| Message ID | 20160824082738.703060-1-arnd@arndb.de (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Wed, Aug 24, 2016 at 10:27 AM, Arnd Bergmann <arnd@arndb.de> wrote: > A cleanup in v4.8 dropped several dependencies, but caused the udl driver > to fail building when CONFIG_FB_DEFERRED_IO is not set: > > drivers/gpu/drm/udl/udl_fb.c: In function 'udl_fb_open': > drivers/gpu/drm/udl/udl_fb.c:206:23: error: 'struct fb_info' has no member named 'fbdefio' > if (fb_defio && (info->fbdefio == NULL)) { > ^~ > > This adds an #ifdef check around the code that fails. We know this > is safe because it always checks info->fbdefio for being non-NULL, > and that member is not defined here. > > Signed-off-by: Arnd Bergmann <arnd@arndb.de> > Fixes: c0f0c177b5a9 ("drm/udl: make fbdev support really optional") There's a pile more needed, which is why I just ended up stuffing them all into linux-next for 4.9. Is your patch still needed on that? If we really want this fixed in 4.8 I guess I can assemble a cherry-pick list. -Daniel > --- > drivers/gpu/drm/udl/udl_fb.c | 9 ++++++++- > 1 file changed, 8 insertions(+), 1 deletion(-) > > diff --git a/drivers/gpu/drm/udl/udl_fb.c b/drivers/gpu/drm/udl/udl_fb.c > index d5df555aeba0..fec47733e57e 100644 > --- a/drivers/gpu/drm/udl/udl_fb.c > +++ b/drivers/gpu/drm/udl/udl_fb.c > @@ -24,11 +24,14 @@ > > #define DL_DEFIO_WRITE_DELAY (HZ/20) /* fb_deferred_io.delay in jiffies */ > > -static int fb_defio = 0; /* Optionally enable experimental fb_defio mmap support */ > static int fb_bpp = 16; > > module_param(fb_bpp, int, S_IWUSR | S_IRUSR | S_IWGRP | S_IRGRP); > + > +#ifdef CONFIG_FB_DEFERRED_IO > +static int fb_defio = 0; /* Optionally enable experimental fb_defio mmap support */ > module_param(fb_defio, int, S_IWUSR | S_IRUSR | S_IWGRP | S_IRGRP); > +#endif > > struct udl_fbdev { > struct drm_fb_helper helper; > @@ -203,6 +206,7 @@ static int udl_fb_open(struct fb_info *info, int user) > > ufbdev->fb_count++; > > +#ifdef CONFIG_FB_DEFERRED_IO > if (fb_defio && (info->fbdefio == NULL)) { > /* enable defio at last moment if not disabled by client */ > > @@ -218,6 +222,7 @@ static int udl_fb_open(struct fb_info *info, int user) > info->fbdefio = fbdefio; > fb_deferred_io_init(info); > } > +#endif > > pr_notice("open /dev/fb%d user=%d fb_info=%p count=%d\n", > info->node, user, info, ufbdev->fb_count); > @@ -235,12 +240,14 @@ static int udl_fb_release(struct fb_info *info, int user) > > ufbdev->fb_count--; > > +#ifdef CONFIG_FB_DEFERRED_IO > if ((ufbdev->fb_count == 0) && (info->fbdefio)) { > fb_deferred_io_cleanup(info); > kfree(info->fbdefio); > info->fbdefio = NULL; > info->fbops->fb_mmap = udl_fb_mmap; > } > +#endif > > pr_warn("released /dev/fb%d user=%d count=%d\n", > info->node, user, ufbdev->fb_count); > -- > 2.9.0 >
On Wednesday, August 24, 2016 11:01:17 AM CEST Daniel Vetter wrote: > On Wed, Aug 24, 2016 at 10:27 AM, Arnd Bergmann <arnd@arndb.de> wrote: > > A cleanup in v4.8 dropped several dependencies, but caused the udl driver > > to fail building when CONFIG_FB_DEFERRED_IO is not set: > > > > drivers/gpu/drm/udl/udl_fb.c: In function 'udl_fb_open': > > drivers/gpu/drm/udl/udl_fb.c:206:23: error: 'struct fb_info' has no member named 'fbdefio' > > if (fb_defio && (info->fbdefio == NULL)) { > > ^~ > > > > This adds an #ifdef check around the code that fails. We know this > > is safe because it always checks info->fbdefio for being non-NULL, > > and that member is not defined here. > > > > Signed-off-by: Arnd Bergmann <arnd@arndb.de> > > Fixes: c0f0c177b5a9 ("drm/udl: make fbdev support really optional") > > There's a pile more needed, which is why I just ended up stuffing them > all into linux-next for 4.9. Is your patch still needed on that? If we > really want this fixed in 4.8 I guess I can assemble a cherry-pick > list. Ok, got it. I think your 0c756134cf76 ("drm: Protect fb_defio in drivers with CONFIG_KMS_FBDEV_EMULATION") is actually sufficient for the build break, it's almost the same as my patch anyway. Alternatively, reverting c0f0c177b5a9 ("drm/udl: make fbdev support really optional") for v4.8 would work too. Arnd
