[UPSTREAM 0/2] pcf50633 changes

Andy Green andy at openmoko.com
Mon Nov 17 09:13:17 CET 2008


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Somebody in the thread at some point said:
| On Sun, Nov 16, 2008 at 08:56:10PM +0000, Andy Green wrote:
|> -----BEGIN PGP SIGNED MESSAGE-----
|> Hash: SHA1
|>
|> Somebody in the thread at some point said:
|> | Hi,
|> |
|> |
|> | Here's a pcf50633 driver rewrite to use the MFD model. This has
resulted
|> | in pretty large number of changes.
|>
|> Some warnings came out:
|>
|> drivers/mfd/pcf50633-core.c:119: warning: initialization from
|> incompatible pointer type
|>
|> drivers/mfd/pcf50633-core.c: In function ‘pcf50633_suspend’:
|>
|> drivers/mfd/pcf50633-core.c:266: warning: unused variable ‘ret’
|>
|> drivers/mfd/pcf50633-mbc.c:88: warning: initialization from incompatible
|> pointer type
|>
|
| Yea, I've seen these things. Will fix them.

OK.  Giving birth to these big patches is so exhausting it's normal that
there's stuff to clean.

|> | 4. Emergency 8s shutdown has been temporarily removed from the driver
|> | as it certainly does not belong there.
|>
|> On pcf50606 this was a hardware feature: it is on pcf50633 too but it's
|> disabled on our variant.  So it is arguable that reproducing it belongs
|> in the driver.
|>
| Our way of initiating a powerdown is not fool proof, The IRQ worker may
| fail to run a all if events is blocked. But it's still better than
| nothing. We could check for the revision and use it only where don't
| have a better option.

Right, IRQ lockout in a panic for example can also mean we can't even
detect power button situation.  It's a best effort to allow the user to
stop the phone quickly, it's not so good as the hardware implementation,
but it should mostly be better than the nothing that we otherwise have.

|> | 5. 'Suppress onkey events on resume' - must be a better way to handle
|> | this. Can't it be done from userspace ?
|>
|> Dunno where the issue comes that it suppresses the notifcation for, if
|> it is opowerd controlled it should know it is coming from wake.

mwester explained the logic, input subsystem takes care to echo all
events to all open file descriptors... does touchscreen also suppress
wake from blanking click?  One wouldn't want the tap taken seriously
there either when waking from blanking by blindly tapping it.

|> | 6. Resume reason - Removed currenty. More discussion required on format
|> | of sysfs attribute.
|>
|> We have to capture the first resume interrupt source bytes in this
|> driver because they don't exist anywhere else.  They're currently
|> exposed into sysfs elsewhere, but that can be done here too and make the
|> resume reason parsing in two parts.
|
| Two parts as in ?

Firstly there's capture of CPU interrupt source register to identify the
CPU view of which wake source did the deed, then if we see it was PMU
interrupt that woke us, the userspace stuff can go look in a
PMU-specific /sys to see what interrupt sources were signalled on first
service of PMU after wake.  It's more scalable than current system
although it does need userspace to know magic sys paths per interrupt
source.

- -Andy
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org

iEYEARECAAYFAkkhJ50ACgkQOjLpvpq7dMrYrACfTHz6vzEDF6lLjiDGCMNpCczl
NhIAnAzXc7/DZgkxqASFd1sl3spouuCx
=RPXF
-----END PGP SIGNATURE-----



More information about the openmoko-kernel mailing list