- 26 Jun, 2019 1 commit
-
-
Daniel Vetter authored
Fixes linking fail when fbcon/fbdev is modular and vgaswitcheroo is enabled: x86_64-linux-gnu-ld: drivers/gpu/vga/vga_switcheroo.o: in function `vga_switchto_stage2': vga_switcheroo.c:(.text+0x997): undefined reference to `fbcon_remap_all' Unfortunately this means that in a default allmodconfig we will disable vgaswitcheroo. v2: I first tried to fix this with an EXPORT_SYMBOL, but vga_switcheroo is never a module, so this doesn't work. Reported-by:
Stephen Rothwell <sfr@canb.auug.org.au> Reported-by:
kbuild test robot <lkp@intel.com> Fixes: 1cd51b5d ("vgaswitcheroo: call fbcon_remap_all directly") Signed-off-by:
Daniel Vetter <daniel.vetter@intel.com> Cc: Stephen Rothwell <sfr@canb.auug.org.au> Cc: Lukas Wunner <lukas@wunner.de> Cc: Daniel Vetter <daniel.vetter@intel.com> Cc: Sam Ravnborg <sam@ravnborg.org> Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Cc: David Airlie <airlied@linux.ie> Cc: Daniel Vetter <daniel@ffwll.ch> Cc: Maxime Ripard <maxime.ripard@bootlin.com> Cc: Sean Paul <sean@poorly.run> Cc: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com> Cc: Hans de Goede <hdegoede@redhat.com> Cc: Yisheng Xie <ysxie@foxmail.com> Cc: linux-fbdev@vger.kernel.org Cc: Daniel Thompson <daniel.thompson@linaro.org> Cc: Lee Jones <lee.jones@linaro.org> Cc: dri-devel@lists.freedesktop.org Cc: linux-fbdev@vger.kernel.org Signed-off-by:
Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20190624092153.23375-1-daniel.vetter@ffwll.ch
-
- 19 Jun, 2019 1 commit
-
-
Daniel Vetter authored
I failed to spot this while compile-testing. Oops. Reported-by:
-
- 13 Jun, 2019 3 commits
-
-
Daniel Vetter authored
With all the work I've done on replacing fb notifier calls with direct calls into fbcon the backlight/lcd notifier is the only user left. It will only receive events now that it cares about, hence we can remove this check. Signed-off-by:
Sam Ravnborg <sam@ravnborg.org> Reviewed-by:
Daniel Thompson <daniel.thompson@linaro.org> Cc: Lee Jones <lee.jones@linaro.org> Cc: Daniel Thompson <daniel.thompson@linaro.org> Cc: Jingoo Han <jingoohan1@gmail.com> Link: https://patchwork.freedesktop.org/patch/msgid/20190528090304.9388-34-daniel.vetter@ffwll.ch
-
Daniel Vetter authored
this driver is pretty horrible from a design pov, and needs a complete overhaul. Concrete thing that annoys me is that it looks at registered_fb, which is an internal thing to fbmem.c and fbcon.c. And ofc it gets the lifetime rules all wrong (it should at least use get/put_fb_info). Looking at the history, there's been an attempt at dropping this from staging in 2016, but that had to be reverted. Since then not real effort except the usual stream of trivial patches, and fbdev has been formally closed for any new hw support. Time to try again and drop this? Signed-off-by:
Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by:
-
Daniel Vetter authored
It's not pretty. Signed-off-by:
-
- 12 Jun, 2019 30 commits
-
-
Daniel Vetter authored
These are actually fbcon ioctls which just happen to be exposed through /dev/fb*. They completely ignore which fb_info they're called on, and I think the userspace tool even hardcodes to /dev/fb0. Hence just forward the entire thing to fbcon.c wholesale. Note that this patch drops the fb_lock/unlock on the set side. Since the ioctl can operate on any fb (as passed in through con2fb.framebuffer) this is bogus. Also note that fbcon.c in general never calls fb_lock on anything, so this has been badly broken already. With this the last user of the fbcon notifier callback is gone, and we can garbage collect that too. v2: add missing uaccess.h include (alpha fails to compile otherwise), reported by kbuild. v3: Remember to also drop the #defines (Maarten) v4: Add the static inline to dummy functions. Signed-off-by:
-
Daniel Vetter authored
