| Message ID | 20191003092913.10731-1-brgl@bgdev.pl (mailing list archive) |
|---|---|
| Headers | show |
| Series | drivers: make early_platform code SuperH-specific | expand |
On Thu, Oct 03, 2019 at 11:29:11AM +0200, Bartosz Golaszewski wrote: > From: Bartosz Golaszewski <bgolaszewski@baylibre.com> > > Some time ago I started a discussion about the need for a proper early device > probing mechanism[1]. One that would be based on real platform drivers and > support both platform data and device tree. > > While we're far from reaching any consensus on the implementation, Arnd > suggested that I start off by moving the SuperH-specific early platform > drivers implementation to arch/sh[2]. > > This series is the first attempt at making way for a new, less hacky > implementation. > > The first patch moves all the early_platform code to arch/sh. > > The second patch prefixes all early_platform symbols with 'sh_'. > > [1] https://lkml.org/lkml/2018/4/26/657 > [2] https://lkml.org/lkml/2018/4/27/239 > > v1 -> v2: > - certain drivers are compiled for arm/mach-shmobile too - we need to > add ifdefs for CONFIG_SUPERH around early_platform calls > > v2 -> v3: > - added a stub for is_early_platform_device() which always returns false > on non-SuperH architectures > > v3 -> v4: > - rebased on top of v5.4-rc1 > - removed patches that are already upstream from the series > > Bartosz Golaszewski (2): > drivers: move the early platform device support to arch/sh > sh: add the sh_ prefix to early platform symbols I like this, any objection from anyone if I take this in my driver-core tree for 5.5-rc1? thanks, greg k-h
On Fri, Oct 04, 2019 at 03:00:31PM +0200, Greg Kroah-Hartman wrote: > On Thu, Oct 03, 2019 at 11:29:11AM +0200, Bartosz Golaszewski wrote: > > From: Bartosz Golaszewski <bgolaszewski@baylibre.com> > > > > Some time ago I started a discussion about the need for a proper early device > > probing mechanism[1]. One that would be based on real platform drivers and > > support both platform data and device tree. > > > > While we're far from reaching any consensus on the implementation, Arnd > > suggested that I start off by moving the SuperH-specific early platform > > drivers implementation to arch/sh[2]. > > > > This series is the first attempt at making way for a new, less hacky > > implementation. > > > > The first patch moves all the early_platform code to arch/sh. > > > > The second patch prefixes all early_platform symbols with 'sh_'. > > > > [1] https://lkml.org/lkml/2018/4/26/657 > > [2] https://lkml.org/lkml/2018/4/27/239 > > > > v1 -> v2: > > - certain drivers are compiled for arm/mach-shmobile too - we need to > > add ifdefs for CONFIG_SUPERH around early_platform calls > > > > v2 -> v3: > > - added a stub for is_early_platform_device() which always returns false > > on non-SuperH architectures > > > > v3 -> v4: > > - rebased on top of v5.4-rc1 > > - removed patches that are already upstream from the series > > > > Bartosz Golaszewski (2): > > drivers: move the early platform device support to arch/sh > > sh: add the sh_ prefix to early platform symbols > > I like this, any objection from anyone if I take this in my driver-core > tree for 5.5-rc1? I don't think I have any objection. It will probably make gratuitous merge conflicts with Sato-san's old device tree sh4 work when we get back to finishing that, but that's not really a big deal. Rich
On Fri, Oct 04, 2019 at 09:20:25AM -0400, Rich Felker wrote: > On Fri, Oct 04, 2019 at 03:00:31PM +0200, Greg Kroah-Hartman wrote: > > On Thu, Oct 03, 2019 at 11:29:11AM +0200, Bartosz Golaszewski wrote: > > > From: Bartosz Golaszewski <bgolaszewski@baylibre.com> > > > > > > Some time ago I started a discussion about the need for a proper early device > > > probing mechanism[1]. One that would be based on real platform drivers and > > > support both platform data and device tree. > > > > > > While we're far from reaching any consensus on the implementation, Arnd > > > suggested that I start off by moving the SuperH-specific early platform > > > drivers implementation to arch/sh[2]. > > > > > > This series is the first attempt at making way for a new, less hacky > > > implementation. > > > > > > The first patch moves all the early_platform code to arch/sh. > > > > > > The second patch prefixes all early_platform symbols with 'sh_'. > > > > > > [1] https://lkml.org/lkml/2018/4/26/657 > > > [2] https://lkml.org/lkml/2018/4/27/239 > > > > > > v1 -> v2: > > > - certain drivers are compiled for arm/mach-shmobile too - we need to > > > add ifdefs for CONFIG_SUPERH around early_platform calls > > > > > > v2 -> v3: > > > - added a stub for is_early_platform_device() which always returns false > > > on non-SuperH architectures > > > > > > v3 -> v4: > > > - rebased on top of v5.4-rc1 > > > - removed patches that are already upstream from the series > > > > > > Bartosz Golaszewski (2): > > > drivers: move the early platform device support to arch/sh > > > sh: add the sh_ prefix to early platform symbols > > > > I like this, any objection from anyone if I take this in my driver-core > > tree for 5.5-rc1? > > I don't think I have any objection. It will probably make gratuitous > merge conflicts with Sato-san's old device tree sh4 work when we get > back to finishing that, but that's not really a big deal. Ok, I've queued it up in my tree now, thanks, greg k-h
On Thu, Oct 03, 2019 at 11:29:12AM +0200, Bartosz Golaszewski wrote: > From: Bartosz Golaszewski <bgolaszewski@baylibre.com> > > SuperH is the only user of the current implementation of early platform > device support. We want to introduce a more robust approach to early > probing. As the first step - move all the current early platform code > to arch/sh. > > In order not to export internal drivers/base functions to arch code for > this temporary solution - copy the two needed routines for driver > matching from drivers/base/platform.c to arch/sh/drivers/platform_early.c. > > Also: call early_platform_cleanup() from subsys_initcall() so that it's > called after all early devices are probed. > > Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com> Wondering ... has anyone tested this patch on affected hardware ? All my qemu boot tests (both sh and sheb) fail because of it. Bisect log below. Guenter --- # bad: [72c0870e3a05d9cd5466d08c3d2a3069ed0a2f9f] Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input # good: [89d57dddd7d319ded00415790a0bb3c954b7e386] Merge tag 'media/v5.5-1' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media git bisect start 'HEAD' '89d57dddd7d3' # good: [0a6cad5df541108cfd3fbd79eef48eb824c89bdc] Merge branch 'vmwgfx-coherent' of git://people.freedesktop.org/~thomash/linux into drm-next git bisect good 0a6cad5df541108cfd3fbd79eef48eb824c89bdc # bad: [9a3d7fd275be4559277667228902824165153c80] Merge tag 'driver-core-5.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core git bisect bad 9a3d7fd275be4559277667228902824165153c80 # good: [59274c7164807d27b24e6c068dfe734f7bea4623] Merge tag 'usb-5.