For bus-based driver, device removal is implemented as:
device_remove() => bus->remove() => driver->remove()
Driver core needs no feedback from bus driver about the result of
remove callback. In which case, commit fc7a6209d571 ("bus: Make
remove callback return void") forces bus_type::remove be void-returned.
Now we have the situation that both 1st & 2nd part of calling chain
are void returned, so it does not make much sense for the last one
(driver->remove) to return non-void to its caller.
So the basic idea behind this patchset is making remove() callback of
any bus-based driver to be void returned.
This patchset includes changes for drivers below:
1. hyperv
2. macio
3. apr
4. xen
5. ac87
6. soundbus
Q: Why not platform drivers?
A: Too many of them.(maybe 4K+)
Dawei Li (6):
hyperv: Make remove callback of hyperv driver void returned
macio: Make remove callback of macio driver void returned
apr: make remove callback of apr driver void returned
xen: make remove callback of xen driver void returned
ac97: make remove callback of ac97 driver void returned
soundbus: make remove callback of soundbus driver void returned
arch/powerpc/include/asm/macio.h | 12 ++++++------
drivers/ata/pata_macio.c | 4 +---
drivers/block/xen-blkback/xenbus.c | 4 +---
drivers/block/xen-blkfront.c | 3 +--
drivers/char/tpm/xen-tpmfront.c | 3 +--
drivers/gpu/drm/hyperv/hyperv_drm_drv.c | 4 +---
drivers/gpu/drm/xen/xen_drm_front.c | 3 +--
drivers/hid/hid-hyperv.c | 4 +---
drivers/hv/hv_balloon.c | 5 +----
drivers/hv/hv_util.c | 4 +---
drivers/input/misc/xen-kbdfront.c | 5 ++---
drivers/input/serio/hyperv-keyboard.c | 4 +---
drivers/macintosh/rack-meter.c | 4 +---
drivers/mfd/wm97xx-core.c | 4 +---
drivers/net/ethernet/apple/bmac.c | 4 +---
drivers/net/ethernet/apple/mace.c | 4 +---
drivers/net/hyperv/netvsc_drv.c | 4 +---
drivers/net/wireless/intersil/orinoco/airport.c | 4 +---
drivers/net/xen-netback/xenbus.c | 3 +--
drivers/net/xen-netfront.c | 4 +---
drivers/pci/controller/pci-hyperv.c | 3 +--
drivers/pci/xen-pcifront.c | 4 +---
drivers/scsi/mac53c94.c | 5 +----
drivers/scsi/mesh.c | 5 +----
drivers/scsi/storvsc_drv.c | 4 +---
drivers/scsi/xen-scsifront.c | 4 +---
drivers/tty/hvc/hvc_xen.c | 4 ++--
drivers/tty/serial/pmac_zilog.c | 7 ++-----
drivers/uio/uio_hv_generic.c | 5 ++---
drivers/usb/host/xen-hcd.c | 4 +---
drivers/video/fbdev/hyperv_fb.c | 5 +----
drivers/video/fbdev/xen-fbfront.c | 6 ++----
drivers/xen/pvcalls-back.c | 3 +--
drivers/xen/pvcalls-front.c | 3 +--
drivers/xen/xen-pciback/xenbus.c | 4 +---
drivers/xen/xen-scsiback.c | 4 +---
include/linux/hyperv.h | 2 +-
include/linux/soc/qcom/apr.h | 2 +-
include/sound/ac97/codec.h | 6 +++---
include/xen/xenbus.h | 2 +-
net/9p/trans_xen.c | 3 +--
net/vmw_vsock/hyperv_transport.c | 4 +---
sound/ac97/bus.c | 5 ++---
sound/aoa/fabrics/layout.c | 3 +--
sound/aoa/soundbus/i2sbus/core.c | 4 +---
sound/aoa/soundbus/soundbus.h | 6 +++---
sound/soc/qcom/qdsp6/q6core.c | 4 +---
sound/xen/xen_snd_front.c | 3 +--
48 files changed, 63 insertions(+), 137 deletions(-)
--
2.25.1
On Mon, Dec 05, 2022 at 11:36:38PM +0800, Dawei Li wrote:
> For bus-based driver, device removal is implemented as:
> device_remove() => bus->remove() => driver->remove()
>
> Driver core needs no feedback from bus driver about the result of
> remove callback. In which case, commit fc7a6209d571 ("bus: Make
> remove callback return void") forces bus_type::remove be void-returned.
>
> Now we have the situation that both 1st & 2nd part of calling chain
> are void returned, so it does not make much sense for the last one
> (driver->remove) to return non-void to its caller.
>
> So the basic idea behind this patchset is making remove() callback of
> any bus-based driver to be void returned.
>
> This patchset includes changes for drivers below:
> 1. hyperv
> 2. macio
> 3. apr
> 4. xen
> 5. ac87
> 6. soundbus
Then that should be 6 different patchsets going to 6 different
subsystems. No need to make this seems like a unified set of patches at
all.
> Q: Why not platform drivers?
> A: Too many of them.(maybe 4K+)
That will have to be done eventually, right?
thanks,
greg k-h
On Mon, Dec 05, 2022 at 05:00:52PM +0100, Greg KH wrote:
> On Mon, Dec 05, 2022 at 11:36:38PM +0800, Dawei Li wrote:
> > For bus-based driver, device removal is implemented as:
> > device_remove() => bus->remove() => driver->remove()
> >
> > Driver core needs no feedback from bus driver about the result of
> > remove callback. In which case, commit fc7a6209d571 ("bus: Make
> > remove callback return void") forces bus_type::remove be void-returned.
> >
> > Now we have the situation that both 1st & 2nd part of calling chain
> > are void returned, so it does not make much sense for the last one
> > (driver->remove) to return non-void to its caller.
> >
> > So the basic idea behind this patchset is making remove() callback of
> > any bus-based driver to be void returned.
> >
> > This patchset includes changes for drivers below:
> > 1. hyperv
> > 2. macio
> > 3. apr
> > 4. xen
> > 5. ac87
> > 6. soundbus
Hi Greg:
Thanks for the reviewing.
>
> Then that should be 6 different patchsets going to 6 different
> subsystems. No need to make this seems like a unified set of patches at
> all.
Right, will fix all the issues for this patchset and resend them in 6
independent patches.
Thanks
Dawei
>
> > Q: Why not platform drivers?
> > A: Too many of them.(maybe 4K+)
>
> That will have to be done eventually, right?
>
> thanks,
>
> greg k-h