While at it, clean up the interface a bit and push the console locking into fbcon.c. v2: Remove now outdated comment (Lukas). v3: Forgot to add static inline to the dummy function. Acked-by:
Lukas Wunner <lukas@wunner.de> Signed-off-by:
-
Daniel Vetter authored
Create a new wrapper function for this, feels like there's some refactoring room here between the two modes. v2: backlight notifier is also interested in the mode change event, it calls lcd->set_mode, of which there are 3 implementations. Thanks to Maarten for spotting this. So we keep that. We can ditch the differentiation between mode change and all mode changes (because backlight notifier doesn't care), and we can drop the FBINFO_MISC_USEREVENT stuff too, because that's just to prevent recursion between fbmem.c and fbcon.c. While at it flatten the control flow a bit. v3: Need to add a static inline to the dummy function. v4: Add missing #include <fbcon.h> to sh_mob (Sam). Cc: Sam Ravnborg <sam@ravnborg.org> Signed-off-by:
-
Daniel Vetter authored
Instead of wiring almost everything down to the very last line using goto soup (but not consistently, where would the fun be otherwise) drop out early when checks fail. This allows us to flatten the huge indent levels to just 1. Aside: If a driver doesn't set ->fb_check_var, then FB_ACTIVATE_NOW does nothing. This bug exists ever since this code was extracted as a common helper in 2002, hence I decided against fixing it. Everyone just better have a fb_check_var to make sure things work correctly. Signed-off-by:
-
Daniel Vetter authored
With the recursion broken in the previous patch we can drop the FBINFO_MISC_USEREVENT flag around calls to fb_blank - recursion prevention was it's only job. Signed-off-by:
-
Daniel Vetter authored
There's a callchain of: fbcon_fb_blanked -> do_(un)blank_screen -> consw->con_blank -> fbcon_blank -> fb_blank Things don't go horribly wrong because the BKL console_lock safes the day, but that's about it. And the seeming recursion is broken in 2 ways: - Starting from the fbdev ioctl we set FBINFO_MISC_USEREVENT, which tells the fbcon_blank code to not call fb_blank. This was required to not deadlock when recursing on the fb_notifier_chain mutex. - Starting from the con_blank hook we're getting saved by the console_blanked checks in do_blank/unblank_screen. Or at least that's my theory. Anyway, recursion isn't awesome, so let's stop it. Breaking the recursion avoids the need to be in the FBINFO_MISC_USEREVENT critical section, so lets move it out of that too. The astute reader will notice that fb_blank seems to require lock_fb_info(), which the fbcon code seems to ignore. I have no idea how to fix that problem, so let's keep ignoring it. v2: I forgot the sysfs blanking code. v3: Fix typo in callchain in the commmit message (Sam). Signed-off-by:
-
Daniel Vetter authored
This reverts commit 994efacd. The justification is that if hw blanking fails (i.e. fbops->fb_blank) fails, then we still want to shut down the backlight. Which is exactly _not_ what fb_blank() does and so rather inconsistent if we end up with different behaviour between fbcon and direct fbdev usage. Given that the entire notifier maze is getting in the way anyway I figured it's simplest to revert this not well justified commit. v2: Add static inline to the dummy version. Cc: Richard Purdie <rpurdie@rpsys.net> Signed-off-by:
-
Daniel Vetter authored
Pretty simple case really. v2: Forgot to remove a break; v3: Add static inline to the dummy versions. Signed-off-by:
-
Daniel Vetter authored
I'm not entirely clear on what new_modelist actually does, it seems exclusively for a sysfs interface. Which in the end does amount to a normal fb_set_par to check the mode, but then takes a different path in both fbmem.c and fbcon.c. I have no idea why these 2 paths are different, but then I also don't really want to find out. So just do the simple conversion to a direct function call. v2: static inline for the dummy versions, I forgot. Signed-off-by:
