nr: number of sectors to erase
> > - * @arg: erase command argument (SD supports only %SD_ERASE_ARG)
> > + * @arg: erase command argument
> > *
> > * Caller must claim host before calling this function.
> > */
> > @@ -2181,6 +2181,9 @@ int mmc_erase(struct mmc_card *card, unsigned
> int from, unsigned int nr,
> > if (!card->erase_size)
> > return -EOPNOTSUPP;
> >
> > + if (mmc_card_sd(card) && arg == SD_DISCARD_ARG)
> > + goto skip_arg_testing;
> > +
>
> This isn't consistent with the rest of the code path in this function,
> please adopt to that.
Done.
>
> > if (mmc_card_sd(card) && arg != SD_ERASE_ARG)
>
> Couldn't you add a check for !SD_DISCARD_ARG here instead?
Done.
>
> > return -EOPNOTSUPP;
> >
> > @@ -2200,6 +2203,7 @@ int mmc_erase(struct mmc_card *card, unsigned
> int from, unsigned int nr,
> > if (arg == MMC_ERASE_ARG)
> > nr = mmc_align_erase_size(card, &from, &to, nr);
> >
> > +skip_arg_testing:
> > if (nr == 0)
> > return 0;
> >
> > diff --git a/drivers/mmc/core/sd.c b/drivers/mmc/core/sd.c
> > index c2db94d..2b4fc22 100644
> > --- a/drivers/mmc/core/sd.c
> > +++ b/drivers/mmc/core/sd.c
> > @@ -231,6 +231,8 @@ static int mmc_read_ssr(struct mmc_card *card)
> > {
> > unsigned int au, es, et, eo;
> > __be32 *raw_ssr;
> > + u32 resp[4] = {};
> > + u8 discard_support;
> > int i;
> >
> > if (!(card->csd.cmdclass & CCC_APP_SPEC)) {
> > @@ -276,7 +278,13 @@ static int mmc_read_ssr(struct mmc_card *card)
> > }
> > }
> >
> > - card->erase_arg = SD_ERASE_ARG;
> > + /*
> > + * starting SD5.1 discard is supported if DISCARD_SUPPORT (b313) is
> set
> > + */
> > + resp[3] = card->raw_ssr[6];
> > + discard_support = UNSTUFF_BITS(resp, 313 - 288, 1);
>
> Couldn't you just replace this with "discard_support =
> UNSTUFF_BITS(card->raw_ssr, 313 - 256, 1);" ?
SD status register is transmitted from 512bit onwards.
The 512th bit is the one transmitted first and so on.
So the 313th bit has to be checked at byte offset 0x18 in which both discard and FULE were enabled.
Byte offset 0x18 means dword offset 6, and UNSTUFF_BITS accounts for 4 dwords only.
Using UNSTUFF_BITS(card->raw_ssr,....) you can only access bits 511..383, but not 313.
>
> > + card->erase_arg = (card->scr.sda_specx && discard_support) ?
> > + SD_DISCARD_ARG : SD_ERASE_ARG;
> >
> > return 0;
> > }
> > diff --git a/include/linux/mmc/sd.h b/include/linux/mmc/sd.h
> > index 1a6d10f..ec94a5a 100644
> > --- a/include/linux/mmc/sd.h
> > +++ b/include/linux/mmc/sd.h
> > @@ -95,5 +95,6 @@
> > * Erase/discard
> > */
> > #define SD_ERASE_ARG 0x00000000
> > +#define SD_DISCARD_ARG 0x00000001
> >
> > #endif /* LINUX_MMC_SD_H */
> > --
> > 1.9.1
> >
>
> Besides the above changes, there is more important thing I think you
> have left out to consider. With discard the operation should be
> completed by the card within 250ms, while for erase the timeout
> depends on the number of blocks to be erased. I am fine if we address
> that on top this change though, just want to point it out so we don't
> forget it.
Done.
Will add another patch to account for that.
Thanks a lot,
Avri
>
> Kind regards
> Uffe
Powered by blists - more mailing lists