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb git bisect good 59274c7164807d27b24e6c068dfe734f7bea4623 # good: [e71903106721dc53923e90aa484d78bc86c039a9] staging: mt7621-dma: align to match open parenthesis git bisect good e71903106721dc53923e90aa484d78bc86c039a9 # good: [8f56e4ebe05c26c30e167519273843476e39e244] Merge tag 'char-misc-5.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc git bisect good 8f56e4ebe05c26c30e167519273843476e39e244 # good: [8bde9f3d2a217d1635a7c7bdf8ad4c25c9a34b50] Merge tag 'iio-for-5.5c' of https://git.kernel.org/pub/scm/linux/kernel/git/jic23/iio into staging-next git bisect good 8bde9f3d2a217d1635a7c7bdf8ad4c25c9a34b50 # bad: [0628cda318df6baec439ca6e6e274007492f1ccd] mac80211: Use debugfs_create_xul() helper git bisect bad 0628cda318df6baec439ca6e6e274007492f1ccd # bad: [03324507e66c7664c754b1ef92c5c3be24c78aa2] driver core: Allow fwnode_operations.add_links to differentiate errors git bisect bad 03324507e66c7664c754b1ef92c5c3be24c78aa2 # bad: [313f5dbba41d905d59c820bb2d91ee6c661aff99] debugfs: remove return value of debugfs_create_u16() git bisect bad 313f5dbba41d905d59c820bb2d91ee6c661aff99 # good: [d4387cd117414ba80230f27a514be5ca4a09cfcc] of: property: Create device links for all child-supplier depencencies git bisect good d4387cd117414ba80230f27a514be5ca4a09cfcc # bad: [c31e73121f4c1ec45a3e523ac6ce3ce6dafdcec1] base: soc: Handle custom soc information sysfs entries git bisect bad c31e73121f4c1ec45a3e523ac6ce3ce6dafdcec1 # bad: [201e91091b1d47047f55580b5474e1239f4d17aa] sh: add the sh_ prefix to early platform symbols git bisect bad 201e91091b1d47047f55580b5474e1239f4d17aa # bad: [507fd01d5333338753a1cc26322dfc9f856c109f] drivers: move the early platform device support to arch/sh git bisect bad 507fd01d5333338753a1cc26322dfc9f856c109f # first bad commit: [507fd01d5333338753a1cc26322dfc9f856c109f] drivers: move the early platform device support to arch/sh
pon., 2 gru 2019 o 05:14 Guenter Roeck <linux@roeck-us.net> napisał(a): > > On Thu, Oct 03, 2019 at 11:29:12AM +0200, Bartosz Golaszewski wrote: > > From: Bartosz Golaszewski <bgolaszewski@baylibre.com> > > > > SuperH is the only user of the current implementation of early platform > > device support. We want to introduce a more robust approach to early > > probing. As the first step - move all the current early platform code > > to arch/sh. > > > > In order not to export internal drivers/base functions to arch code for > > this temporary solution - copy the two needed routines for driver > > matching from drivers/base/platform.c to arch/sh/drivers/platform_early.c. > > > > Also: call early_platform_cleanup() from subsys_initcall() so that it's > > called after all early devices are probed. > > > > Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com> > > Wondering ... has anyone tested this patch on affected hardware ? > All my qemu boot tests (both sh and sheb) fail because of it. > Bisect log below. > > Guenter > > --- > # bad: [72c0870e3a05d9cd5466d08c3d2a3069ed0a2f9f] Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input > # good: [89d57dddd7d319ded00415790a0bb3c954b7e386] Merge tag 'media/v5.5-1' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media > git bisect start 'HEAD' '89d57dddd7d3' > # good: [0a6cad5df541108cfd3fbd79eef48eb824c89bdc] Merge branch 'vmwgfx-coherent' of git://people.freedesktop.org/~thomash/linux into drm-next > git bisect good 0a6cad5df541108cfd3fbd79eef48eb824c89bdc > # bad: [9a3d7fd275be4559277667228902824165153c80] Merge tag 'driver-core-5.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core > git bisect bad 9a3d7fd275be4559277667228902824165153c80 > # good: [59274c7164807d27b24e6c068dfe734f7bea4623] Merge tag 'usb-5.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb > git bisect good 59274c7164807d27b24e6c068dfe734f7bea4623 > # good: [e71903106721dc53923e90aa484d78bc86c039a9] staging: mt7621-dma: align to match open parenthesis > git bisect good e71903106721dc53923e90aa484d78bc86c039a9 > # good: [8f56e4ebe05c26c30e167519273843476e39e244] Merge tag 'char-misc-5.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc > git bisect good 8f56e4ebe05c26c30e167519273843476e39e244 > # good: [8bde9f3d2a217d1635a7c7bdf8ad4c25c9a34b50] Merge tag 'iio-for-5.5c' of https://git.kernel.org/pub/scm/linux/kernel/git/jic23/iio into staging-next > git bisect good 8bde9f3d2a217d1635a7c7bdf8ad4c25c9a34b50 > # bad: [0628cda318df6baec439ca6e6e274007492f1ccd] mac80211: Use debugfs_create_xul() helper > git bisect bad 0628cda318df6baec439ca6e6e274007492f1ccd > # bad: [03324507e66c7664c754b1ef92c5c3be24c78aa2] driver core: Allow fwnode_operations.add_links to differentiate errors > git bisect bad 03324507e66c7664c754b1ef92c5c3be24c78aa2 > # bad: [313f5dbba41d905d59c820bb2d91ee6c661aff99] debugfs: remove return value of debugfs_create_u16() > git bisect bad 313f5dbba41d905d59c820bb2d91ee6c661aff99 > # good: [d4387cd117414ba80230f27a514be5ca4a09cfcc] of: property: Create device links for all child-supplier depencencies > git bisect good d4387cd117414ba80230f27a514be5ca4a09cfcc > # bad: [c31e73121f4c1ec45a3e523ac6ce3ce6dafdcec1] base: soc: Handle custom soc information sysfs entries > git bisect bad c31e73121f4c1ec45a3e523ac6ce3ce6dafdcec1 > # bad: [201e91091b1d47047f55580b5474e1239f4d17aa] sh: add the sh_ prefix to early platform symbols > git bisect bad 201e91091b1d47047f55580b5474e1239f4d17aa > # bad: [507fd01d5333338753a1cc26322dfc9f856c109f] drivers: move the early platform device support to arch/sh > git bisect bad 507fd01d5333338753a1cc26322dfc9f856c109f > # first bad commit: [507fd01d5333338753a1cc26322dfc9f856c109f] drivers: move the early platform device support to arch/sh Hi Guenter, can you post some bootlogs? Is it the same problem everywhere? Bart
On 12/1/19 11:40 PM, Bartosz Golaszewski wrote: > pon., 2 gru 2019 o 05:14 Guenter Roeck <linux@roeck-us.net> napisał(a): >> >> On Thu, Oct 03, 2019 at 11:29:12AM +0200, Bartosz Golaszewski wrote: >>> From: Bartosz Golaszewski <bgolaszewski@baylibre.com> >>> >>> SuperH is the only user of the current implementation of early platform >>> device support. We want to introduce a more robust approach to early >>> probing. As the first step - move all the current early platform code >>> to arch/sh. >>> >>> In order not to export internal drivers/base functions to arch code for >>> this temporary solution - copy the two needed routines for driver >>> matching from drivers/base/platform.c to arch/sh/drivers/platform_early.c. >>> >>> Also: call early_platform_cleanup() from subsys_initcall() so that it's >>> called after all early devices are probed. >>> >>> Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com> >> >> Wondering ... has anyone tested this patch on affected hardware ? >> All my qemu boot tests (both sh and sheb) fail because of it. >> Bisect log below. >> >> Guenter >> >> --- >> # bad: [72c0870e3a05d9cd5466d08c3d2a3069ed0a2f9f] Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input >> # good: [89d57dddd7d319ded00415790a0bb3c954b7e386] Merge tag 'media/v5.5-1' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media >> git bisect start 'HEAD' '89d57dddd7d3' >> # good: [0a6cad5df541108cfd3fbd79eef48eb824c89bdc] Merge branch 'vmwgfx-coherent' of git://people.freedesktop.org/~thomash/linux into drm-next >> git bisect good 0a6cad5df541108cfd3fbd79eef48eb824c89bdc >> # bad: [9a3d7fd275be4559277667228902824165153c80] Merge tag 'driver-core-5.