-
Daniel Vetter authored
With the sh_mobile notifier removed we can just directly call the fbcon code here. v2: Remove now unused local variable. v3: fixup !CONFIG_FRAMEBUFFER_CONSOLE, noticed by kbuild Signed-off-by:
-
Daniel Vetter authored
This seems to be entirely defunct: - The FB_EVEN_SUSPEND/RESUME events are only sent out by fb_set_suspend. Which is supposed to be called by drivers in their suspend/resume hooks, and not itself call into drivers. Luckily sh_mob doesn't call fb_set_suspend, so this seems to do nothing useful. - The notify hook calls sh_mobile_fb_reconfig() which in turn can call into the fb notifier. Or attempt too, since that would deadlock. So looks like leftover hacks from when this was originally introduced in commit 6011bdea Author: Guennadi Liakhovetski <g.liakhovetski@gmx.de> Date: Wed Jul 21 10:13:21 2010 +0000 fbdev: sh-mobile: HDMI support for SH-Mobile SoCs So let's just remove it. Signed-off-by:
Geert Uytterhoeven <geert+renesas@glider.be> Tested-by:
-
Daniel Vetter authored
For some reasons the pm_vt_switch_unregister call was missing from the direct unregister_framebuffer path. Fix this. v2: fbinfo->dev is used to decided whether unlink_framebuffer has been called already. I botched that in v1. Make this all clearer by inlining __unlink_framebuffer. v3: Fix typoe in subject (Maarten). Signed-off-by:
-
Daniel Vetter authored
Except for driver bugs (which we'll catch with a WARN_ON) this is only to report failures of the new driver taking over the console. There's nothing the outgoing driver can do about that, and no one ever bothered to actually look at these return values. So remove them all. v2: fixup unregister_framebuffer in savagefb, fbtft, ivtvfb, and neofb drivers, reported by kbuild. Signed-off-by:
-
Daniel Vetter authored
Also remove the error return value. That's all errors for either driver bugs (trying to unbind something that isn't bound), or errors of the new driver that will take over. There's nothing the outgoing driver can do about this anyway, so switch over to void. Signed-off-by:
-
Daniel Vetter authored
Ever since commit c47747fd Author: Linus Torvalds <torvalds@linux-foundation.org> Date: Wed May 11 14:58:34 2011 -0700 fbmem: make read/write/ioctl use the frame buffer at open time fbdev has gained proper refcounting for the fbinfo attached to any open files, which means that the backing driver (stored in fb_info->fbops) cannot untimely disappear anymore. The only thing that can happen is that the entire device just outright disappears and gets unregistered, but file_fb_info does check for that. Except that it's racy - it only checks once at the start of a file_ops, there's no guarantee that the underlying fbdev won't untimely disappear. Aside: A proper way to fix that race is probably to replicate the srcu trickery we've rolled out in drm. But given that this race has existed since forever it's probably not one we need to fix right away. do_unregister_framebuffer also nowhere clears fb_info->fbops, hence the check in lock_fb_info can't possible catch a disappearing fbdev later on. Long story short: Ever since the above commit the fb_info->fbops checks have essentially become dead code. Remove this all. Aside from the file_ops callbacks, and stuff called from there there's only register/unregister code left. If that goes wrong a driver managed to register/unregister a device instance twice or in the wrong order. That's just a driver bug. v2: - fb_mmap had an open-coded version of the fbinfo->fops check, because it doesn't need the fbinfo->lock. Delete that too. - Use the wrapper function in fb_open/release now, since no difference anymore. Signed-off-by:
-
Daniel Vetter authored
It's properly protected by reboot_lock. Signed-off-by:
Daniel Vetter <daniel.vetter@ffwll.ch> Reviewed-by:
-
Daniel Vetter authored
Simply because olpc never unregisters the damn thing. It also registers the framebuffer directly by poking around in fbdev core internals, so it's all around rather broken. Signed-off-by:
-
Daniel Vetter authored
Which means lock_fb_info can never fail. Remove the error handling. Signed-off-by:
-
Daniel Vetter authored
Which means lock_fb_info can never fail. Remove the error handling. Signed-off-by:
-
Daniel Vetter authored
It's dead code, and removing it avoids me having to understand what it's doing with lock_fb_info. v2: Also remove sh_mobile_lcdc_must_reconfigure, now unused (Sam). Signed-off-by:
-
Daniel Vetter authored
With commit 6104c370 Author: Daniel Vetter <daniel.vetter@ffwll.ch> Date: Tue Aug 1 17:32:07 2017 +0200 fbcon: Make fbcon a built-time depency for fbdev we have a static dependency between fbcon and fbdev, and we can replace the indirection through the notifier chain with a function call. v2: Sam Ravnborg noticed that mach-pxa/am200epd.c has a notifier too, and listens to this. ... Looking at the code it seems to wait for some fb to show up, so that it can get the framebuffer base address from the fb_info struct. I suspect his is some firmware fbdev. Then it uses that information to let the real fbdev driver (metronomefb.c by the looks) get at the framebuffer memory. This doesn't looke like it's easy to fix (except by deleting the entire thing, seems untouched since 2008, we might be able to get away with that), so let's just stuff a few #ifdef into fb.h and fbmem.c and cry over them for a bit. Signed-off-by:
-
Daniel Vetter authored
This is unused code since commit 6104c370 Author: Daniel Vetter <daniel.vetter@ffwll.ch> Date: Tue Aug 1 17:32:07 2017 +0200 fbcon: Make fbcon a built-time depency for fbdev when fbcon was made a compile-time static dependency of fbdev. We can't exit fbcon anymore without exiting fbdev first, which only works if all fbdev drivers have unloaded already. Hence this is all dead code. v2: I missed that fbcon_exit is also called from con_deinit stuff, and there fbcon_has_exited prevents double-cleanup. But we can fix that by properly resetting con2fb_map[] to all -1, which is used everywhere else to indicate "no fb_info allocate to this console". With that change the double-cleanup (which resulted in a module refcount underflow, among other things) is prevented. Aside: con2fb_map is a signed char, so don't register more than 128 fb_info or hilarity will ensue. v3: CI showed me that I still didn't fully understand what's going on here. The leaked references in con2fb_map have been used upon rebinding the fb console in fbcon_init. It worked because fbdev unregistering still cleaned out con2fb_map, and reset it to info_idx. If the last fbdev driver unregistered, then it also reset info_idx, and unregistered the fbcon driver. Imo that's all a bit fragile, so let's keep the con2fb_map reset to -1, and in fbcon_init pick info_idx if we're starting fresh. That means unbinding and rebinding will cleanse the mapping, but why are you doing that if you want to retain the mapping, so should be fine. Also, I think info_idx == -1 is impossible in fbcon_init - we unregister the fbcon in that case. So catch&warn about that. v4: Drop unecessary assignment - I forgot to delete the first assignment of info in fbcon_init. Cc: Sam Ravnborg <sam@ravnborg.org> Reviewed-by:
-
Daniel Vetter authored
This was formerly used in fbdev drivers (not sure why, predates most git history), but now it's entirely an fbcon internal thing. Give it a more specific name. Signed-off-by:
-
Daniel Vetter authored
Motivated because it contains a struct display, which is a fbcon internal data structure that I want to rename. It seems to have been formerly used in drivers, but that's very long time ago. Signed-off-by:
