On Thu, Oct 8, 2020 at 11:25 AM Thomas Zimmermann <tzimmermann@xxxxxxx> wrote:
>
> Hi
>
> Am 02.10.20 um 20:44 schrieb Daniel Vetter:
> > On Fri, Oct 2, 2020 at 8:05 PM Daniel Vetter <daniel@xxxxxxxx> wrote:
> >>
> >> On Tue, Sep 29, 2020 at 05:14:36PM +0200, Thomas Zimmermann wrote:
> >>> At least sparc64 requires I/O-specific access to framebuffers. This
> >>> patch updates the fbdev console accordingly.
> >>>
> >>> For drivers with direct access to the framebuffer memory, the callback
> >>> functions in struct fb_ops test for the type of memory and call the rsp
> >>> fb_sys_ of fb_cfb_ functions.
> >>>
> >>> For drivers that employ a shadow buffer, fbdev's blit function retrieves
> >>> the framebuffer address as struct dma_buf_map, and uses dma_buf_map
> >>> interfaces to access the buffer.
> >>>
> >>> The bochs driver on sparc64 uses a workaround to flag the framebuffer as
> >>> I/O memory and avoid a HW exception. With the introduction of struct
> >>> dma_buf_map, this is not required any longer. The patch removes the rsp
> >>> code from both, bochs and fbdev.
> >>>
> >>> Signed-off-by: Thomas Zimmermann <tzimmermann@xxxxxxx>
> >
> > Argh, I accidentally hit send before finishing this ...
> >
> >>> ---
> >>> drivers/gpu/drm/bochs/bochs_kms.c | 1 -
> >>> drivers/gpu/drm/drm_fb_helper.c | 217 ++++++++++++++++++++++++++++--
> >>> include/drm/drm_mode_config.h | 12 --
> >>> include/linux/dma-buf-map.h | 72 ++++++++--
> >>> 4 files changed, 265 insertions(+), 37 deletions(-)
> >>>
> >>> diff --git a/drivers/gpu/drm/bochs/bochs_kms.c b/drivers/gpu/drm/bochs/bochs_kms.c
> >>> index 13d0d04c4457..853081d186d5 100644
> >>> --- a/drivers/gpu/drm/bochs/bochs_kms.c
> >>> +++ b/drivers/gpu/drm/bochs/bochs_kms.c
> >>> @@ -151,7 +151,6 @@ int bochs_kms_init(struct bochs_device *bochs)
> >>> bochs->dev->mode_config.preferred_depth = 24;
> >>> bochs->dev->mode_config.prefer_shadow = 0;
> >>> bochs->dev->mode_config.prefer_shadow_fbdev = 1;
> >>> - bochs->dev->mode_config.fbdev_use_iomem = true;
> >>> bochs->dev->mode_config.quirk_addfb_prefer_host_byte_order = true;
> >>>
> >>> bochs->dev->mode_config.funcs = &bochs_mode_funcs;
> >>> diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c
> >>> index 343a292f2c7c..f345a314a437 100644
> >>> --- a/drivers/gpu/drm/drm_fb_helper.c
> >>> +++ b/drivers/gpu/drm/drm_fb_helper.c
> >>> @@ -388,24 +388,22 @@ static void drm_fb_helper_resume_worker(struct work_struct *work)
> >>> }
> >>>
> >>> static void drm_fb_helper_dirty_blit_real(struct drm_fb_helper *fb_helper,
> >>> - struct drm_clip_rect *clip)
> >>> + struct drm_clip_rect *clip,
> >>> + struct dma_buf_map *dst)
> >>> {
> >>> struct drm_framebuffer *fb = fb_helper->fb;
> >>> unsigned int cpp = fb->format->cpp[0];
> >>> size_t offset = clip->y1 * fb->pitches[0] + clip->x1 * cpp;
> >>> void *src = fb_helper->fbdev->screen_buffer + offset;
> >>> - void *dst = fb_helper->buffer->map.vaddr + offset;
> >>> size_t len = (clip->x2 - clip->x1) * cpp;
> >>> unsigned int y;
> >>>
> >>> - for (y = clip->y1; y < clip->y2; y++) {
> >>> - if (!fb_helper->dev->mode_config.fbdev_use_iomem)
> >>> - memcpy(dst, src, len);
> >>> - else
