Re: [PATCH v1 3/4] firmware loader: allow distribution to choose default search paths

[Greg Kroah-Hartman <gregkh@linuxfoundation.org>]

On Fri, Jun 07, 2013 at 09:24:25AM +0800, Ming Lei wrote:
> On Fri, Jun 7, 2013 at 3:47 AM, Greg Kroah-Hartman
> <gregkh@linuxfoundation.org> wrote:
> > On Thu, Jun 06, 2013 at 08:01:49PM +0800, Ming Lei wrote:
> >> For some distributions(e.g. android), firmware images aren't put
> >> under kernel built-in search paths, so introduce one Kconfig
> >> option to allow distributions or users to choose its specific default
> >> search paths, which are always tried before searching from kernel
> >> built-in paths in direct loading.
> >>
> >> Also this patch introduces fw_get_fw_file_from_paths to cover all
> >> search paths, and fw_get_filesystem_firmware is simpified a bit.
> >>
> >> Cc: Takashi Iwai <tiwai@suse.de>
> >> Signed-off-by: Ming Lei <ming.lei@canonical.com>
> >> ---
> >>  drivers/base/Kconfig          |   14 ++++++++++
> >>  drivers/base/firmware_class.c |   61 ++++++++++++++++++++++++++++++++++-------
> >>  2 files changed, 65 insertions(+), 10 deletions(-)
> >>
> >> diff --git a/drivers/base/Kconfig b/drivers/base/Kconfig
> >> index 07abd9d..2be10e4 100644
> >> --- a/drivers/base/Kconfig
> >> +++ b/drivers/base/Kconfig
> >> @@ -156,6 +156,20 @@ config FW_LOADER_USER_HELPER
> >>         no longer required unless you have a special firmware file that
> >>         resides in a non-standard path.
> >>
> >> +config FW_CUSTOMIZED_PATH
> >> +     string "default firmware search paths for direct loading"
> >> +     help
> >> +       On some distribution(e.g. android), firmware images aren't
> >> +       put under kernel built-in search paths, so provide this option
> >> +       for distributions to choose a distribution specific firmware
> >> +       search path. The option allows to choose more than one path,
> >> +       and paths are seperated with colon like $PATH(e.g. on android,
> >> +       the option might look as "/etc/firmware:/vendor/firmware").
> >> +       Each path should be a absolute path, and relative path will be
> >> +       ignored.
> >> +
> >> +       If you are unsure about this, don't choose here.
> >> +
> >>  config DEBUG_DRIVER
> >>       bool "Driver Core verbose debug messages"
> >>       depends on DEBUG_KERNEL
> >> diff --git a/drivers/base/firmware_class.c b/drivers/base/firmware_class.c
> >> index 6ede229..051db83 100644
> >> --- a/drivers/base/firmware_class.c
> >> +++ b/drivers/base/firmware_class.c
> >> @@ -265,8 +265,13 @@ static void fw_free_buf(struct firmware_buf *buf)
> >>
> >>  /* direct firmware loading support */
> >>  static char fw_path_para[256];
> >> +
> >> +/* search runtime paths first, then static pre-defined paths */
> >>  static const char * const fw_path[] = {
> >>       fw_path_para,
> >> +#ifdef CONFIG_FW_CUSTOMIZED_PATH
> >> +     CONFIG_FW_CUSTOMIZED_PATH,
> >> +#endif
> >>       "/lib/firmware/updates/" UTS_RELEASE,
> >>       "/lib/firmware/updates",
> >>       "/lib/firmware/" UTS_RELEASE,
> >> @@ -314,6 +319,50 @@ static bool fw_read_file_contents(struct file *file, struct firmware_buf *fw_buf
> >>       return true;
> >>  }
> >>
> >> +static bool fw_get_file_firmware(const char *path,
> >> +                              struct firmware_buf *buf)
> >> +{
> >> +     struct file *file;
> >> +     bool success;
> >> +
> >> +     file = filp_open(path, O_RDONLY, 0);
> >> +     if (IS_ERR(file))
> >> +             return false;
> >> +     success = fw_read_file_contents(file, buf);
> >> +     fput(file);
> >> +
> >> +     return success;
> >> +}
> >> +
> >> +/* The path in @paths is seperated by ';' */
> >
> > No it isn't.
> 
> Will fix it.
> 
> >
> >> +static bool fw_get_file_fw_from_paths(const char *paths, char *path,
> >> +                                   struct firmware_buf *buf)
> >> +{
> >> +     int len, start, end = -1;
> >> +     char *pos;
> >> +
> >> +     do {
> >> +             start = end + 1;
> >> +             pos = strchr(&paths[start], ':');
> >
> > As you have an array of paths now, why are you doing the ':' check
> > still?  Don't do that, just allow the person building the kernel to add
> > one path to the loader, that should be all that we need, right?
> 
> The distribution may have put firmware under more than one places,
> for example, android put firmwares under "/etc/firmware" and
> "/vendor/firmware", so I am afraid that only allowing one path or two
> paths isn't flexible enough.

How about we wait until someone complains about this?  Don't add
features that aren't needed, especially complex ones like this.

thanks,

greg k-h