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core >> git bisect bad 9a3d7fd275be4559277667228902824165153c80 >> # good: [59274c7164807d27b24e6c068dfe734f7bea4623] Merge tag 'usb-5.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb >> git bisect good 59274c7164807d27b24e6c068dfe734f7bea4623 >> # good: [e71903106721dc53923e90aa484d78bc86c039a9] staging: mt7621-dma: align to match open parenthesis >> git bisect good e71903106721dc53923e90aa484d78bc86c039a9 >> # good: [8f56e4ebe05c26c30e167519273843476e39e244] Merge tag 'char-misc-5.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc >> git bisect good 8f56e4ebe05c26c30e167519273843476e39e244 >> # good: [8bde9f3d2a217d1635a7c7bdf8ad4c25c9a34b50] Merge tag 'iio-for-5.5c' of https://git.kernel.org/pub/scm/linux/kernel/git/jic23/iio into staging-next >> git bisect good 8bde9f3d2a217d1635a7c7bdf8ad4c25c9a34b50 >> # bad: [0628cda318df6baec439ca6e6e274007492f1ccd] mac80211: Use debugfs_create_xul() helper >> git bisect bad 0628cda318df6baec439ca6e6e274007492f1ccd >> # bad: [03324507e66c7664c754b1ef92c5c3be24c78aa2] driver core: Allow fwnode_operations.add_links to differentiate errors >> git bisect bad 03324507e66c7664c754b1ef92c5c3be24c78aa2 >> # bad: [313f5dbba41d905d59c820bb2d91ee6c661aff99] debugfs: remove return value of debugfs_create_u16() >> git bisect bad 313f5dbba41d905d59c820bb2d91ee6c661aff99 >> # good: [d4387cd117414ba80230f27a514be5ca4a09cfcc] of: property: Create device links for all child-supplier depencencies >> git bisect good d4387cd117414ba80230f27a514be5ca4a09cfcc >> # bad: [c31e73121f4c1ec45a3e523ac6ce3ce6dafdcec1] base: soc: Handle custom soc information sysfs entries >> git bisect bad c31e73121f4c1ec45a3e523ac6ce3ce6dafdcec1 >> # bad: [201e91091b1d47047f55580b5474e1239f4d17aa] sh: add the sh_ prefix to early platform symbols >> git bisect bad 201e91091b1d47047f55580b5474e1239f4d17aa >> # bad: [507fd01d5333338753a1cc26322dfc9f856c109f] drivers: move the early platform device support to arch/sh >> git bisect bad 507fd01d5333338753a1cc26322dfc9f856c109f >> # first bad commit: [507fd01d5333338753a1cc26322dfc9f856c109f] drivers: move the early platform device support to arch/sh > > Hi Guenter, > > can you post some bootlogs? Is it the same problem everywhere? > I don't have any useful logs. The boot fails too early for that. Sorry, I don't understand "everywhere". It fails in mainline and -next, if that is what you mean. And, yes, it has failed in -next for a while, looking back at the log history. Guenter
pon., 2 gru 2019 o 14:03 Guenter Roeck <linux@roeck-us.net> napisał(a): > > On 12/1/19 11:40 PM, Bartosz Golaszewski wrote: > > pon., 2 gru 2019 o 05:14 Guenter Roeck <linux@roeck-us.net> napisał(a): > >> > >> On Thu, Oct 03, 2019 at 11:29:12AM +0200, Bartosz Golaszewski wrote: > >>> From: Bartosz Golaszewski <bgolaszewski@baylibre.com> > >>> > >>> SuperH is the only user of the current implementation of early platform > >>> device support. We want to introduce a more robust approach to early > >>> probing. As the first step - move all the current early platform code > >>> to arch/sh. > >>> > >>> In order not to export internal drivers/base functions to arch code for > >>> this temporary solution - copy the two needed routines for driver > >>> matching from drivers/base/platform.c to arch/sh/drivers/platform_early.c. > >>> > >>> Also: call early_platform_cleanup() from subsys_initcall() so that it's > >>> called after all early devices are probed. > >>> > >>> Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com> > >> > >> Wondering ... has anyone tested this patch on affected hardware ? > >> All my qemu boot tests (both sh and sheb) fail because of it. > >> Bisect log below. > >> > >> Guenter > >> > >> --- > >> # bad: [72c0870e3a05d9cd5466d08c3d2a3069ed0a2f9f] Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input > >> # good: [89d57dddd7d319ded00415790a0bb3c954b7e386] Merge tag 'media/v5.5-1' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media > >> git bisect start 'HEAD' '89d57dddd7d3' > >> # good: [0a6cad5df541108cfd3fbd79eef48eb824c89bdc] Merge branch 'vmwgfx-coherent' of git://people.freedesktop.org/~thomash/linux into drm-next > >> git bisect good 0a6cad5df541108cfd3fbd79eef48eb824c89bdc > >> # bad: [9a3d7fd275be4559277667228902824165153c80] Merge tag 'driver-core-5.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core > >> git bisect bad 9a3d7fd275be4559277667228902824165153c80 > >> # good: [59274c7164807d27b24e6c068dfe734f7bea4623] Merge tag 'usb-5.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb > >> git bisect good 59274c7164807d27b24e6c068dfe734f7bea4623 > >> # good: [e71903106721dc53923e90aa484d78bc86c039a9] staging: mt7621-dma: align to match open parenthesis > >> git bisect good e71903106721dc53923e90aa484d78bc86c039a9 > >> # good: [8f56e4ebe05c26c30e167519273843476e39e244] Merge tag 'char-misc-5.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc > >> git bisect good 8f56e4ebe05c26c30e167519273843476e39e244 > >> # good: [8bde9f3d2a217d1635a7c7bdf8ad4c25c9a34b50] Merge tag 'iio-for-5.5c' of https://git.kernel.org/pub/scm/linux/kernel/git/jic23/iio into staging-next > >> git bisect good 8bde9f3d2a217d1635a7c7bdf8ad4c25c9a34b50 > >> # bad: [0628cda318df6baec439ca6e6e274007492f1ccd] mac80211: Use debugfs_create_xul() helper > >> git bisect bad 0628cda318df6baec439ca6e6e274007492f1ccd > >> # bad: [03324507e66c7664c754b1ef92c5c3be24c78aa2] driver core: Allow fwnode_operations.add_links to differentiate errors > >> git bisect bad 03324507e66c7664c754b1ef92c5c3be24c78aa2 > >> # bad: [313f5dbba41d905d59c820bb2d91ee6c661aff99] debugfs: remove return value of debugfs_create_u16() > >> git bisect bad 313f5dbba41d905d59c820bb2d91ee6c661aff99 > >> # good: [d4387cd117414ba80230f27a514be5ca4a09cfcc] of: property: Create device links for all child-supplier depencencies > >> git bisect good d4387cd117414ba80230f27a514be5ca4a09cfcc > >> # bad: [c31e73121f4c1ec45a3e523ac6ce3ce6dafdcec1] base: soc: Handle custom soc information sysfs entries > >> git bisect bad c31e73121f4c1ec45a3e523ac6ce3ce6dafdcec1 > >> # bad: [201e91091b1d47047f55580b5474e1239f4d17aa] sh: add the sh_ prefix to early platform symbols > >> git bisect bad 201e91091b1d47047f55580b5474e1239f4d17aa > >> # bad: [507fd01d5333338753a1cc26322dfc9f856c109f] drivers: move the early platform device support to arch/sh > >> git bisect bad 507fd01d5333338753a1cc26322dfc9f856c109f > >> # first bad commit: [507fd01d5333338753a1cc26322dfc9f856c109f] drivers: move the early platform device support to arch/sh > > > > Hi Guenter, > > > > can you post some bootlogs? Is it the same problem everywhere? > > > > I don't have any useful logs. The boot fails too early for that. > Yeah, it touched "early" devices after all... > Sorry, I don't understand "everywhere". It fails in mainline and -next, if that is what you mean. I referred to "All my qemu boot tests" when saying everywhere. I don't have any relevant HW - the idea for this patch is to stop compiling for everyone a bunch of code that's only used by one largely irrelevant architecture and free the namespace for a generic early platform drivers implementation. How are you creating your SH qemu images? I'm seeing there's something in buildroot for superh - maybe I'll be able to build something useful. Bartosz > And, yes, it has failed in -next for a while, looking back at the log history. > > Guenter