> >>> - memcpy_toio((void __iomem *)dst, src, len);
> >>> + dma_buf_map_incr(dst, offset); /* go to first pixel within clip rect */
> >>>
> >>> + for (y = clip->y1; y < clip->y2; y++) {
> >>> + dma_buf_map_memcpy_to(dst, src, len);
> >>> + dma_buf_map_incr(dst, fb->pitches[0]);
> >>> src += fb->pitches[0];
> >>> - dst += fb->pitches[0];
> >>> }
> >>> }
> >>>
> >>> @@ -433,8 +431,9 @@ static void drm_fb_helper_dirty_work(struct work_struct *work)
> >>> ret = drm_client_buffer_vmap(helper->buffer, &map);
> >>> if (ret)
> >>> return;
> >>> - drm_fb_helper_dirty_blit_real(helper, &clip_copy);
> >>> + drm_fb_helper_dirty_blit_real(helper, &clip_copy, &map);
> >>> }
> >>> +
> >>> if (helper->fb->funcs->dirty)
> >>> helper->fb->funcs->dirty(helper->fb, NULL, 0, 0,
> >>> &clip_copy, 1);
> >>> @@ -771,6 +770,136 @@ void drm_fb_helper_sys_imageblit(struct fb_info *info,
> >>> }
> >>> EXPORT_SYMBOL(drm_fb_helper_sys_imageblit);
> >>>
> >>> +static ssize_t drm_fb_helper_cfb_read(struct fb_info *info, char __user *buf,
> >>> + size_t count, loff_t *ppos)
> >>> +{
> >>> + unsigned long p = *ppos;
> >>> + u8 *dst;
> >>> + u8 __iomem *src;
> >>> + int c, err = 0;
> >>> + unsigned long total_size;
> >>> + unsigned long alloc_size;
> >>> + ssize_t ret = 0;
> >>> +
> >>> + if (info->state != FBINFO_STATE_RUNNING)
> >>> + return -EPERM;
> >>> +
> >>> + total_size = info->screen_size;
> >>> +
> >>> + if (total_size == 0)
> >>> + total_size = info->fix.smem_len;
> >>> +
> >>> + if (p >= total_size)
> >>> + return 0;
> >>> +
> >>> + if (count >= total_size)
> >>> + count = total_size;
> >>> +
> >>> + if (count + p > total_size)
> >>> + count = total_size - p;
> >>> +
> >>> + src = (u8 __iomem *)(info->screen_base + p);
> >>> +
> >>> + alloc_size = min(count, PAGE_SIZE);
> >>> +
> >>> + dst = kmalloc(alloc_size, GFP_KERNEL);
> >>> + if (!dst)
> >>> + return -ENOMEM;
> >>> +
> >>> + while (count) {
> >>> + c = min(count, alloc_size);
> >>> +
> >>> + memcpy_fromio(dst, src, c);
> >>> + if (copy_to_user(buf, dst, c)) {
> >>> + err = -EFAULT;
> >>> + break;
> >>> + }
> >>> +
> >>> + src += c;
> >>> + *ppos += c;
> >>> + buf += c;
> >>> + ret += c;
> >>> + count -= c;
> >>> + }
> >>> +
> >>> + kfree(dst);
> >>> +
> >>> + if (err)
> >>> + return err;
> >>> +
> >>> + return ret;
> >>> +}
> >>> +
> >>> +static ssize_t drm_fb_helper_cfb_write(struct fb_info *info, const char __user *buf,
> >>> + size_t count, loff_t *ppos)
> >>> +{
> >>> + unsigned long p = *ppos;
> >>> + u8 *src;
> >>> + u8 __iomem *dst;
> >>> + int c, err = 0;
> >>> + unsigned long total_size;
> >>> + unsigned long alloc_size;
> >>> + ssize_t ret = 0;
> >>> +
> >>> + if (info->state != FBINFO_STATE_RUNNING)
> >>> + return -EPERM;
> >>> +
> >>> + total_size = info->screen_size;
> >>> +
> >>> + if (total_size == 0)
> >>> + total_size = info->fix.smem_len;
> >>> +
> >>> + if (p > total_size)
> >>> + return -EFBIG;
> >>> +
> >>> + if (count > total_size) {
> >>> + err = -EFBIG;
> >>> + count = total_size;
> >>> + }
> >>> +
> >>> + if (count + p > total_size) {
> >>> + /*
> >>> + * The framebuffer is too small. We do the
> >>> + * copy operation, but return an error code
> >>> + * afterwards. Taken from fbdev.
