Message ID | 1481219279-6982-1-git-send-email-linus.walleij@linaro.org |
---|---|
State | Deferred |
Headers | show |
On 2016-12-08 18:47, Linus Walleij wrote: > The compatible string for an EEPROM in the device tree is currently > completely ignored by the kernel, simply stated it will not make the > corresponding AT24 EEPROM driver probe properly. It is instead still > relying on the DT node name to be set to one of the I2C device IDs > which works due to a side effect in the I2C DT parsing code. > > Fix this up by making the DT probe mechanism a bit more elaborate: > actually match on the compatible strings defined in the device > tree bindings in Documentation/devicetree/bindings/eeprom/eeprom.txt: > map these to the corresponding I2C IDs by name and look up the > magic flags from the I2C ID before proceeding, also make the DT > compatible string take precedence. > > Keep the second DT parsing callback that sets up per-chip flags as > this needs to happen after mangling the magic flags passed from the > I2C ID table. > > All vendor compatible strings listed in the binding document are > added to the driver. > > After this it is possible to name the device tree node for the EEPROM > whatever you actually like to call it, and the probing will be done > from the compatible string. > > Before this patch, the following device tree node does not probe, > which might be considered a bug: > > eeprom@52 { > compatible = "atmel,at24c128"; The way I read it, that should be "atmel,24c128", i.e. w/o the "at" prefix. > reg = <0x52>; > pagesize = <64>; > }; > > After this patch, the driver probes fine from this node. The bindings says: Required properties: - compatible should be "<manufacturer>,<type>", like these: and then lists the compatibles you have added to the match table (but you have this extra "at" prefix for the atmel manufacturer). The way I read the above, you are free to use any manufacturer and still have it work, and indeed, I have success with this node: eeprom@50 { compatible = "nxp,24c02"; reg = <0x50>; pagesize = <16>; }; I fear that your patch will regress this matching on the wildcard manufacturer. I haven't tested that though, but there are enough question marks anyway... Cheers, Peter -- To unsubscribe from this list: send the line "unsubscribe linux-i2c" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Thu, Dec 8, 2016 at 7:23 PM, Peter Rosin <peda@axentia.se> wrote: > On 2016-12-08 18:47, Linus Walleij wrote: >> Before this patch, the following device tree node does not probe, >> which might be considered a bug: >> >> eeprom@52 { >> compatible = "atmel,at24c128"; > > The way I read it, that should be "atmel,24c128", i.e. w/o the "at" prefix. (...) > and then lists the compatibles you have added to the match table (but you > have this extra "at" prefix for the atmel manufacturer). > > The way I read the above, you are free to use any manufacturer and still > have it work, and indeed, I have success with this node: > > eeprom@50 { > compatible = "nxp,24c02"; > reg = <0x50>; > pagesize = <16>; > }; > > I fear that your patch will regress this matching on the wildcard > manufacturer. I haven't tested that though, but there are enough > question marks anyway... Bah I probably just screwed up syntactically and now worked around a non-existing problem. I will try as you suggest, just "vendor,type" and see if it works. It probably does. Some days I feel just utterly stupid. Sorry for the fuzz. Wolfram: ignore the patch for now. Yours, Linus Walleij -- To unsubscribe from this list: send the line "unsubscribe linux-i2c" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 2016-12-09 00:32, Linus Walleij wrote: > On Thu, Dec 8, 2016 at 7:23 PM, Peter Rosin <peda@axentia.se> wrote: >> On 2016-12-08 18:47, Linus Walleij wrote: > >>> Before this patch, the following device tree node does not probe, >>> which might be considered a bug: >>> >>> eeprom@52 { >>> compatible = "atmel,at24c128"; >> >> The way I read it, that should be "atmel,24c128", i.e. w/o the "at" prefix. > (...) >> and then lists the compatibles you have added to the match table (but you >> have this extra "at" prefix for the atmel manufacturer). >> >> The way I read the above, you are free to use any manufacturer and still >> have it work, and indeed, I have success with this node: >> >> eeprom@50 { >> compatible = "nxp,24c02"; >> reg = <0x50>; >> pagesize = <16>; >> }; >> >> I fear that your patch will regress this matching on the wildcard >> manufacturer. I haven't tested that though, but there are enough >> question marks anyway... > > Bah I probably just screwed up syntactically and now worked around > a non-existing problem. I will try as you suggest, just "vendor,type" > and see if it works. It probably does. But it is a bit strange. Grepping for compatible.*24c finds quite a few instances of "bad" compatible strings. Many on the patterns at,24c256 and at24,24c256 (should be probably be atmel,24c256) but also a few atmel,at24c16 and atmel,at24c128b. I don't understand how those last ones ever worked, if it is not working for you? Especially those with the trailing "b". WTF? > Some days I feel just utterly stupid. Sorry for the fuzz. Join the club... Cheers, Peter > Wolfram: ignore the patch for now. -- To unsubscribe from this list: send the line "unsubscribe linux-i2c" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
> Many on the patterns at,24c256 and at24,24c256 (should be probably > be atmel,24c256) I remember patches fixing that. Since I usually don't take DTS patches, I can't recall what happened to them. It might well be that those were accepted and meanwhile new bad bindings got in. > but also a few atmel,at24c16 and atmel,at24c128b. > I don't understand how those last ones ever worked, if it is not > working for you? Especially those with the trailing "b". WTF? I'd simply assume it was never tested. EEPROMs are often convenience storage and not essential for a working board. Also, for historic reasons, they are often used via the i2c-dev interface directly, so a wrong binding with the kernel driver might simply go unnoticed.
