[RFC 1/7] mmc: fsl_esdhc_imx: add OF_PLATDATA support
Walter Lozano
walter.lozano at collabora.com
Fri Apr 17 22:05:31 CEST 2020
Hi Simon,
On 9/4/20 16:50, Simon Glass wrote:
> Hi Walter,
>
> On Thu, 9 Apr 2020 at 13:44, Walter Lozano <walter.lozano at collabora.com> wrote:
>> Hi Simon,
>>
>> On 9/4/20 16:36, Simon Glass wrote:
>>> Hi Walter,
>>>
>>> On Thu, 9 Apr 2020 at 13:07, Walter Lozano <walter.lozano at collabora.com> wrote:
>>>> Hi Simon,
>>>>
>>>> On 8/4/20 00:14, Simon Glass wrote:
>>>>> Hi Walter,
>>>>>
>>>>> On Tue, 7 Apr 2020 at 14:05, Walter Lozano <walter.lozano at collabora.com> wrote:
>>>>>> Hi Simon,
>>>>>>
>>>>>> Thank you for taking the time to review this series.
>>>>>>
>>>>>> On 6/4/20 00:42, Simon Glass wrote:
>>>>>>> Hi Walter,
>>>>>>>
>>>>>>> On Sun, 29 Mar 2020 at 21:32, Walter Lozano<walter.lozano at collabora.com> wrote:
>>>>>>>> Signed-off-by: Walter Lozano<walter.lozano at collabora.com>
>>>>>>>> ---
>>>>>>>> drivers/mmc/fsl_esdhc_imx.c | 46 +++++++++++++++++++++++++++++++++----
>>>>>>>> 1 file changed, 42 insertions(+), 4 deletions(-)
>>>>>>>>
>>>>>>>> diff --git a/drivers/mmc/fsl_esdhc_imx.c b/drivers/mmc/fsl_esdhc_imx.c
>>>>>>>> index 4900498e9b..761a4b46e9 100644
>>>>>>>> --- a/drivers/mmc/fsl_esdhc_imx.c
>>>>>>>> +++ b/drivers/mmc/fsl_esdhc_imx.c
>>>>>>>> @@ -29,6 +29,8 @@
>>>>>>>> #include <dm.h>
>>>>>>>> #include <asm-generic/gpio.h>
>>>>>>>> #include <dm/pinctrl.h>
>>>>>>>> +#include <dt-structs.h>
>>>>>>>> +#include <mapmem.h>
>>>>>>>>
>>>>>>>> #if !CONFIG_IS_ENABLED(BLK)
>>>>>>>> #include "mmc_private.h"
>>>>>>>> @@ -98,6 +100,11 @@ struct fsl_esdhc {
>>>>>>>> };
>>>>>>>>
>>>>>>>> struct fsl_esdhc_plat {
>>>>>>>> +#if CONFIG_IS_ENABLED(OF_PLATDATA)
>>>>>>>> + /* Put this first since driver model will copy the data here */
>>>>>>>> + struct dtd_fsl_imx6q_usdhc dtplat;
>>>>>>>> +#endif
>>>>>>>> +
>>>>>>>> struct mmc_config cfg;
>>>>>>>> struct mmc mmc;
>>>>>>>> };
>>>>>>>> @@ -1377,14 +1384,18 @@ static int fsl_esdhc_probe(struct udevice *dev)
>>>>>>>> struct mmc_uclass_priv *upriv = dev_get_uclass_priv(dev);
>>>>>>>> struct fsl_esdhc_plat *plat = dev_get_platdata(dev);
>>>>>>>> struct fsl_esdhc_priv *priv = dev_get_priv(dev);
>>>>>>>> +#if !CONFIG_IS_ENABLED(OF_PLATDATA)
>>>>>>>> const void *fdt = gd->fdt_blob;
>>>>>>>> int node = dev_of_offset(dev);
>>>>>>>> + fdt_addr_t addr;
>>>>>>>> +#else
>>>>>>>> + struct dtd_fsl_imx6q_usdhc *dtplat = &plat->dtplat;
>>>>>>>> +#endif
>>>>>>>> struct esdhc_soc_data *data =
>>>>>>>> (struct esdhc_soc_data *)dev_get_driver_data(dev);
>>>>>>>> #if CONFIG_IS_ENABLED(DM_REGULATOR)
>>>>>>>> struct udevice *vqmmc_dev;
>>>>>>>> #endif
>>>>>>>> - fdt_addr_t addr;
