Date: Wed, 5 Mar 2014 16:56:02 -0500 From: Patrick Kelsey <kelsey@ieee.org> To: Warner Losh <imp@bsdimp.com> Cc: "freebsd-hackers@freebsd.org" <freebsd-hackers@freebsd.org>, freebsd-arm <freebsd-arm@freebsd.org>, freebsd-embedded@freebsd.org, Ian Lepore <ian@freebsd.org> Subject: Re: [PATCH] simplebus child device probe order control via FDT (motivated by BeagleBone Black) Message-ID: <CAD44qMU46nqsw155qqKs7DrxXqawURgaUVEAaetzDExYM2LhYg@mail.gmail.com> In-Reply-To: <16A5203B-B06D-4129-A54F-F34D5FA28D2B@bsdimp.com> References: <CAD44qMUyqzaFtjgXdgThgHcHjPctx-oZAdhvHp4Kf0G7N4HVog@mail.gmail.com> <7C2B7036-51CC-4C97-80C4-0A439874357D@bsdimp.com> <CAD44qMXe8bh0cR60tm8%2BLZ9W3WhJDuGX6xz9FLNHrzmXNd5FDQ@mail.gmail.com> <1393939277.1149.300.camel@revolution.hippie.lan> <CAD44qMWLox0yY1-%2B2bn4dQ=z0jWKow95b=mnCJEkwmSiSipf4g@mail.gmail.com> <438620C4-7712-4B01-A46F-CB57946A30BF@bsdimp.com> <CAD44qMWNGPY70NMkTWoWcMTf1pywhSPGdbJpmbxxSxEdENSOJQ@mail.gmail.com> <16A5203B-B06D-4129-A54F-F34D5FA28D2B@bsdimp.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, Mar 5, 2014 at 4:24 PM, Warner Losh <imp@bsdimp.com> wrote:
>
> On Mar 5, 2014, at 11:55 AM, Patrick Kelsey <kelsey@ieee.org> wrote:
>
> >
> >
> >
> > On Wed, Mar 5, 2014 at 8:31 AM, Warner Losh <imp@bsdimp.com> wrote:
> >
> > On Mar 4, 2014, at 11:53 PM, Patrick Kelsey <kelsey@ieee.org> wrote:
> >
> > >
> > >
> > >
> > > On Tue, Mar 4, 2014 at 8:21 AM, Ian Lepore <ian@freebsd.org> wrote:
> > >
> > > There's a standard property for mmc/sd devices, "non-removable" whose
> > > presence denotes things like soldered-on eMMC parts. Only one of our
> > > many sdhci drivers supports it right now (imx). In general the core
> > > part of the driver (dev/sdhci) doesn't have good support for
> > > insert/remove/presence detection that's handled off to the side
> (whether
> > > it's non-removable or a gpio pin).
> > >
> > > That alone doesn't solve the wider problem, though, which I think
> breaks
> > > down into two pieces. Let's say you've got two slots, call them left
> > > and right. You end up with these two problems...
> > >
> > > * Sometimes the right slot is mmcsd0, but it turns into mmcsd1 if
> > > there's a card in the left slot when I boot; I don't want it to
> > > change.
> > >
> > > And not just a boot-time issue, of course. If you were to remove
> those two cards and then reinsert them in the opposite time-order, their
> device names would swap.
> > >
> > > * I want the slot on the left to be named '1' and the right to be
> > > '0'.
> > >
> > > The first problem is easily solved without impacting dts in any way.
> We
> > > just wire down the relationship controllerN -> mmcN -> mmcsdN. This is
> > > exactly equivelent to the old ATA_STATIC_ID option in its effect -- you
> > > don't get to choose the names, but you know they won't change based on
> > > which devices are present. It could be controlled with a tunable.
> > >
> > > It's harder to envision the fix for the second part without adding an
> > > ad-hoc property for the devices. Even with a property I'm not sure how
> > > easy it would be.
> > >
> > > We should be able to assign a geographic address (controllerX:slotY)
> to each mmcsd device in a given system (let's ignore for now the
> theoretical possibility of multiple cards on one bus). The 'controllerX'
> part of the address could be the controller's pathname from the devicetree,
> or an index assigned by mmcbr machinery at attach time. The "slotY" part
> of the address is assigned by the specific controller device driver using
> some internally-determined fixed mapping between the assigned values and
> its physical slots. This geographic address could be used to create an
> additional set of /dev entries with stable names. There could be a
> mechanism for user-configurable aliases for the geographical addresses.
> >
> > There's already a chance to run a script when a device is attached that
> can create /dev/slot0 or /dev/slot1 that has geographical information
> available to it. People use ddvd for this in the USB world all the time to
> make sure that tty devices get a symlink based on their location or serial
> number.
> >
> > Why is mmc so different it needs its own mechanism?
> >
> > I'm laying out my view of the information that would be needed and the
> types of actions that would have to be taken based on that information to
> solve the issue. I'm not saying devd can't be the piece that is used to
> implement the actions (indeed, I noted devd as a potential building block
> for a solution in my initial response). I'm also not saying mmc needs its
> own mechanism, I'm saying it needs /a/ mechanism, and so far there still
> seems to be something missing (because it's not there, or I'm still
> ignorant of it).
> >
> > What seems to be missing is geographical addressing for mmc devices.
> >
> > I think what you might be saying is that the existing mmcsd add/remove
> code could be augmented to send devctl notifications, along the lines of:
> >
> > devctl_notify("MMC", "DEVICE", "ATTACH"|"DETACH", "...
> controller=<controller_device_name_or_better_yet_devicetree_path>
> slot=<slot_number> rca=<rca>")
> >
> > and then I and the fine author of devctl and devd would be pleased.
>
> MMC doesn't need anything special here. That's already present. Looking at
> the device tree we see on one of the platforms that's handy for me to
> access:
>
> at91_mci0
> mmc0
> mmcsd0 at rca=0xb368
>
> So you know that mmcsd0 is on mmc0 bus attached to at91_mci0, which is
> ultimately the FDT node where things came from. There's not a user-defined
> slot associated with this (and we should have a SLOT A vs SLOT B as a
> location info for this platform, because we can have two cards on the one
> bus in the MMC case), true, but for your use case I don't think that you
> need it. We should be attaching the host controller regardless of whether
> the or not there's a card in there, so it will be fixed. While some
> additional information would be useful to publish, I don't think your use
> case requires it...
>
>
The reason you need something extra here is that what is there now breaks
down whenever you don't have a one-to-one mapping between controllers and
buses. Any controller with more than one slot can yield something of the
form:
sdhci0
mmc0
mmcsd0 at rca=0xabcd
mmc1
mmcsd1 at rca=0x1234
and you have no idea what physical slot in the system mmcsd0 and mmcsd1 are
in.
For my immediate use case, sure, I can rely on the one-to-one relationship
between controllers and buses. At this point, though, rather than skate by
on that happy coincidence, I'd rather invest what now seems to be a rather
small amount of effort adding mmcsd devctl notifications that would cover
the multiple-slots-per-controller case as well, and then build the rest of
what I want on top of that information coming out of devd.
On the at91 platform cited above, that allows you to connect two MMC cards
to the same bus (i.e., multidrop configuration), is there any way to
distinguish which card is in which physical slot? I'm still under the
impression that this is the one case where we aren't going to be able to
determine the physical location of every mmcsd device.
-Patrick
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAD44qMU46nqsw155qqKs7DrxXqawURgaUVEAaetzDExYM2LhYg>