On Wed, Aug 24, 2016 at 12:02 PM, Arnd Bergmann <arnd@arndb.de> wrote: > On Wednesday, August 24, 2016 11:01:17 AM CEST Daniel Vetter wrote: >> On Wed, Aug 24, 2016 at 10:27 AM, Arnd Bergmann <arnd@arndb.de> wrote: >> > A cleanup in v4.8 dropped several dependencies, but caused the udl driver >> > to fail building when CONFIG_FB_DEFERRED_IO is not set: >> > >> > drivers/gpu/drm/udl/udl_fb.c: In function 'udl_fb_open': >> > drivers/gpu/drm/udl/udl_fb.c:206:23: error: 'struct fb_info' has no member named 'fbdefio' >> > if (fb_defio && (info->fbdefio == NULL)) { >> > ^~ >> > >> > This adds an #ifdef check around the code that fails. We know this >> > is safe because it always checks info->fbdefio for being non-NULL, >> > and that member is not defined here. >> > >> > Signed-off-by: Arnd Bergmann <arnd@arndb.de> >> > Fixes: c0f0c177b5a9 ("drm/udl: make fbdev support really optional") >> >> There's a pile more needed, which is why I just ended up stuffing them >> all into linux-next for 4.9. Is your patch still needed on that? If we >> really want this fixed in 4.8 I guess I can assemble a cherry-pick >> list. > > Ok, got it. I think your 0c756134cf76 ("drm: Protect fb_defio in > drivers with CONFIG_KMS_FBDEV_EMULATION") is actually sufficient > for the build break, it's almost the same as my patch anyway. Dave, can you please cherry-pick that one from -next to -fixes? Thanks, Daniel
diff --git a/drivers/gpu/drm/udl/udl_fb.c b/drivers/gpu/drm/udl/udl_fb.c index d5df555aeba0..fec47733e57e 100644 --- a/drivers/gpu/drm/udl/udl_fb.c +++ b/drivers/gpu/drm/udl/udl_fb.c @@ -24,11 +24,14 @@ #define DL_DEFIO_WRITE_DELAY (HZ/20) /* fb_deferred_io.delay in jiffies */ -static int fb_defio = 0; /* Optionally enable experimental fb_defio mmap support */ static int fb_bpp = 16; module_param(fb_bpp, int, S_IWUSR | S_IRUSR | S_IWGRP | S_IRGRP); + +#ifdef CONFIG_FB_DEFERRED_IO +static int fb_defio = 0; /* Optionally enable experimental fb_defio mmap support */ module_param(fb_defio, int, S_IWUSR | S_IRUSR | S_IWGRP | S_IRGRP); +#endif struct udl_fbdev { struct drm_fb_helper helper; @@ -203,6 +206,7 @@ static int udl_fb_open(struct fb_info *info, int user) ufbdev->fb_count++; +#ifdef CONFIG_FB_DEFERRED_IO if (fb_defio && (info->fbdefio == NULL)) { /* enable defio at last moment if not disabled by client */ @@ -218,6 +222,7 @@ static int udl_fb_open(struct fb_info *info, int user) info->fbdefio = fbdefio; fb_deferred_io_init(info); } +#endif pr_notice("open /dev/fb%d user=%d fb_info=%p count=%d\n", info->node, user, info, ufbdev->fb_count); @@ -235,12 +240,14 @@ static int udl_fb_release(struct fb_info *info, int user) ufbdev->fb_count--; +#ifdef CONFIG_FB_DEFERRED_IO if ((ufbdev->fb_count == 0) && (info->fbdefio)) { fb_deferred_io_cleanup(info); kfree(info->fbdefio); info->fbdefio = NULL; info->fbops->fb_mmap = udl_fb_mmap; } +#endif pr_warn("released /dev/fb%d user=%d count=%d\n", info->node, user, ufbdev->fb_count);
A cleanup in v4.8 dropped several dependencies, but caused the udl driver to fail building when CONFIG_FB_DEFERRED_IO is not set: drivers/gpu/drm/udl/udl_fb.c: In function 'udl_fb_open': drivers/gpu/drm/udl/udl_fb.c:206:23: error: 'struct fb_info' has no member named 'fbdefio' if (fb_defio && (info->fbdefio == NULL)) { ^~ This adds an #ifdef check around the code that fails. We know this is safe because it always checks info->fbdefio for being non-NULL, and that member is not defined here. Signed-off-by: Arnd Bergmann <arnd@arndb.de> Fixes: c0f0c177b5a9 ("drm/udl: make fbdev support really optional") --- drivers/gpu/drm/udl/udl_fb.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-)