On Mon, Dec 02, 2019 at 06:18:22PM +0100, Bartosz Golaszewski wrote: > pon., 2 gru 2019 o 14:03 Guenter Roeck <linux@roeck-us.net> napisał(a): > > > > On 12/1/19 11:40 PM, Bartosz Golaszewski wrote: > > > pon., 2 gru 2019 o 05:14 Guenter Roeck <linux@roeck-us.net> napisał(a): > > >> > > >> On Thu, Oct 03, 2019 at 11:29:12AM +0200, Bartosz Golaszewski wrote: > > >>> From: Bartosz Golaszewski <bgolaszewski@baylibre.com> > > >>> > > >>> SuperH is the only user of the current implementation of early platform > > >>> device support. We want to introduce a more robust approach to early > > >>> probing. As the first step - move all the current early platform code > > >>> to arch/sh. > > >>> > > >>> In order not to export internal drivers/base functions to arch code for > > >>> this temporary solution - copy the two needed routines for driver > > >>> matching from drivers/base/platform.c to arch/sh/drivers/platform_early.c. > > >>> > > >>> Also: call early_platform_cleanup() from subsys_initcall() so that it's > > >>> called after all early devices are probed. > > >>> > > >>> Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com> > > >> > > >> Wondering ... has anyone tested this patch on affected hardware ? > > >> All my qemu boot tests (both sh and sheb) fail because of it. > > >> Bisect log below. > > >> > > >> Guenter > > >> > > >> --- > > >> # bad: [72c0870e3a05d9cd5466d08c3d2a3069ed0a2f9f] Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input > > >> # good: [89d57dddd7d319ded00415790a0bb3c954b7e386] Merge tag 'media/v5.5-1' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media > > >> git bisect start 'HEAD' '89d57dddd7d3' > > >> # good: [0a6cad5df541108cfd3fbd79eef48eb824c89bdc] Merge branch 'vmwgfx-coherent' of git://people.freedesktop.org/~thomash/linux into drm-next > > >> git bisect good 0a6cad5df541108cfd3fbd79eef48eb824c89bdc > > >> # bad: [9a3d7fd275be4559277667228902824165153c80] Merge tag 'driver-core-5.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core > > >> git bisect bad 9a3d7fd275be4559277667228902824165153c80 > > >> # good: [59274c7164807d27b24e6c068dfe734f7bea4623] Merge tag 'usb-5.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb > > >> git bisect good 59274c7164807d27b24e6c068dfe734f7bea4623 > > >> # good: [e71903106721dc53923e90aa484d78bc86c039a9] staging: mt7621-dma: align to match open parenthesis > > >> git bisect good e71903106721dc53923e90aa484d78bc86c039a9 > > >> # good: [8f56e4ebe05c26c30e167519273843476e39e244] Merge tag 'char-misc-5.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc > > >> git bisect good 8f56e4ebe05c26c30e167519273843476e39e244 > > >> # good: [8bde9f3d2a217d1635a7c7bdf8ad4c25c9a34b50] Merge tag 'iio-for-5.5c' of https://git.kernel.org/pub/scm/linux/kernel/git/jic23/iio into staging-next > > >> git bisect good 8bde9f3d2a217d1635a7c7bdf8ad4c25c9a34b50 > > >> # bad: [0628cda318df6baec439ca6e6e274007492f1ccd] mac80211: Use debugfs_create_xul() helper > > >> git bisect bad 0628cda318df6baec439ca6e6e274007492f1ccd > > >> # bad: [03324507e66c7664c754b1ef92c5c3be24c78aa2] driver core: Allow fwnode_operations.add_links to differentiate errors > > >> git bisect bad 03324507e66c7664c754b1ef92c5c3be24c78aa2 > > >> # bad: [313f5dbba41d905d59c820bb2d91ee6c661aff99] debugfs: remove return value of debugfs_create_u16() > > >> git bisect bad 313f5dbba41d905d59c820bb2d91ee6c661aff99 > > >> # good: [d4387cd117414ba80230f27a514be5ca4a09cfcc] of: property: Create device links for all child-supplier depencencies > > >> git bisect good d4387cd117414ba80230f27a514be5ca4a09cfcc > > >> # bad: [c31e73121f4c1ec45a3e523ac6ce3ce6dafdcec1] base: soc: Handle custom soc information sysfs entries > > >> git bisect bad c31e73121f4c1ec45a3e523ac6ce3ce6dafdcec1 > > >> # bad: [201e91091b1d47047f55580b5474e1239f4d17aa] sh: add the sh_ prefix to early platform symbols > > >> git bisect bad 201e91091b1d47047f55580b5474e1239f4d17aa > > >> # bad: [507fd01d5333338753a1cc26322dfc9f856c109f] drivers: move the early platform device support to arch/sh > > >> git bisect bad 507fd01d5333338753a1cc26322dfc9f856c109f > > >> # first bad commit: [507fd01d5333338753a1cc26322dfc9f856c109f] drivers: move the early platform device support to arch/sh > > > > > > Hi Guenter, > > > > > > can you post some bootlogs? Is it the same problem everywhere? > > > > > > > I don't have any useful logs. The boot fails too early for that. > > > > Yeah, it touched "early" devices after all... > > > Sorry, I don't understand "everywhere". It fails in mainline and -next, if that is what you mean. > > I referred to "All my qemu boot tests" when saying everywhere. > Yes, it affects all my sh and sheb qemu boot tests. See https://kerneltests.org/builders/. > I don't have any relevant HW - the idea for this patch is to stop > compiling for everyone a bunch of code that's only used by one largely > irrelevant architecture and free the namespace for a generic early > platform drivers implementation. > > How are you creating your SH qemu images? I'm seeing there's something > in buildroot for superh - maybe I'll be able to build something > useful. > I use buildroot. Various root file systems and scripts are at https://github.com/groeck/linux-build-test/tree/master/rootfs/{sh,sheb} Guenter