> >>> + */
> >>> + if (!err)
> >>> + err = -ENOSPC;
> >>> + count = total_size - p;
> >>> + }
> >>> +
> >>> + alloc_size = min(count, PAGE_SIZE);
> >>> +
> >>> + src = kmalloc(alloc_size, GFP_KERNEL);
> >>> + if (!src)
> >>> + return -ENOMEM;
> >>> +
> >>> + dst = (u8 __iomem *)(info->screen_base + p);
> >>> +
> >>> + while (count) {
> >>> + c = min(count, alloc_size);
> >>> +
> >>> + if (copy_from_user(src, buf, c)) {
> >>> + err = -EFAULT;
> >>> + break;
> >>> + }
> >>> + memcpy_toio(dst, src, c);
> >>> +
> >>> + dst += c;
> >>> + *ppos += c;
> >>> + buf += c;
> >>> + ret += c;
> >>> + count -= c;
> >>> + }
> >>> +
> >>> + kfree(src);
> >>> +
> >>> + if (err)
> >>> + return err;
> >>> +
> >>> + return ret;
> >>> +}
> >
> > The duplication is a bit annoying here, but can't really be avoided. I
> > do think though we should maybe go a bit further, and have drm
> > implementations of this stuff instead of following fbdev concepts as
> > closely as possible. So here roughly:
> >
> > - if we have a shadow fb, construct a dma_buf_map for that, otherwise
> > take the one from the driver
> > - have a full generic implementation using that one directly (and
> > checking size limits against the underlying gem buffer)
> > - ideally also with some testcases in the fbdev testcase we have (very
> > bare-bones right now) in igt
> >
> > But I'm not really sure whether that's worth all the trouble. It's
> > just that the fbdev-ness here in this copied code sticks out a lot :-)
> >
> >>> +
> >>> /**
> >>> * drm_fb_helper_cfb_fillrect - wrapper around cfb_fillrect
> >>> * @info: fbdev registered by the helper
> >>> @@ -2043,6 +2172,66 @@ static int drm_fbdev_fb_mmap(struct fb_info *info, struct vm_area_struct *vma)
> >>> return -ENODEV;
> >>> }
> >>>
> >>> +static ssize_t drm_fbdev_fb_read(struct fb_info *info, char __user *buf,
> >>> + size_t count, loff_t *ppos)
> >>> +{
> >>> + struct drm_fb_helper *fb_helper = info->par;
> >>> + struct drm_client_buffer *buffer = fb_helper->buffer;
> >>> +
> >>> + if (drm_fbdev_use_shadow_fb(fb_helper) || !buffer->map.is_iomem)
> >>> + return drm_fb_helper_sys_read(info, buf, count, ppos);
> >>> + else
> >>> + return drm_fb_helper_cfb_read(info, buf, count, ppos);
> >>> +}
> >>> +
> >>> +static ssize_t drm_fbdev_fb_write(struct fb_info *info, const char __user *buf,
> >>> + size_t count, loff_t *ppos)
> >>> +{
> >>> + struct drm_fb_helper *fb_helper = info->par;
> >>> + struct drm_client_buffer *buffer = fb_helper->buffer;
> >>> +
> >>> + if (drm_fbdev_use_shadow_fb(fb_helper) || !buffer->map.is_iomem)
> >>> + return drm_fb_helper_sys_write(info, buf, count, ppos);