>>>>>>>> unsigned int val;
>>>>>>>> struct mmc *mmc;
>>>>>>>> #if !CONFIG_IS_ENABLED(BLK)
>>>>>>>> @@ -1392,14 +1403,23 @@ static int fsl_esdhc_probe(struct udevice *dev)
>>>>>>>> #endif
>>>>>>>> int ret;
>>>>>>>>
>>>>>>>> +#if CONFIG_IS_ENABLED(OF_PLATDATA)
>>>>>>>> + priv->esdhc_regs = map_sysmem(dtplat->reg[0], dtplat->reg[1]);
>>>>>>>> + val = plat->dtplat.bus_width;
>>>>>>>> + if (val == 8)
>>>>>>>> + priv->bus_width = 8;
>>>>>>>> + else if (val == 4)
>>>>>>>> + priv->bus_width = 4;
>>>>>>>> + else
>>>>>>>> + priv->bus_width = 1;
>>>>>>>> + priv->non_removable = 1;
>>>>>>>> +#else
>>>>>>>> addr = dev_read_addr(dev);
>>>>>>>> if (addr == FDT_ADDR_T_NONE)
>>>>>>>> return -EINVAL;
>>>>>>>> priv->esdhc_regs = (struct fsl_esdhc *)addr;
>>>>>>>> priv->dev = dev;
>>>>>>>> priv->mode = -1;
>>>>>>>> - if (data)
>>>>>>>> - priv->flags = data->flags;
>>>>>>>>
>>>>>>>> val = dev_read_u32_default(dev, "bus-width", -1);
>>>>>>>> if (val == 8)
>>>>>>>> @@ -1462,7 +1482,9 @@ static int fsl_esdhc_probe(struct udevice *dev)
>>>>>>>> priv->vs18_enable = 1;
>>>>>>>> }
>>>>>>>> #endif
>>>>>>>> -
>>>>>>>> +#endif
>>>>>>>> + if (data)
>>>>>>>> + priv->flags = data->flags;
>>>>>>>> /*
>>>>>>>> * TODO:
>>>>>>>> * Because lack of clk driver, if SDHC clk is not enabled,
>>>>>>>> @@ -1513,9 +1535,11 @@ static int fsl_esdhc_probe(struct udevice *dev)
>>>>>>>> return ret;
>>>>>>>> }
>>>>>>>>
>>>>>>>> +#if !CONFIG_IS_ENABLED(OF_PLATDATA)
>>>>>>> Maybe can use if() for this one?
>>>>>> Thank you for the suggestion
>>>>>>
>>>>>>>> ret = mmc_of_parse(dev, &plat->cfg);
>>>>>>>> if (ret)
>>>>>>>> return ret;
>>>>>>>> +#endif
>>>>>>>>
>>>>>>>> mmc = &plat->mmc;
>>>>>>>> mmc->cfg = &plat->cfg;
>>>>>>>> @@ -1648,4 +1672,18 @@ U_BOOT_DRIVER(fsl_esdhc) = {
>>>>>>>> .platdata_auto_alloc_size = sizeof(struct fsl_esdhc_plat),
>>>>>>>> .priv_auto_alloc_size = sizeof(struct fsl_esdhc_priv),
>>>>>>>> };
>>>>>>>> +
>>>>>>>> +#if CONFIG_IS_ENABLED(OF_PLATDATA)
>>>>>>> Don't you already have a U_BOOT_DRIVER declaration?
>>>>>>>
>>>>>>> You may need to change the name of your existing driver though (see
>>>>>>> of-platdata docs).
>>>>>>>
>>>>>>> So if it is because of that, please add a comment.
>>>>>> I have my doubts regarding this issue. As I see, this driver is used by
>>>>>> many different DT with different compatible strings, so I'm thinking in
>>>>>> trying to find a more generic approach. Would it be useful to have a
>>>>>> more elaborated solution searching for the compatible string when
>>>>>> matching drivers with device?
>>>>> Yes I think so.
>>>>>
>>>>> Actually searching for a string is not great anyway. I wonder if we
>>>>> can use the linker-list idea in the previous email somehow?