On 12/2/19 9:18 AM, Bartosz Golaszewski wrote: > pon., 2 gru 2019 o 14:03 Guenter Roeck <linux@roeck-us.net> napisał(a): >> >> On 12/1/19 11:40 PM, Bartosz Golaszewski wrote: >>> pon., 2 gru 2019 o 05:14 Guenter Roeck <linux@roeck-us.net> napisał(a): >>>> >>>> On Thu, Oct 03, 2019 at 11:29:12AM +0200, Bartosz Golaszewski wrote: >>>>> From: Bartosz Golaszewski <bgolaszewski@baylibre.com> >>>>> >>>>> SuperH is the only user of the current implementation of early platform >>>>> device support. We want to introduce a more robust approach to early >>>>> probing. As the first step - move all the current early platform code >>>>> to arch/sh. >>>>> >>>>> In order not to export internal drivers/base functions to arch code for >>>>> this temporary solution - copy the two needed routines for driver >>>>> matching from drivers/base/platform.c to arch/sh/drivers/platform_early.c. >>>>> >>>>> Also: call early_platform_cleanup() from subsys_initcall() so that it's >>>>> called after all early devices are probed. >>>>> >>>>> Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com> >>>> >>>> Wondering ... has anyone tested this patch on affected hardware ? >>>> All my qemu boot tests (both sh and sheb) fail because of it. >>>> Bisect log below. >>>> >>>> Guenter >>>> >>>> --- >>>> # bad: [72c0870e3a05d9cd5466d08c3d2a3069ed0a2f9f] Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input >>>> # good: [89d57dddd7d319ded00415790a0bb3c954b7e386] Merge tag 'media/v5.5-1' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media >>>> git bisect start 'HEAD' '89d57dddd7d3' >>>> # good: [0a6cad5df541108cfd3fbd79eef48eb824c89bdc] Merge branch 'vmwgfx-coherent' of git://people.freedesktop.org/~thomash/linux into drm-next >>>> git bisect good 0a6cad5df541108cfd3fbd79eef48eb824c89bdc >>>> # bad: [9a3d7fd275be4559277667228902824165153c80] Merge tag 'driver-core-5.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core >>>> git bisect bad 9a3d7fd275be4559277667228902824165153c80 >>>> # good: [59274c7164807d27b24e6c068dfe734f7bea4623] Merge tag 'usb-5.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb >>>> git bisect good 59274c7164807d27b24e6c068dfe734f7bea4623 >>>> # good: [e71903106721dc53923e90aa484d78bc86c039a9] staging: mt7621-dma: align to match open parenthesis >>>> git bisect good e71903106721dc53923e90aa484d78bc86c039a9 >>>> # good: [8f56e4ebe05c26c30e167519273843476e39e244] Merge tag 'char-misc-5.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc >>>> git bisect good 8f56e4ebe05c26c30e167519273843476e39e244 >>>> # good: [8bde9f3d2a217d1635a7c7bdf8ad4c25c9a34b50] Merge tag 'iio-for-5.5c' of https://git.kernel.org/pub/scm/linux/kernel/git/jic23/iio into staging-next >>>> git bisect good 8bde9f3d2a217d1635a7c7bdf8ad4c25c9a34b50 >>>> # bad: [0628cda318df6baec439ca6e6e274007492f1ccd] mac80211: Use debugfs_create_xul() helper >>>> git bisect bad 0628cda318df6baec439ca6e6e274007492f1ccd >>>> # bad: [03324507e66c7664c754b1ef92c5c3be24c78aa2] driver core: Allow fwnode_operations.add_links to differentiate errors >>>> git bisect bad 03324507e66c7664c754b1ef92c5c3be24c78aa2 >>>> # bad: [313f5dbba41d905d59c820bb2d91ee6c661aff99] debugfs: remove return value of debugfs_create_u16() >>>> git bisect bad 313f5dbba41d905d59c820bb2d91ee6c661aff99 >>>> # good: [d4387cd117414ba80230f27a514be5ca4a09cfcc] of: property: Create device links for all child-supplier depencencies >>>> git bisect good d4387cd117414ba80230f27a514be5ca4a09cfcc >>>> # bad: [c31e73121f4c1ec45a3e523ac6ce3ce6dafdcec1] base: soc: Handle custom soc information sysfs entries >>>> git bisect bad c31e73121f4c1ec45a3e523ac6ce3ce6dafdcec1 >>>> # bad: [201e91091b1d47047f55580b5474e1239f4d17aa] sh: add the sh_ prefix to early platform symbols >>>> git bisect bad 201e91091b1d47047f55580b5474e1239f4d17aa >>>> # bad: [507fd01d5333338753a1cc26322dfc9f856c109f] drivers: move the early platform device support to arch/sh >>>> git bisect bad 507fd01d5333338753a1cc26322dfc9f856c109f >>>> # first bad commit: [507fd01d5333338753a1cc26322dfc9f856c109f] drivers: move the early platform device support to arch/sh >>> >>> Hi Guenter, >>> >>> can you post some bootlogs? Is it the same problem everywhere? >>> >> >> I don't have any useful logs. The boot fails too early for that. >> > > Yeah, it touched "early" devices after all... > >> Sorry, I don't understand "everywhere". It fails in mainline and -next, if that is what you mean. > > I referred to "All my qemu boot tests" when saying everywhere. > > I don't have any relevant HW - the idea for this patch is to stop > compiling for everyone a bunch of code that's only used by one largely > irrelevant architecture and free the namespace for a generic early > platform drivers implementation. > > How are you creating your SH qemu images? I'm seeing there's something > in buildroot for superh - maybe I'll be able to bearly_platform_cleanupuild something > useful. > Below is a possible fix. As I had suspected, the call to early_platform_cleanup() is in the wrong place. Guenter --- arch/sh/drivers/platform_early.c | 11 ++--------- drivers/base/platform.c | 4 ++++ 2 files changed, 6 insertions(+), 9 deletions(-) diff --git a/arch/sh/drivers/platform_early.c b/arch/sh/drivers/platform_early.c index f6d148451dfc..16f33bffd8fc 100644 --- a/arch/sh/drivers/platform_early.c +++ b/arch/sh/drivers/platform_early.c @@ -325,9 +325,9 @@ int __init sh_early_platform_driver_probe(char *class_str, } /** - * sh_early_platform_cleanup - clean up early platform code + * early_platform_cleanup - clean up early platform code */ -static int __init sh_early_platform_cleanup(void) +void early_platform_cleanup(void) { struct platform_device *pd, *pd2; @@ -337,11 +337,4 @@ static int __init sh_early_platform_cleanup(void) list_del(&pd->dev.devres_head); memset(&pd->dev.devres_head, 0, sizeof(pd->dev.devres_head)); } - - return 0; } -/* - * This must happen once after all early devices are probed but before probing - * real platform devices. - */ -subsys_initcall(sh_early_platform_cleanup); diff --git a/drivers/base/platform.c b/drivers/base/platform.c index 7c532548b0a6..3ba153e356ee 100644 --- a/drivers/base/platform.c +++ b/drivers/base/platform.c @@ -1325,10 +1325,14 @@ struct device *platform_find_device_by_driver(struct device *start, } EXPORT_SYMBOL_GPL(platform_find_device_by_driver); +void __weak early_platform_cleanup(void) { } + int __init platform_bus_init(void) { int error; + early_platform_cleanup(); + error = device_register(&platform_bus); if (error) { put_device(&platform_bus);
wt., 3 gru 2019 o 06:24 Guenter Roeck <linux@roeck-us.net> napisał(a): > > On 12/2/19 9:18 AM, Bartosz Golaszewski wrote: > > pon., 2 gru 2019 o 14:03 Guenter Roeck <linux@roeck-us.net> napisał(a): > >> > >> On 12/1/19 11:40 PM, Bartosz Golaszewski wrote: > >>> pon., 2 gru 2019 o 05:14 Guenter Roeck <linux@roeck-us.net> napisał(a): > >>>> > >>>> On Thu, Oct 03, 2019 at 11:29:12AM +0200, Bartosz Golaszewski wrote: > >>>>> From: Bartosz Golaszewski <bgolaszewski@baylibre.com> > >>>>> > >>>>> SuperH is the only user of the current implementation of early platform > >>>>> device support. We want to introduce a more robust approach to early > >>>>> probing. As the first step - move all the current early platform code > >>>>> to arch/sh. > >>>>> > >>>>> In order not to export internal drivers/base functions to arch code for > >>>>> this temporary solution - copy the two needed routines for driver > >>>>> matching from drivers/base/platform.c to arch/sh/drivers/platform_early.c. > >>>>> > >>>>> Also: call early_platform_cleanup() from subsys_initcall() so that it's > >>>>> called after all early devices are probed. > >>>>> > >>>>> Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com> > >>>> > >>>> Wondering ... has anyone tested this patch on affected hardware ? > >>>> All my qemu boot tests (both sh and sheb) fail because of it. > >>>> Bisect log below. > >>>> > >>>> Guenter > >>>> > >>>> --- > >>>> # bad: [72c0870e3a05d9cd5466d08c3d2a3069ed0a2f9f] Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input > >>>> # good: [89d57dddd7d319ded00415790a0bb3c954b7e386] Merge tag 'media/v5.5-1' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media > >>>> git bisect start 'HEAD' '89d57dddd7d3' > >>>> # good: [0a6cad5df541108cfd3fbd79eef48eb824c89bdc] Merge branch 'vmwgfx-coherent' of git://people.freedesktop.org/~thomash/linux into drm-next > >>>> git bisect good 0a6cad5df541108cfd3fbd79eef48eb824c89bdc > >>>> # bad: [9a3d7fd275be4559277667228902824165153c80] Merge tag 'driver-core-5.