Re: [Uclinux-dist-devel] [PATCH 1/2] mtd: m25p80: Reworkprobing/JEDEC code
From: Barry Song <hidden>
Date: 2010-06-22 06:38:15
Also in:
lkml
On Mon, Jun 21, 2010 at 7:20 PM, Anton Vorontsov [off-list ref] w= rote:
On Mon, Jun 21, 2010 at 06:31:44PM +0800, Barry Song wrote:quoted
On Mon, Jun 21, 2010 at 3:39 PM, Anton Vorontsov <cbouatmailru@gmail.com=wrote:quoted
quoted
On Mon, Jun 21, 2010 at 03:22:48PM +0800, Barry Song wrote:quoted
On Mon, Jun 21, 2010 at 3:15 PM, Anton Vorontsov <cbouatmailru@gmail.=
com> wrote:
quoted
quoted
quoted
quoted
On Mon, Jun 21, 2010 at 11:27:31AM +0800, Barry Song wrote: [...]quoted
quoted
How about we add a non_jedec flag in platform_data, if the flag =
is 1, we
quoted
quoted
quoted
quoted
quoted
quoted
let the detection pass even though the ID is 0? Otherwise, we ne=
ed a
quoted hunk ↗ jump to hunk
quoted
quoted
quoted
quoted
quoted
quoted
valid ID?Here i mean:This will break at least OF-enabled platforms (e.g. PowerPC), they assume that the driver will success for non-JEDEC flashes. OF platforms don't pass platform data, and even if they did, device tree doesn't specify if the flash is JEDEC or non-JEDEC. Which is why I think that, by default, the driver should successfully register the flash even if JEDEC probe fails. So, instead of checking for "!non_jedec", I would recommend "force_jedec" check.Mike Frysinger suggested to use non_jedec since most devices are standard jedec devices.Well, on OF platforms most devices that I'm aware of are non-JEDEC.quoted
Only if non_jedec=3D1, we let the detection pass if ID is 0.Then please #ifdef it with CONFIG_OF.I think the patch has nothing to do with platform. Here SPI Flash is a peripherals, doesn't depend on any platform. Adding a CONFIG_OF doesn't make sense very much.With OF we don't place non-existent devices into the device tree (or we mark them with status =3D "not-ok/disabled/absent" property).quoted
If you think most devices are non-JEDEC, we can change non_JEDEC to force_JEDEC as you said. But anyway, is that real that most devices are non_JEDEC?Why would this matter? We have to support both.quoted
If not, I think we should change OF platform codes to fit with this patch.You can't easily change OF. It's like "let's change ACPI tables or BIOS in these PCs". Doable, but involves things like reflashing. And we usually have to support old BIOSes as well. OTOH, I see (git grep m25p arch/powerpc/boot/dts/) that in mainline kernel only MPC8569 board has a correct m25p node, and it is STMicro variant (it is JEDEC capable). As we don't really have to support out of tree code, I'd just go with this patch, assuming that we have to change device tree for boards with non-JEDEC flashes. It's effectively the same thing as platform data flag, except that it works automatically for OF platforms. Signed-off-by: Anton Vorontsov <redacted> ---diff --git a/drivers/mtd/devices/m25p80.c b/drivers/mtd/devices/m25p80.c index 81e49a9..a610ca9 100644 --- a/drivers/mtd/devices/m25p80.c +++ b/drivers/mtd/devices/m25p80.c@@ -680,6 +680,16 @@ static const struct spi_device_id m25p_ids[] =3D {=C2=A0 =C2=A0 =C2=A0 =C2=A0{ "m25p64", =C2=A0INFO(0x202017, =C2=A00, =C2=
=A064 * 1024, 128, 0) },
=C2=A0 =C2=A0 =C2=A0 =C2=A0{ "m25p128", INFO(0x202018, =C2=A00, 256 * 102=4, =C2=A064, 0) },
+ =C2=A0 =C2=A0 =C2=A0 { "m25p05-nonjedec", =C2=A0INFO(0, 0, =C2=A032 * 1=024, =C2=A0 2, 0) },
+ =C2=A0 =C2=A0 =C2=A0 { "m25p10-nonjedec", =C2=A0INFO(0, 0, =C2=A032 * 1=024, =C2=A0 4, 0) },
+ =C2=A0 =C2=A0 =C2=A0 { "m25p20-nonjedec", =C2=A0INFO(0, 0, =C2=A064 * 1=024, =C2=A0 4, 0) },
+ =C2=A0 =C2=A0 =C2=A0 { "m25p40-nonjedec", =C2=A0INFO(0, 0, =C2=A064 * 1=024, =C2=A0 8, 0) },
+ =C2=A0 =C2=A0 =C2=A0 { "m25p80-nonjedec", =C2=A0INFO(0, 0, =C2=A064 * 1=024, =C2=A016, 0) },
+ =C2=A0 =C2=A0 =C2=A0 { "m25p16-nonjedec", =C2=A0INFO(0, 0, =C2=A064 * 1=024, =C2=A032, 0) },
+ =C2=A0 =C2=A0 =C2=A0 { "m25p32-nonjedec", =C2=A0INFO(0, 0, =C2=A064 * 1=024, =C2=A064, 0) },
+ =C2=A0 =C2=A0 =C2=A0 { "m25p64-nonjedec", =C2=A0INFO(0, 0, =C2=A064 * 1=024, 128, 0) },
+ =C2=A0 =C2=A0 =C2=A0 { "m25p128-nonjedec", INFO(0, 0, 256 * 1024, =C2==A064, 0) },
+
=C2=A0 =C2=A0 =C2=A0 =C2=A0{ "m45pe10", INFO(0x204011, =C2=A00, 64 * 1024=, =C2=A0 =C2=A02, 0) },
=C2=A0 =C2=A0 =C2=A0 =C2=A0{ "m45pe80", INFO(0x204014, =C2=A00, 64 * 1024=, =C2=A0 16, 0) },
=C2=A0 =C2=A0 =C2=A0 =C2=A0{ "m45pe16", INFO(0x204015, =C2=A00, 64 * 1024=, =C2=A0 32, 0) },
quoted hunk ↗ jump to hunk
@@ -795,8 +805,7 @@ static int __devinit m25p_probe(struct spi_device *sp=
i)
=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0jid =3D jedec_prob=
e(spi);
=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0if (!jid) {
- =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 ==C2=A0 dev_info(&spi->dev, "non-JEDEC variant of %s\n",
- =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =
=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0id->name);
+ =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =
=C2=A0 return -ENODEV;
The patch looks good to me. Only problem is NULL is also returned by
spi_write_then_read() fail:
static const struct spi_device_id *__devinit jedec_probe(struct spi_device =
*spi)
{
int tmp;
u8 code =3D OPCODE_RDID;
u8 id[5];
u32 jedec;
u16 ext_jedec;
struct flash_info *info;
/* JEDEC also defines an optional "extended device information"
* string for after vendor-specific data, after the three bytes
* we use here. Supporting some chips might require using it.
*/
tmp =3D spi_write_then_read(spi, &code, 1, id, 5);
if (tmp < 0) {
DEBUG(MTD_DEBUG_LEVEL0, "%s: error %d reading JEDEC ID\n",
dev_name(&spi->dev), tmp);
return NULL;
}
...
}
Here much better for -EIO (return tmp)?
=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0} else if (jid !=
=3D id) {=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=
=A0 =C2=A0/*
=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=
=A0 =C2=A0 * JEDEC knows better, so overwrite platform ID. We