-
Daniel Vetter authored
Entirely unused. Signed-off-by:
-
Daniel Vetter authored
Motivated because it contains a struct display, which is a fbcon internal data structure that I want to rename. It seems to have been formerly used in drivers, but that's very long time ago. Signed-off-by:
-
Daniel Vetter authored
I honestly have no idea what the subtle differences between con_is_visible, con_is_fg (internal to vt.c) and con_is_bound are. But it looks like both vc->vc_display_fg and con_driver_map are protected by the console_lock, so probably better if we hold that when checking this. To do that I had to deinline the con_is_visible function. Signed-off-by:
-
Daniel Vetter authored
For symmetry reasons with do_unblank_screen, except without the oops_in_progress special case. Just a drive-by annotation while I'm trying to untangle the fbcon vs. fbdev screen blank/unblank maze. Signed-off-by:
-
Daniel Vetter authored
Just drive-by, nothing systematic yet. Signed-off-by:
-
Daniel Vetter authored
As part of trying to understand the locking (or lack thereof) in the fbcon/vt/fbdev maze, annotate everything. Signed-off-by:
-
- 09 Jun, 2019 1 commit
-
-
Linus Torvalds authored
-
- 08 Jun, 2019 4 commits
-
-
git://github.com/ceph/ceph-clientLinus Torvalds authored
Pull ceph fixes from Ilya Dryomov: "A change to call iput() asynchronously to avoid a possible deadlock when iput_final() needs to wait for in-flight I/O (e.g. readahead) and a fixup for a cleanup that went into -rc1" * tag 'ceph-for-5.2-rc4' of git://github.com/ceph/ceph-client: ceph: fix error handling in ceph_get_caps() ceph: avoid iput_final() while holding mutex or in dispatch thread ceph: single workqueue for inode related works
-
git://git.kernel.org/pub/scm/linux/kernel/git/xen/tipLinus Torvalds authored
Pull xen fix from Juergen Gross: "Just one fix for the Xen block frontend driver avoiding allocations with order > 0" * tag 'for-linus-5.2b-rc4-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip: xen-blkfront: switch kcalloc to kvcalloc for large array allocation
-
git://git.kernel.org/pub/scm/linux/kernel/git/s390/linuxLinus Torvalds authored
Pull s390 fixes from Heiko Carstens: - fix stack unwinder: the stack unwinder rework has on off-by-one bug which prevents following stack backchains over more than one context (e.g. irq -> process). - fix address space detection in exception handler: if user space switches to access register mode, which is not supported anymore, the exception handler may resolve to the wrong address space. * tag 's390-5.2-4' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux: s390/unwind: correct stack switching during unwind s390/mm: fix address space detection in exception handling
-
git://git.kernel.org/pub/scm/linux/kernel/git/mips/linuxLinus Torvalds authored
Pull MIPS fixes from Paul Burton: - Declare ginvt() __always_inline due to its use of an argument as an inline asm immediate. - A VDSO build fix following Kbuild changes made this cycle. - A fix for boot failures on txx9 systems following memory initialization changes made this cycle. - Bounds check virt_addr_valid() to prevent it spuriously indicating that bogus addresses are valid, in turn fixing hardened usercopy failures that have been present since v4.12. - Build uImage.gz for pistachio systems by default, since this is the image we need in order to actually boot on a board. - Remove an unused variable in our uprobes code. * tag 'mips_fixes_5.2_1' of git://git.kernel.org/pub/scm/linux/kernel/git/mips/linux: MIPS: uprobes: remove set but not used variable 'epc' MIPS: pistachio: Build uImage.gz by default MIPS: Make virt_addr_valid() return bool MIPS: Bounds check virt_addr_valid MIPS: TXx9: Fix boot crash in free_initmem() MIPS: remove a space after -I to cope with header search paths for VDSO MIPS: mark ginvt() as __always_inline
-
