From: Pratyush Yadav <p.yadav@ti.com>
To: Michael Walle <michael@walle.cc>
Cc: <linux-kernel@vger.kernel.org>, <linux-mtd@lists.infradead.org>,
Tudor Ambarus <tudor.ambarus@microchip.com>,
Miquel Raynal <miquel.raynal@bootlin.com>,
Richard Weinberger <richard@nod.at>,
Vignesh Raghavendra <vigneshr@ti.com>
Subject: Re: [PATCH 1/2] mtd: spi-nor: sfdp: save a copy of the SFDP data
Date: Tue, 23 Mar 2021 15:07:49 +0530 [thread overview]
Message-ID: <20210323093747.snojk5kq7p6im4rq@ti.com> (raw)
In-Reply-To: <7e00a5fc05a186d5b34916e5a9f45a48@walle.cc>
On 22/03/21 11:31PM, Michael Walle wrote:
> Am 2021-03-22 19:42, schrieb Pratyush Yadav:
> > On 22/03/21 04:32PM, Michael Walle wrote:
> > > Am 2021-03-22 15:21, schrieb Pratyush Yadav:
> > > > On 18/03/21 10:24AM, Michael Walle wrote:
> > > > > +
> > > > > + sfdp->num_dwords = DIV_ROUND_UP(sfdp_size, sizeof(*sfdp->dwords));
> > > >
> > > > The SFDP spec says that Parameter Table Pointer should be DWORD aligned
> > > > and Parameter Table length is specified in number of DWORDs. So,
> > > > sfdp_size should always be a multiple of 4. Any SFDP table where this is
> > > > not true is an invalid one.
> > > >
> > > > Also, the spec says "Device behavior when the Read SFDP command crosses
> > > > the SFDP structure boundary is not defined".
> > > >
> > > > So I think this should be a check for alignment instead of a round-up.
> > >
> > > Well, that woundn't help for debugging. I.e. you also want the SFDP
> > > data
> > > in cases like this. IMHO we should try hard enough to actually get a
> > > reasonable dump.
> > >
> > > OTOH we also rely on the header and the pointers in the header. Any
> > > other ideas, but just to chicken out?
> >
> > Honestly, I don't think reading past the SFDP boundary would be too bad.
> > It probably will just be some garbage data. But if you want to avoid
> > that, you can always round down instead of up.
>
> Like I said, while the storage will be rounded up to a multiple of
> DWORDs, only sfdp_size is transferred. Thus it case a pointer is not
> DWORD aligned, we end up with zeros at the end.
>
> I'll add a comment.
Right.
> > This way you will only
> > miss the last DWORD at most. In either case, a warning should be printed
> > so this problem can be brought to the user's attention.
>
> I was about to add a warning/debug message. But its the wrong place.
> It should really be checked in the for loop which iterates over the
> headers before parsing them. You could check sfdp_size but then two
> unaligned param pointers might cancel each other out.
>
> This can be a seperate patch, besides adding a warning, should there
> be any other things to do, e.g. stop parsing and error out?
Just removing the bad table from the "tables to parse" list should be
the most conservative option.
>
> ..
>
> > > > > + goto exit;
> > > > > + }
> > > > > +
> > > > > + err = spi_nor_read_sfdp_dma_unsafe(nor, 0, sfdp_size, sfdp->dwords);
>
> Btw, this can be spi_nor_read_sfdp(). But I'm not sure, what this
> whole dma capable buffer should be. Is kmalloc(GFP_KERNEL)
> considered DMA safe?
I think spi_nor_read_sfdp_dma_unsafe() is meant for buffers that are
allocated on stack. Both its current users pass in buffers on the stack.
Also see bfa4133795e5 (mtd: spi-nor: fix DMA unsafe buffer issue in
spi_nor_read_sfdp(), 2017-09-06).
sfdp->dwords is a DMA safe buffer, so you should directly use
spi_nor_read_sfdp() here. All spi_nor_read_sfdp_dma_unsafe() does is
allocate a buffer using kmalloc(GFP_KERNEL), pass it to
spi_nor_read_sfdp() and copy the contents back.
>
> The buffer ends in spi_nor_read_data(), which is also called from
> mtdcore:
>
> spi_nor_read_sfdp()
> spi_nor_read_raw()
> spi_nor_read_data()
>
> mtd_read()
> mtd_read_oob()
> mtd_read_oob_std()
> spi_nor_read()
> spi_nor_read_data()
>
> Is the buffer passed from mtd_read() also DMA-safe? Doesn't the SPI
> drivers allocate DMA safe buffers if they need them?
SPI MEM at least requires the buffers to be DMA-safe. The comment for
data.buf.in says "input buffer (must be DMA-able)".
Dunno if mtd_read() always passes a DMA-safe buffer though.
--
Regards,
Pratyush Yadav
Texas Instruments Inc.
next prev parent reply other threads:[~2021-03-23 9:38 UTC|newest]
Thread overview: 22+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-03-18 9:24 [PATCH 0/2] mtd: spi-nor: support dumping sfdp tables Michael Walle
2021-03-18 9:24 ` [PATCH 1/2] mtd: spi-nor: sfdp: save a copy of the SFDP data Michael Walle
2021-03-22 14:21 ` Pratyush Yadav
2021-03-22 15:32 ` Michael Walle
2021-03-22 15:48 ` Michael Walle
2021-03-22 18:42 ` Pratyush Yadav
2021-03-22 22:31 ` Michael Walle
2021-03-23 9:37 ` Pratyush Yadav [this message]
2021-04-05 13:11 ` Tudor.Ambarus
2021-04-05 15:07 ` Michael Walle
2021-04-05 15:42 ` Tudor.Ambarus
2021-04-05 16:03 ` Michael Walle
2021-03-18 9:24 ` [PATCH 2/2] mtd: spi-nor: add initial sysfs support Michael Walle
2021-03-20 4:16 ` Yicong Yang
2021-03-20 19:17 ` Michael Walle
2021-03-22 14:43 ` Pratyush Yadav
2021-03-22 14:57 ` Michael Walle
2021-04-06 7:56 ` Vignesh Raghavendra
2021-04-06 8:47 ` Michael Walle
2021-04-06 11:43 ` Vignesh Raghavendra
2021-04-29 15:37 ` Alexander Williams
2021-04-29 15:46 ` Michael Walle
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20210323093747.snojk5kq7p6im4rq@ti.com \
--to=p.yadav@ti.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mtd@lists.infradead.org \
--cc=michael@walle.cc \
--cc=miquel.raynal@bootlin.com \
--cc=richard@nod.at \
--cc=tudor.ambarus@microchip.com \
--cc=vigneshr@ti.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).