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core > >>>> git bisect bad 9a3d7fd275be4559277667228902824165153c80 > >>>> # good: [59274c7164807d27b24e6c068dfe734f7bea4623] Merge tag 'usb-5.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb > >>>> git bisect good 59274c7164807d27b24e6c068dfe734f7bea4623 > >>>> # good: [e71903106721dc53923e90aa484d78bc86c039a9] staging: mt7621-dma: align to match open parenthesis > >>>> git bisect good e71903106721dc53923e90aa484d78bc86c039a9 > >>>> # good: [8f56e4ebe05c26c30e167519273843476e39e244] Merge tag 'char-misc-5.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc > >>>> git bisect good 8f56e4ebe05c26c30e167519273843476e39e244 > >>>> # good: [8bde9f3d2a217d1635a7c7bdf8ad4c25c9a34b50] Merge tag 'iio-for-5.5c' of https://git.kernel.org/pub/scm/linux/kernel/git/jic23/iio into staging-next > >>>> git bisect good 8bde9f3d2a217d1635a7c7bdf8ad4c25c9a34b50 > >>>> # bad: [0628cda318df6baec439ca6e6e274007492f1ccd] mac80211: Use debugfs_create_xul() helper > >>>> git bisect bad 0628cda318df6baec439ca6e6e274007492f1ccd > >>>> # bad: [03324507e66c7664c754b1ef92c5c3be24c78aa2] driver core: Allow fwnode_operations.add_links to differentiate errors > >>>> git bisect bad 03324507e66c7664c754b1ef92c5c3be24c78aa2 > >>>> # bad: [313f5dbba41d905d59c820bb2d91ee6c661aff99] debugfs: remove return value of debugfs_create_u16() > >>>> git bisect bad 313f5dbba41d905d59c820bb2d91ee6c661aff99 > >>>> # good: [d4387cd117414ba80230f27a514be5ca4a09cfcc] of: property: Create device links for all child-supplier depencencies > >>>> git bisect good d4387cd117414ba80230f27a514be5ca4a09cfcc > >>>> # bad: [c31e73121f4c1ec45a3e523ac6ce3ce6dafdcec1] base: soc: Handle custom soc information sysfs entries > >>>> git bisect bad c31e73121f4c1ec45a3e523ac6ce3ce6dafdcec1 > >>>> # bad: [201e91091b1d47047f55580b5474e1239f4d17aa] sh: add the sh_ prefix to early platform symbols > >>>> git bisect bad 201e91091b1d47047f55580b5474e1239f4d17aa > >>>> # bad: [507fd01d5333338753a1cc26322dfc9f856c109f] drivers: move the early platform device support to arch/sh > >>>> git bisect bad 507fd01d5333338753a1cc26322dfc9f856c109f > >>>> # first bad commit: [507fd01d5333338753a1cc26322dfc9f856c109f] drivers: move the early platform device support to arch/sh > >>> > >>> Hi Guenter, > >>> > >>> can you post some bootlogs? Is it the same problem everywhere? > >>> > >> > >> I don't have any useful logs. The boot fails too early for that. > >> > > > > Yeah, it touched "early" devices after all... > > > >> Sorry, I don't understand "everywhere". It fails in mainline and -next, if that is what you mean. > > > > I referred to "All my qemu boot tests" when saying everywhere. > > > > I don't have any relevant HW - the idea for this patch is to stop > > compiling for everyone a bunch of code that's only used by one largely > > irrelevant architecture and free the namespace for a generic early > > platform drivers implementation. > > > > How are you creating your SH qemu images? I'm seeing there's something > > in buildroot for superh - maybe I'll be able to bearly_platform_cleanupuild something > > useful. > > > > Below is a possible fix. As I had suspected, the call to early_platform_cleanup() > is in the wrong place. > This makes sense. If it fixes the issue - do you want to send it to Greg? Bart > Guenter > > --- > arch/sh/drivers/platform_early.c | 11 ++--------- > drivers/base/platform.c | 4 ++++ > 2 files changed, 6 insertions(+), 9 deletions(-) > > diff --git a/arch/sh/drivers/platform_early.c b/arch/sh/drivers/platform_early.c > index f6d148451dfc..16f33bffd8fc 100644 > --- a/arch/sh/drivers/platform_early.c > +++ b/arch/sh/drivers/platform_early.c > @@ -325,9 +325,9 @@ int __init sh_early_platform_driver_probe(char *class_str, > } > > /** > - * sh_early_platform_cleanup - clean up early platform code > + * early_platform_cleanup - clean up early platform code > */ > -static int __init sh_early_platform_cleanup(void) > +void early_platform_cleanup(void) > { > struct platform_device *pd, *pd2; > > @@ -337,11 +337,4 @@ static int __init sh_early_platform_cleanup(void) > list_del(&pd->dev.devres_head); > memset(&pd->dev.devres_head, 0, sizeof(pd->dev.devres_head)); > } > - > - return 0; > } > -/* > - * This must happen once after all early devices are probed but before probing > - * real platform devices. > - */ > -subsys_initcall(sh_early_platform_cleanup); > diff --git a/drivers/base/platform.c b/drivers/base/platform.c > index 7c532548b0a6..3ba153e356ee 100644 > --- a/drivers/base/platform.c > +++ b/drivers/base/platform.c > @@ -1325,10 +1325,14 @@ struct device *platform_find_device_by_driver(struct device *start, > } > EXPORT_SYMBOL_GPL(platform_find_device_by_driver); > > +void __weak early_platform_cleanup(void) { } > + > int __init platform_bus_init(void) > { > int error; > > + early_platform_cleanup(); > + > error = device_register(&platform_bus); > if (error) { > put_device(&platform_bus); >
Hi Bartosz, Günter, On Tue, Dec 3, 2019 at 10:50 AM Bartosz Golaszewski <brgl@bgdev.pl> wrote: > wt., 3 gru 2019 o 06:24 Guenter Roeck <linux@roeck-us.net> napisał(a): > > On 12/2/19 9:18 AM, Bartosz Golaszewski wrote: > > > pon., 2 gru 2019 o 14:03 Guenter Roeck <linux@roeck-us.net> napisał(a): > > >> > > >> On 12/1/19 11:40 PM, Bartosz Golaszewski wrote: > > >>> pon., 2 gru 2019 o 05:14 Guenter Roeck <linux@roeck-us.net> napisał(a): > > >>>> > > >>>> On Thu, Oct 03, 2019 at 11:29:12AM +0200, Bartosz Golaszewski wrote: > > >>>>> From: Bartosz Golaszewski <bgolaszewski@baylibre.com> > > >>>>> > > >>>>> SuperH is the only user of the current implementation of early platform > > >>>>> device support. We want to introduce a more robust approach to early > > >>>>> probing. As the first step - move all the current early platform code > > >>>>> to arch/sh. > > >>>>> > > >>>>> In order not to export internal drivers/base functions to arch code for > > >>>>> this temporary solution - copy the two needed routines for driver > > >>>>> matching from drivers/base/platform.c to arch/sh/drivers/platform_early.c. > > >>>>> > > >>>>> Also: call early_platform_cleanup() from subsys_initcall() so that it's > > >>>>> called after all early devices are probed. > > >>>>> > > >>>>> Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com> > > >>>> > > >>>> Wondering ... has anyone tested this patch on affected hardware ? > > >>>> All my qemu boot tests (both sh and sheb) fail because of it. > > >>>> Bisect log below. > > >>>> > > >>>> Guenter > > >>>> > > >>>> --- > > >>>> # bad: [72c0870e3a05d9cd5466d08c3d2a3069ed0a2f9f] Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input > > >>>> # good: [89d57dddd7d319ded00415790a0bb3c954b7e386] Merge tag 'media/v5.5-1' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media > > >>>> git bisect start 'HEAD' '89d57dddd7d3' > > >>>> # good: [0a6cad5df541108cfd3fbd79eef48eb824c89bdc] Merge branch 'vmwgfx-coherent' of git://people.freedesktop.org/~thomash/linux into drm-next > > >>>> git bisect good 0a6cad5df541108cfd3fbd79eef48eb824c89bdc > > >>>> # bad: [9a3d7fd275be4559277667228902824165153c80] Merge tag 'driver-core-5.