>>>> I'm sure I' understand the idea you try to share with me. Sorry, I
>>>> understand that one limitation of the current implementation of
>>>> OF_PLATDATA is the fact that the U_BOOT_DRIVER name should match the one
>>>> used as first entry in DT. As in particular this driver has several
>>>> compatible
>>>>
>>>> { .compatible = "fsl,imx53-esdhc", },
>>>> { .compatible = "fsl,imx6ul-usdhc", },
>>>> { .compatible = "fsl,imx6sx-usdhc", },
>>>> { .compatible = "fsl,imx6sl-usdhc", },
>>>> { .compatible = "fsl,imx6q-usdhc", },
>>>> { .compatible = "fsl,imx7d-usdhc", .data =
>>>> (ulong)&usdhc_imx7d_data,},
>>>> { .compatible = "fsl,imx7ulp-usdhc", },
>>>> { .compatible = "fsl,imx8qm-usdhc", .data =
>>>> (ulong)&usdhc_imx8qm_data,},
>>>> { .compatible = "fsl,imx8mm-usdhc", .data =
>>>> (ulong)&usdhc_imx8qm_data,},
>>>> { .compatible = "fsl,imx8mn-usdhc", .data =
>>>> (ulong)&usdhc_imx8qm_data,},
>>>> { .compatible = "fsl,imx8mq-usdhc", .data =
>>>> (ulong)&usdhc_imx8qm_data,},
>>>> { .compatible = "fsl,imxrt-usdhc", },
>>>>
>>>> and in order to create a general solution, we need to search for the
>>>> compatible string instead of matching for driver name.
>>>>
>>>> Could you please elaborate a bit more your suggestion in order to
>>>> understand your approach.
>>>>
>>>> Thanks in advance.
>>> I am wondering if we can use the DM_GET_DRIVER() macro somehow in
>>> dt_platdata.c. At present we emit a string, and that string matches
>>> the driver name, so we should be able to. That will give a compile
>>> error if something is wrong, much better than the current behaviour of
>>> not being able to bind the driver at runtime.
>>>
>>> This is just to improve the current impl, not to do what you are asking here.
>>>
>>> For what you want, our current approach is to use the first compatible
>>> string to find the driver name. Since all compatible strings point to
>>> the same driver, perhaps that is good enough? We should at least
>>> understand the limitations before going further.
>>>
>>> The main one I am aware of is that you need to duplicate the
>>> U_BOOT_DRIVER()xxx for each compatible string. To fix that we could
>>> add:
>>>
>>> U_BOOT_DRIVER_ALIAS(xxx, new_name)
>>>
>>> which creates a linker list symbol that points to the original driver,
>>> perhaps using ll_entry_get(). That should be easy enough I think. Then
>>> whichever symbol you use you will get the same driver since all the
>>> symbols point to it.
>>>
>>> Unfortunately the .data field is not available with of_platdata. That
>>> could be added to the dtd_... struct automatically by dtoc, I suspect.
>>> However that requires some clever parsing of the C code...
>>>
>>> As you can tell I would really like to avoid string comparisons and
>>> tables of compatible strings in the image itself. It adds overheade.
>>
>> Thanks for taking the time to elaborate your idea, now is clear. I
>> totally agree with you, the whole idea behind it to reduce the image
>> size, so we need to work to avoid any kind of table of strings.
>>
>>
>> I will investigate you approach and come back to you.
> OK sounds good. I should mention that the dtoc tool should be
> upstreamed to dtc. I was thinking of sending something very simple to
> start.
I have been thinking in your suggestions and the main goal behind all
these. With that in mind I think that the best approach we can use is to
move as much complexity as we can to dtoc, which will give us
- Reduction in TPL/SPL image
- Warnings at compile time
So I was thinking in add the support for aliases to doc as
- Add U_BOOT_DRIVER_ALIAS(name, new_name) in drivers which generate no
code for U-Boot
- Parse U_BOOT_DRIVER(name) to build a list of drivers in dtoc
- Parse U_BOOT_DRIVER_ALIAS(name, new_name) to populate the list of
drivers with aliases in dtoc
- When parsing dts file, look for the proper driver name based on the
list created, is no one is found raise an error
I think the parser could be simple, something like (this is only a draft
to show the idea)
#!/usr/bin/python
import os
import re
class UBootDriverList:
def parse_u_boot_driver(self, fn):
f = open(fn)
b = f.read()
drivers = re.findall('U_BOOT_DRIVER\((.*)\)', b)
for d in drivers:
self.driver_list.append(d)
def parse_u_boot_drivers(self):
self.driver_list = []
for (dirpath, dirnames, filenames) in os.walk('./'):
for fn in filenames:
if not fn.endswith('.c'):
continue
self.parse_u_boot_driver(dirpath + '/' + fn)
def print_list(self):
for d in self.driver_list:
print d
driver_list = UBootDriverList()
driver_list.parse_u_boot_drivers()
driver_list.print_list()
What do you think?
If this make sense we can try to go also in this way for other improvements.
Regards,
Walter
More information about the U-Boot
mailing list