> >>> + else
> >>> + return drm_fb_helper_cfb_write(info, buf, count, ppos);
> >>> +}
> >>> +
> >>> +static void drm_fbdev_fb_fillrect(struct fb_info *info,
> >>> + const struct fb_fillrect *rect)
> >>> +{
> >>> + struct drm_fb_helper *fb_helper = info->par;
> >>> + struct drm_client_buffer *buffer = fb_helper->buffer;
> >>> +
> >>> + if (drm_fbdev_use_shadow_fb(fb_helper) || !buffer->map.is_iomem)
> >>> + drm_fb_helper_sys_fillrect(info, rect);
> >>> + else
> >>> + drm_fb_helper_cfb_fillrect(info, rect);
> >>> +}
> >>> +
> >>> +static void drm_fbdev_fb_copyarea(struct fb_info *info,
> >>> + const struct fb_copyarea *area)
> >>> +{
> >>> + struct drm_fb_helper *fb_helper = info->par;
> >>> + struct drm_client_buffer *buffer = fb_helper->buffer;
> >>> +
> >>> + if (drm_fbdev_use_shadow_fb(fb_helper) || !buffer->map.is_iomem)
> >>> + drm_fb_helper_sys_copyarea(info, area);
> >>> + else
> >>> + drm_fb_helper_cfb_copyarea(info, area);
> >>> +}
> >>> +
> >>> +static void drm_fbdev_fb_imageblit(struct fb_info *info,
> >>> + const struct fb_image *image)
> >>> +{
> >>> + struct drm_fb_helper *fb_helper = info->par;
> >>> + struct drm_client_buffer *buffer = fb_helper->buffer;
> >>> +
> >>> + if (drm_fbdev_use_shadow_fb(fb_helper) || !buffer->map.is_iomem)
> >>> + drm_fb_helper_sys_imageblit(info, image);
> >>> + else
> >>> + drm_fb_helper_cfb_imageblit(info, image);
> >>> +}
> >
> > I think a todo.rst entry to make the new generic functions the real ones, and
> > drivers not using the sys/cfb ones anymore would be a good addition.
> > It's kinda covered by the move to the generic helpers, but maybe we
> > can convert a few more drivers over to these here. Would also allow us
> > to maybe flatten the code a bit and use more of the dma_buf_map stuff
> > directly (instead of reusing crusty fbdev code written 20 years ago or
> > so).
>
> I wouldn't mind doing our own thing, but dma_buf_map is not a good fit
> here. Mostly because the _cfb_ code first does a reads from I/O to
> system memory, and then copies to userspace. The _sys_ functions copy
> directly to userspace. (Same for write, but in the other direction.)
>
> There's some code at the top and bottom of these functions that could be
> shared. If we want to share the copy loops, we'd probably end up with
> additional memcpys in the _sys_ case.
Yeah I noticed that. I'd just ignore it. If someone is using a) fbdev
and b) read/write on it, they don't care much about performance. We
can do another copy or two, no problem. But the duplication is also ok
I guess, just a bit less pretty.