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core > > >>>> git bisect bad 9a3d7fd275be4559277667228902824165153c80 > > >>>> # good: [59274c7164807d27b24e6c068dfe734f7bea4623] Merge tag 'usb-5.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb > > >>>> git bisect good 59274c7164807d27b24e6c068dfe734f7bea4623 > > >>>> # good: [e71903106721dc53923e90aa484d78bc86c039a9] staging: mt7621-dma: align to match open parenthesis > > >>>> git bisect good e71903106721dc53923e90aa484d78bc86c039a9 > > >>>> # good: [8f56e4ebe05c26c30e167519273843476e39e244] Merge tag 'char-misc-5.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc > > >>>> git bisect good 8f56e4ebe05c26c30e167519273843476e39e244 > > >>>> # good: [8bde9f3d2a217d1635a7c7bdf8ad4c25c9a34b50] Merge tag 'iio-for-5.5c' of https://git.kernel.org/pub/scm/linux/kernel/git/jic23/iio into staging-next > > >>>> git bisect good 8bde9f3d2a217d1635a7c7bdf8ad4c25c9a34b50 > > >>>> # bad: [0628cda318df6baec439ca6e6e274007492f1ccd] mac80211: Use debugfs_create_xul() helper > > >>>> git bisect bad 0628cda318df6baec439ca6e6e274007492f1ccd > > >>>> # bad: [03324507e66c7664c754b1ef92c5c3be24c78aa2] driver core: Allow fwnode_operations.add_links to differentiate errors > > >>>> git bisect bad 03324507e66c7664c754b1ef92c5c3be24c78aa2 > > >>>> # bad: [313f5dbba41d905d59c820bb2d91ee6c661aff99] debugfs: remove return value of debugfs_create_u16() > > >>>> git bisect bad 313f5dbba41d905d59c820bb2d91ee6c661aff99 > > >>>> # good: [d4387cd117414ba80230f27a514be5ca4a09cfcc] of: property: Create device links for all child-supplier depencencies > > >>>> git bisect good d4387cd117414ba80230f27a514be5ca4a09cfcc > > >>>> # bad: [c31e73121f4c1ec45a3e523ac6ce3ce6dafdcec1] base: soc: Handle custom soc information sysfs entries > > >>>> git bisect bad c31e73121f4c1ec45a3e523ac6ce3ce6dafdcec1 > > >>>> # bad: [201e91091b1d47047f55580b5474e1239f4d17aa] sh: add the sh_ prefix to early platform symbols > > >>>> git bisect bad 201e91091b1d47047f55580b5474e1239f4d17aa > > >>>> # bad: [507fd01d5333338753a1cc26322dfc9f856c109f] drivers: move the early platform device support to arch/sh > > >>>> git bisect bad 507fd01d5333338753a1cc26322dfc9f856c109f > > >>>> # first bad commit: [507fd01d5333338753a1cc26322dfc9f856c109f] drivers: move the early platform device support to arch/sh > > >>> > > >>> Hi Guenter, > > >>> > > >>> can you post some bootlogs? Is it the same problem everywhere? > > >>> > > >> > > >> I don't have any useful logs. The boot fails too early for that. > > >> > > > > > > Yeah, it touched "early" devices after all... > > > > > >> Sorry, I don't understand "everywhere". It fails in mainline and -next, if that is what you mean. > > > > > > I referred to "All my qemu boot tests" when saying everywhere. > > > > > > I don't have any relevant HW - the idea for this patch is to stop > > > compiling for everyone a bunch of code that's only used by one largely > > > irrelevant architecture and free the namespace for a generic early > > > platform drivers implementation. > > > > > > How are you creating your SH qemu images? I'm seeing there's something > > > in buildroot for superh - maybe I'll be able to bearly_platform_cleanupuild something > > > useful. > > > > > > > Below is a possible fix. As I had suspected, the call to early_platform_cleanup() > > is in the wrong place. > > > > This makes sense. If it fixes the issue - do you want to send it to Greg? Which tree does this patch apply to? Oh, it's whitespace-damaged. After applying it manually on top of current Linus' tree, (remote) SH7722-based Migo-R boots again, so Tested-by: Geert Uytterhoeven <geert+renesas@glider.be> There's still an issue with crashing userspace, but that happens on v5.4 too. Bisecting... Gr{oetje,eeting}s, Geert
On 12/3/19 2:14 AM, Geert Uytterhoeven wrote: > Hi Bartosz, Günter, > > On Tue, Dec 3, 2019 at 10:50 AM Bartosz Golaszewski <brgl@bgdev.pl> wrote: >> wt., 3 gru 2019 o 06:24 Guenter Roeck <linux@roeck-us.net> napisał(a): >>> On 12/2/19 9:18 AM, Bartosz Golaszewski wrote: >>>> pon., 2 gru 2019 o 14:03 Guenter Roeck <linux@roeck-us.net> napisał(a): >>>>> >>>>> On 12/1/19 11:40 PM, Bartosz Golaszewski wrote: >>>>>> pon., 2 gru 2019 o 05:14 Guenter Roeck <linux@roeck-us.net> napisał(a): >>>>>>> >>>>>>> On Thu, Oct 03, 2019 at 11:29:12AM +0200, Bartosz Golaszewski wrote: >>>>>>>> From: Bartosz Golaszewski <bgolaszewski@baylibre.com> >>>>>>>> >>>>>>>> SuperH is the only user of the current implementation of early platform >>>>>>>> device support. We want to introduce a more robust approach to early >>>>>>>> probing. As the first step - move all the current early platform code >>>>>>>> to arch/sh. >>>>>>>> >>>>>>>> In order not to export internal drivers/base functions to arch code for >>>>>>>> this temporary solution - copy the two needed routines for driver >>>>>>>> matching from drivers/base/platform.c to arch/sh/drivers/platform_early.c. >>>>>>>> >>>>>>>> Also: call early_platform_cleanup() from subsys_initcall() so that it's >>>>>>>> called after all early devices are probed. >>>>>>>> >>>>>>>> Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com> >>>>>>> >>>>>>> Wondering ... has anyone tested this patch on affected hardware ? >>>>>>> All my qemu boot tests (both sh and sheb) fail because of it. >>>>>>> Bisect log below. >>>>>>> >>>>>>> Guenter >>>>>>> >>>>>>> --- >>>>>>> # bad: [72c0870e3a05d9cd5466d08c3d2a3069ed0a2f9f] Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input >>>>>>> # good: [89d57dddd7d319ded00415790a0bb3c954b7e386] Merge tag 'media/v5.5-1' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media >>>>>>> git bisect start 'HEAD' '89d57dddd7d3' >>>>>>> # good: [0a6cad5df541108cfd3fbd79eef48eb824c89bdc] Merge branch 'vmwgfx-coherent' of git://people.freedesktop.org/~thomash/linux into drm-next >>>>>>> git bisect good 0a6cad5df541108cfd3fbd79eef48eb824c89bdc >>>>>>> # bad: [9a3d7fd275be4559277667228902824165153c80] Merge tag 'driver-core-5.