On Wed, 27 May 2020 at 10:00, Pali Rohár <pali@xxxxxxxxxx> wrote: > > On Wednesday 27 May 2020 09:39:50 Ulf Hansson wrote: > > On Tue, 26 May 2020 at 17:43, Pali Rohár <pali@xxxxxxxxxx> wrote: > > > > > > Device/vendor ids from Common CIS (Card Information Structure) may be > > > different as device/vendor ids from CIS on particular SDIO function. > > > > > > Export these "main" device/vendor ids for SDIO and SD combo cards at top > > > level mmc device in sysfs so userspace can do better identification of > > > connected SDIO and SD combo cards. > > > > What would userspace do with this information, more exactly? > > Userspace can e.g. write udev rules based on Common CIS vendor/device > id. Or can exactly identify SDIO card by CIS vendor/device id. Also it > can be suitable for "lssdio" tool to print all information about SDIO > card. > > Currently I do not know any way how userspace can retrieve these ids for > particular SDIO card. And correct identification is important. > > Other important thing is that kernel on some places (e.g. mmc quirks) > uses Common CIS vendor/device id and on other places (e.g. binding > drivers) it uses SDIO function device/vendor ids. > > So for debugging and developing kernel drivers it is needed to know > correct Common CIS vendor/device id and SDIO functions vendor/device > ids. > > > Isn't it just sufficient to give events per SDIO func, as we already > > do in sdio_bus_uevent()? > > No because some SDIO cards have different Common CIS vendor/device id > and different vendor/device ids for particular SDIO functions. > > Common CIS vendor/device id is the "main" identification of SDIO card, > functions vendor/device ids just identify one of those functions. > > For example look at my patch "mmc: sdio: Fix macro name for Marvell > device with ID 0x9134" [1]. Without knowing correct CIS vendor/device id > I was not able to fully understand problem and mess with names and ids. > Because mmc quirks list uses CIS vendor/device ids (I guess for obvious > reason as SDIO functions are not enumerated yet). Good points, much appreciated to understand the use cases better! May I suggest that you put some of this information into the commit message? > > [1] - https://lore.kernel.org/linux-mmc/20200522144412.19712-2-pali@xxxxxxxxxx/ > [...] Kind regards Uffe