On Fri, Dec 9, 2016 at 6:48 AM, Peter Rosin <peda@axentia.se> wrote: > On 2016-12-09 00:32, Linus Walleij wrote: >> Bah I probably just screwed up syntactically and now worked around >> a non-existing problem. I will try as you suggest, just "vendor,type" >> and see if it works. It probably does. (It does) > But it is a bit strange. Grepping for compatible.*24c finds quite > a few instances of "bad" compatible strings. > > Many on the patterns at,24c256 and at24,24c256 (should be probably > be atmel,24c256) but also a few atmel,at24c16 and atmel,at24c128b. > I don't understand how those last ones ever worked, if it is not > working for you? Especially those with the trailing "b". WTF? Those are the erroneous device trees that I copied as inspiration instead of looking directly at the bindings. Yours, Linus Walleij -- To unsubscribe from this list: send the line "unsubscribe linux-i2c" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/Documentation/devicetree/bindings/vendor-prefixes.txt b/Documentation/devicetree/bindings/vendor-prefixes.txt index f0a48ea78659..40bdf9aa590c 100644 --- a/Documentation/devicetree/bindings/vendor-prefixes.txt +++ b/Documentation/devicetree/bindings/vendor-prefixes.txt @@ -47,6 +47,7 @@ brcm Broadcom Corporation buffalo Buffalo, Inc. calxeda Calxeda capella Capella Microsystems, Inc +catalyst Catalyst Semiconductor Inc. cavium Cavium, Inc. cdns Cadence Design Systems Inc. ceva Ceva, Inc. diff --git a/drivers/misc/eeprom/at24.c b/drivers/misc/eeprom/at24.c index 051b14766ef9..246b15539d45 100644 --- a/drivers/misc/eeprom/at24.c +++ b/drivers/misc/eeprom/at24.c @@ -20,6 +20,7 @@ #include <linux/bitops.h> #include <linux/jiffies.h> #include <linux/of.h> +#include <linux/of_device.h> #include <linux/acpi.h> #include <linux/i2c.h> #include <linux/nvmem-provider.h> @@ -563,23 +564,70 @@ static int at24_write(void *priv, unsigned int off, void *val, size_t count) } #ifdef CONFIG_OF -static void at24_get_ofdata(struct i2c_client *client, - struct at24_platform_data *chip) +static void at24_get_of_magic(struct device *dev, + kernel_ulong_t *magic) { - const __be32 *val; - struct device_node *node = client->dev.of_node; - - if (node) { - if (of_get_property(node, "read-only", NULL)) - chip->flags |= AT24_FLAG_READONLY; - val = of_get_property(node, "pagesize", NULL); - if (val) - chip->page_size = be32_to_cpup(val); + const char *name; + const struct i2c_device_id *id; + int i; + + name = of_device_get_match_data(dev); + if (!name) + return; + + for (i = 0; i < ARRAY_SIZE(at24_ids); i++) { + id = &at24_ids[i]; + if (!strcmp(id->name, name)) { + *magic = id->driver_data; + break; + } } + if (i == ARRAY_SIZE(at24_ids)) + return; + + dev_dbg(dev, "DT match for %s -> %s\n", name, id->name); +} + +static void at24_get_of_chipdata(struct device_node *np, + struct at24_platform_data *chip) +{ + u32 val; + int ret; + + if (of_property_read_bool(np, "read-only")) + chip->flags |= AT24_FLAG_READONLY; + + ret = of_property_read_u32(np, "pagesize", &val); + if (!ret) + chip->page_size = val; } + +static const struct of_device_id at24_of_match[] = { + { .compatible = "atmel,at24c00", .data = "24c00" }, + { .compatible = "atmel,at24c01", .data = "24c01" }, + { .compatible = "atmel,at24c02", .data = "24c02" }, + { .compatible = "atmel,at24c04", .data = "24c04" }, + { .compatible = "atmel,at24c08", .data = "24c08" }, + { .compatible = "atmel,at24c16", .data = "24c16" }, + { .compatible = "atmel,at24c32", .data = "24c32" }, + { .compatible = "atmel,at24c64", .data = "24c64" }, + { .compatible = "atmel,at24c128", .data = "24c128" }, + { .compatible = "atmel,at24c256", .data = "24c256" }, + { .compatible = "atmel,at24c512", .data = "24c512" }, + { .compatible = "atmel,at24c1024", .data = "24c1024" }, + { .compatible = "catalyst,24c32", .data = "24c32" }, + { .compatible = "ramtron,24c64", .data = "24c64" }, + { .compatible = "renesas,r1ex24002", .data = "24c02" }, + { }, +}; +MODULE_DEVICE_TABLE(of, at24_of_match); + #else -static void at24_get_ofdata(struct i2c_client *client, - struct at24_platform_data *chip) +static void at24_get_of_magic(struct device *dev, + kernel_ulong_t *magic) +{ } +static void at24_get_of_chipdata(struct device_node *np, + struct at24_platform_data *chip) { } #endif /* CONFIG_OF */ @@ -598,7 +646,10 @@ static int at24_probe(struct i2c_client *client, const struct i2c_device_id *id) if (client->dev.platform_data) { chip = *(struct at24_platform_data *)client->dev.platform_data; } else { - if (id) { + if (client->dev.of_node) { + /* Get chipdata if OF is present */ + at24_get_of_magic(&client->dev, &magic); + } else if (id) { magic = id->driver_data; } else { const struct acpi_device_id *aid; @@ -621,7 +672,7 @@ static int at24_probe(struct i2c_client *client, const struct i2c_device_id *id) chip.page_size = 1; /* update chipdata if OF is present */ - at24_get_ofdata(client, &chip); + at24_get_of_chipdata(client->dev.of_node, &chip); chip.setup = NULL; chip.context = NULL; @@ -822,6 +873,7 @@ static struct i2c_driver at24_driver = { .driver = { .name = "at24", .acpi_match_table = ACPI_PTR(at24_acpi_ids), + .of_match_table = of_match_ptr(at24_of_match), }, .probe = at24_probe, .remove = at24_remove,
The compatible string for an EEPROM in the device tree is currently completely ignored by the kernel, simply stated it will not make the corresponding AT24 EEPROM driver probe properly. It is instead still relying on the DT node name to be set to one of the I2C device IDs which works due to a side effect in the I2C DT parsing code. Fix this up by making the DT probe mechanism a bit more elaborate: actually match on the compatible strings defined in the device tree bindings in Documentation/devicetree/bindings/eeprom/eeprom.txt: map these to the corresponding I2C IDs by name and look up the magic flags from the I2C ID before proceeding, also make the DT compatible string take precedence. Keep the second DT parsing callback that sets up per-chip flags as this needs to happen after mangling the magic flags passed from the I2C ID table. All vendor compatible strings listed in the binding document are added to the driver. After this it is possible to name the device tree node for the EEPROM whatever you actually like to call it, and the probing will be done from the compatible string. Before this patch, the following device tree node does not probe, which might be considered a bug: eeprom@52 { compatible = "atmel,at24c128"; reg = <0x52>; pagesize = <64>; }; After this patch, the driver probes fine from this node. As checkpatch is complaining about the vendor "catalyst" not existing in the vendor prefixes, despite being mentioned in the EEPROM DT binding document, we add this as part of this patch so that checkpatch is happy. Cc: devicetree@vger.kernel.org Signed-off-by: Linus Walleij <linus.walleij@linaro.org> --- .../devicetree/bindings/vendor-prefixes.txt | 1 + drivers/misc/eeprom/at24.c | 82 ++++++++++++++++++---- 2 files changed, 68 insertions(+), 15 deletions(-)