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core >>>>>>> git bisect bad 9a3d7fd275be4559277667228902824165153c80 >>>>>>> # good: [59274c7164807d27b24e6c068dfe734f7bea4623] Merge tag 'usb-5.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb >>>>>>> git bisect good 59274c7164807d27b24e6c068dfe734f7bea4623 >>>>>>> # good: [e71903106721dc53923e90aa484d78bc86c039a9] staging: mt7621-dma: align to match open parenthesis >>>>>>> git bisect good e71903106721dc53923e90aa484d78bc86c039a9 >>>>>>> # good: [8f56e4ebe05c26c30e167519273843476e39e244] Merge tag 'char-misc-5.5-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc >>>>>>> git bisect good 8f56e4ebe05c26c30e167519273843476e39e244 >>>>>>> # good: [8bde9f3d2a217d1635a7c7bdf8ad4c25c9a34b50] Merge tag 'iio-for-5.5c' of https://git.kernel.org/pub/scm/linux/kernel/git/jic23/iio into staging-next >>>>>>> git bisect good 8bde9f3d2a217d1635a7c7bdf8ad4c25c9a34b50 >>>>>>> # bad: [0628cda318df6baec439ca6e6e274007492f1ccd] mac80211: Use debugfs_create_xul() helper >>>>>>> git bisect bad 0628cda318df6baec439ca6e6e274007492f1ccd >>>>>>> # bad: [03324507e66c7664c754b1ef92c5c3be24c78aa2] driver core: Allow fwnode_operations.add_links to differentiate errors >>>>>>> git bisect bad 03324507e66c7664c754b1ef92c5c3be24c78aa2 >>>>>>> # bad: [313f5dbba41d905d59c820bb2d91ee6c661aff99] debugfs: remove return value of debugfs_create_u16() >>>>>>> git bisect bad 313f5dbba41d905d59c820bb2d91ee6c661aff99 >>>>>>> # good: [d4387cd117414ba80230f27a514be5ca4a09cfcc] of: property: Create device links for all child-supplier depencencies >>>>>>> git bisect good d4387cd117414ba80230f27a514be5ca4a09cfcc >>>>>>> # bad: [c31e73121f4c1ec45a3e523ac6ce3ce6dafdcec1] base: soc: Handle custom soc information sysfs entries >>>>>>> git bisect bad c31e73121f4c1ec45a3e523ac6ce3ce6dafdcec1 >>>>>>> # bad: [201e91091b1d47047f55580b5474e1239f4d17aa] sh: add the sh_ prefix to early platform symbols >>>>>>> git bisect bad 201e91091b1d47047f55580b5474e1239f4d17aa >>>>>>> # bad: [507fd01d5333338753a1cc26322dfc9f856c109f] drivers: move the early platform device support to arch/sh >>>>>>> git bisect bad 507fd01d5333338753a1cc26322dfc9f856c109f >>>>>>> # first bad commit: [507fd01d5333338753a1cc26322dfc9f856c109f] drivers: move the early platform device support to arch/sh >>>>>> >>>>>> Hi Guenter, >>>>>> >>>>>> can you post some bootlogs? Is it the same problem everywhere? >>>>>> >>>>> >>>>> I don't have any useful logs. The boot fails too early for that. >>>>> >>>> >>>> Yeah, it touched "early" devices after all... >>>> >>>>> Sorry, I don't understand "everywhere". It fails in mainline and -next, if that is what you mean. >>>> >>>> I referred to "All my qemu boot tests" when saying everywhere. >>>> >>>> I don't have any relevant HW - the idea for this patch is to stop >>>> compiling for everyone a bunch of code that's only used by one largely >>>> irrelevant architecture and free the namespace for a generic early >>>> platform drivers implementation. >>>> >>>> How are you creating your SH qemu images? I'm seeing there's something >>>> in buildroot for superh - maybe I'll be able to bearly_platform_cleanupuild something >>>> useful. >>>> >>> >>> Below is a possible fix. As I had suspected, the call to early_platform_cleanup() >>> is in the wrong place. >>> >> >> This makes sense. If it fixes the issue - do you want to send it to Greg? > > Which tree does this patch apply to? > Oh, it's whitespace-damaged. > Sorry, that wasn't meant to be applicable, just a hack to confirm my suspicion. I'll clean it up and submit it. Thanks, Guenter > After applying it manually on top of current Linus' tree, (remote) > SH7722-based Migo-R boots again, so > Tested-by: Geert Uytterhoeven <geert+renesas@glider.be> > > There's still an issue with crashing userspace, but that happens on v5.4 > too. Bisecting... > > Gr{oetje,eeting}s, > > Geert >
From: Bartosz Golaszewski <bgolaszewski@baylibre.com> Some time ago I started a discussion about the need for a proper early device probing mechanism[1]. One that would be based on real platform drivers and support both platform data and device tree. While we're far from reaching any consensus on the implementation, Arnd suggested that I start off by moving the SuperH-specific early platform drivers implementation to arch/sh[2]. This series is the first attempt at making way for a new, less hacky implementation. The first patch moves all the early_platform code to arch/sh. The second patch prefixes all early_platform symbols with 'sh_'. [1] https://lkml.org/lkml/2018/4/26/657 [2] https://lkml.org/lkml/2018/4/27/239 v1 -> v2: - certain drivers are compiled for arm/mach-shmobile too - we need to add ifdefs for CONFIG_SUPERH around early_platform calls v2 -> v3: - added a stub for is_early_platform_device() which always returns false on non-SuperH architectures v3 -> v4: - rebased on top of v5.4-rc1 - removed patches that are already upstream from the series Bartosz Golaszewski (2): drivers: move the early platform device support to arch/sh sh: add the sh_ prefix to early platform symbols arch/sh/drivers/Makefile | 2 +- arch/sh/drivers/platform_early.c | 347 +++++++++++++++++++++++++ arch/sh/include/asm/platform_early.h | 61 +++++ arch/sh/kernel/cpu/sh2/setup-sh7619.c | 3 +- arch/sh/kernel/cpu/sh2a/setup-mxg.c | 3 +- arch/sh/kernel/cpu/sh2a/setup-sh7201.c | 3 +- arch/sh/kernel/cpu/sh2a/setup-sh7203.c | 3 +- arch/sh/kernel/cpu/sh2a/setup-sh7206.c | 3 +- arch/sh/kernel/cpu/sh2a/setup-sh7264.c | 3 +- arch/sh/kernel/cpu/sh2a/setup-sh7269.c | 3 +- arch/sh/kernel/cpu/sh3/setup-sh3.c | 1 + arch/sh/kernel/cpu/sh3/setup-sh7705.c | 3 +- arch/sh/kernel/cpu/sh3/setup-sh770x.c | 3 +- arch/sh/kernel/cpu/sh3/setup-sh7710.c | 3 +- arch/sh/kernel/cpu/sh3/setup-sh7720.c | 3 +- arch/sh/kernel/cpu/sh4/setup-sh4-202.c | 3 +- arch/sh/kernel/cpu/sh4/setup-sh7750.c | 9 +- arch/sh/kernel/cpu/sh4/setup-sh7760.c | 3 +- arch/sh/kernel/cpu/sh4a/setup-sh7343.c | 3 +- arch/sh/kernel/cpu/sh4a/setup-sh7366.c | 3 +- arch/sh/kernel/cpu/sh4a/setup-sh7722.c | 3 +- arch/sh/kernel/cpu/sh4a/setup-sh7723.c | 3 +- arch/sh/kernel/cpu/sh4a/setup-sh7724.c | 3 +- arch/sh/kernel/cpu/sh4a/setup-sh7734.c | 3 +- arch/sh/kernel/cpu/sh4a/setup-sh7757.c | 3 +- arch/sh/kernel/cpu/sh4a/setup-sh7763.c | 3 +- arch/sh/kernel/cpu/sh4a/setup-sh7770.c | 3 +- arch/sh/kernel/cpu/sh4a/setup-sh7780.c | 3 +- arch/sh/kernel/cpu/sh4a/setup-sh7785.c | 3 +- arch/sh/kernel/cpu/sh4a/setup-sh7786.c | 3 +- arch/sh/kernel/cpu/sh4a/setup-shx3.c | 3 +- arch/sh/kernel/cpu/sh5/setup-sh5.c | 3 +- arch/sh/kernel/setup.c | 3 +- arch/sh/kernel/time.c | 5 +- drivers/base/platform.c | 288 -------------------- drivers/clocksource/sh_cmt.c | 13 +- drivers/clocksource/sh_mtu2.c | 13 +- drivers/clocksource/sh_tmu.c | 14 +- drivers/tty/serial/sh-sci.c | 11 +- include/linux/platform_device.h | 64 +---- 40 files changed, 525 insertions(+), 387 deletions(-) create mode 100644 arch/sh/drivers/platform_early.c create mode 100644 arch/sh/include/asm/platform_early.h