-Daniel
> Best regards
> Thomas
>
> >
> >>> +
> >>> static const struct fb_ops drm_fbdev_fb_ops = {
> >>> .owner = THIS_MODULE,
> >>> DRM_FB_HELPER_DEFAULT_OPS,
> >>> @@ -2050,11 +2239,11 @@ static const struct fb_ops drm_fbdev_fb_ops = {
> >>> .fb_release = drm_fbdev_fb_release,
> >>> .fb_destroy = drm_fbdev_fb_destroy,
> >>> .fb_mmap = drm_fbdev_fb_mmap,
> >>> - .fb_read = drm_fb_helper_sys_read,
> >>> - .fb_write = drm_fb_helper_sys_write,
> >>> - .fb_fillrect = drm_fb_helper_sys_fillrect,
> >>> - .fb_copyarea = drm_fb_helper_sys_copyarea,
> >>> - .fb_imageblit = drm_fb_helper_sys_imageblit,
> >>> + .fb_read = drm_fbdev_fb_read,
> >>> + .fb_write = drm_fbdev_fb_write,
> >>> + .fb_fillrect = drm_fbdev_fb_fillrect,
> >>> + .fb_copyarea = drm_fbdev_fb_copyarea,
> >>> + .fb_imageblit = drm_fbdev_fb_imageblit,
> >>> };
> >>>
> >>> static struct fb_deferred_io drm_fbdev_defio = {
> >>> diff --git a/include/drm/drm_mode_config.h b/include/drm/drm_mode_config.h
> >>> index 5ffbb4ed5b35..ab424ddd7665 100644
> >>> --- a/include/drm/drm_mode_config.h
> >>> +++ b/include/drm/drm_mode_config.h
> >>> @@ -877,18 +877,6 @@ struct drm_mode_config {
> >>> */
> >>> bool prefer_shadow_fbdev;
> >>>
> >>> - /**
> >>> - * @fbdev_use_iomem:
> >>> - *
> >>> - * Set to true if framebuffer reside in iomem.
> >>> - * When set to true memcpy_toio() is used when copying the framebuffer in
> >>> - * drm_fb_helper.drm_fb_helper_dirty_blit_real().
> >>> - *
> >>> - * FIXME: This should be replaced with a per-mapping is_iomem
> >>> - * flag (like ttm does), and then used everywhere in fbdev code.
> >>> - */
> >>> - bool fbdev_use_iomem;
> >>> -
> >>> /**
> >>> * @quirk_addfb_prefer_xbgr_30bpp:
> >>> *
> >>> diff --git a/include/linux/dma-buf-map.h b/include/linux/dma-buf-map.h
> >
> > I think the below should be split out as a prep patch.
> >
> >>> index 2e8bbecb5091..6ca0f304dda2 100644
> >>> --- a/include/linux/dma-buf-map.h
> >>> +++ b/include/linux/dma-buf-map.h
> >>> @@ -32,6 +32,14 @@
> >>> * accessing the buffer. Use the returned instance and the helper functions
> >>> * to access the buffer's memory in the correct way.
> >>> *
> >>> + * The type :c:type:`struct dma_buf_map <dma_buf_map>` and its helpers are
> >>> + * actually independent from the dma-buf infrastructure. When sharing buffers
> >>> + * among devices, drivers have to know the location of the memory to access
> >>> + * the buffers in a safe way. :c:type:`struct dma_buf_map <dma_buf_map>`
> >>> + * solves this problem for dma-buf and its users. If other drivers or
> >>> + * sub-systems require similar functionality, the type could be generalized
> >>> + * and moved to a more prominent header file.
> >>> + *
> >>> * Open-coding access to :c:type:`struct dma_buf_map <dma_buf_map>` is
> >>> * considered bad style. Rather then accessing its fields directly, use one
> >>> * of the provided helper functions, or implement your own. For example,
> >>> @@ -51,6 +59,14 @@
> >>> *
> >>> * dma_buf_map_set_vaddr_iomem(&map. 0xdeadbeaf);
> >>> *
> >>> + * Instances of struct dma_buf_map do not have to be cleaned up, but
> >>> + * can be cleared to NULL with dma_buf_map_clear(). Cleared mappings
> >>> + * always refer to system memory.
> >>> + *
> >>> + * .. code-block:: c
> >>> + *
> >>> + * dma_buf_map_clear(&map);
> >>> + *
> >>> * Test if a mapping is valid with either dma_buf_map_is_set() or
> >>> * dma_buf_map_is_null().
> >>> *
> >>> @@ -73,17 +89,19 @@
> >>> * if (dma_buf_map_is_equal(&sys_map, &io_map))
> >>> * // always false
> >>> *
> >>> - * Instances of struct dma_buf_map do not have to be cleaned up, but
> >>> - * can be cleared to NULL with dma_buf_map_clear(). Cleared mappings
> >>> - * always refer to system memory.
> >>> + * A set up instance of struct dma_buf_map can be used to access or manipulate
> >>> + * the buffer memory. Depending on the location of the memory, the provided
> >>> + * helpers will pick the correct operations. Data can be copied into the memory
> >>> + * with dma_buf_map_memcpy_to(). The address can be manipulated with
> >>> + * dma_buf_map_incr().
> >>> *
> >>> - * The type :c:type:`struct dma_buf_map <dma_buf_map>` and its helpers are
> >>> - * actually independent from the dma-buf infrastructure. When sharing buffers
> >>> - * among devices, drivers have to know the location of the memory to access
> >>> - * the buffers in a safe way. :c:type:`struct dma_buf_map <dma_buf_map>`
> >>> - * solves this problem for dma-buf and its users. If other drivers or
> >>> - * sub-systems require similar functionality, the type could be generalized
> >>> - * and moved to a more prominent header file.
> >>> + * .. code-block:: c
> >>> + *
> >>> + * const void *src = ...; // source buffer
> >>> + * size_t len = ...; // length of src
> >>> + *
> >>> + * dma_buf_map_memcpy_to(&map, src, len);
> >>> + * dma_buf_map_incr(&map, len); // go to first byte after the memcpy
> >>> */
> >>>
> >>> /**
> >>> @@ -210,4 +228,38 @@ static inline void dma_buf_map_clear(struct dma_buf_map *map)
> >>> }
> >>> }
> >>>
> >>> +/**
> >>> + * dma_buf_map_memcpy_to - Memcpy into dma-buf mapping
> >>> + * @dst: The dma-buf mapping structure
> >>> + * @src: The source buffer
> >>> + * @len: The number of byte in src
> >>> + *
> >>> + * Copies data into a dma-buf mapping. The source buffer is in system
> >>> + * memory. Depending on the buffer's location, the helper picks the correct
> >>> + * method of accessing the memory.
> >>> + */
> >>> +static inline void dma_buf_map_memcpy_to(struct dma_buf_map *dst, const void *src, size_t len)
> >>> +{
> >>> + if (dst->is_iomem)
> >>> + memcpy_toio(dst->vaddr_iomem, src, len);
> >>> + else
> >>> + memcpy(dst->vaddr, src, len);
> >>> +}
> >>> +
> >>> +/**
> >>> + * dma_buf_map_incr - Increments the address stored in a dma-buf mapping
> >>> + * @map: The dma-buf mapping structure
> >>> + * @incr: The number of bytes to increment
> >>> + *
> >>> + * Increments the address stored in a dma-buf mapping. Depending on the
> >>> + * buffer's location, the correct value will be updated.
> >>> + */
> >>> +static inline void dma_buf_map_incr(struct dma_buf_map *map, size_t incr)
> >>> +{
> >>> + if (map->is_iomem)
> >>> + map->vaddr_iomem += incr;
> >>> + else
> >>> + map->vaddr += incr;
> >>> +}
> >>> +
> >>> #endif /* __DMA_BUF_MAP_H__ */
> >>> --
> >>> 2.28.0
> >
> > Aside from the details I think looks all reasonable.
> > -Daniel
> >
>
> --
> Thomas Zimmermann
> Graphics Driver Developer
> SUSE Software Solutions Germany GmbH
> Maxfeldstr. 5, 90409 Nürnberg, Germany
> (HRB 36809, AG Nürnberg)
> Geschäftsführer: Felix Imendörffer
>
--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
