Skip to contentSkip to conversation

drm: qemu hardware patches

PATCH Applied Gerd Hoffmann 3 participants 10 messages Oct 11, 2013 – Nov 6, 2013

Subsystems (3)
lore
Share & tools
lore

Gerd Hoffmann

author maintainer
#0 lore raw

cover drm: qemu hardware patches

Hi,

Here comes a collection of drm patches for qemu emulated virtual graphics cards. Small improvements for cirrus and qxl. A new kms driver for the qemu standard vga.

Patches 1-4 can be queued up for merge, unless someone finds bugs / problems in review of course.

Patches 5+6 have known issues and are still work-in-progress. I'm posing them for early review nevertheless.

git tree with these patches is here:
git://git.kraxel.org/linux qemu-drm

cheers,
Gerd

Gerd Hoffmann (6):
drm: add drm_set_preferred_mode
drm/cirrus: use drm_set_preferred_mode
drm/qxl: support 64bit surface bar
drm/qxl: add some surface memory logging
[wip] drm/qxl: request regions
[wip] drm/bochs: new driver

 drivers/gpu/drm/Kconfig              |   2 +
 drivers/gpu/drm/Makefile             |   1 +
 drivers/gpu/drm/bochs/Kconfig        |  11 +
 drivers/gpu/drm/bochs/Makefile       |   4 +
 drivers/gpu/drm/bochs/bochs.h        | 167 +++++++++++
 drivers/gpu/drm/bochs/bochs_drv.c    | 179 ++++++++++++
 drivers/gpu/drm/bochs/bochs_fbdev.c  | 215 ++++++++++++++
 drivers/gpu/drm/bochs/bochs_hw.c     | 162 ++++++++++
 drivers/gpu/drm/bochs/bochs_kms.c    | 294 +++++++++++++++++++
 drivers/gpu/drm/bochs/bochs_mm.c     | 552 +++++++++++++++++++++++++++++++++++
 drivers/gpu/drm/cirrus/cirrus_mode.c |  11 +-
 drivers/gpu/drm/drm_edid.c           |  13 +
 drivers/gpu/drm/qxl/qxl_kms.c        |  48 ++-
 drivers/gpu/drm/qxl/qxl_ttm.c        |   2 +
 include/drm/drm_crtc.h               |   2 +
 15 files changed, 1647 insertions(+), 16 deletions(-)
 create mode 100644 drivers/gpu/drm/bochs/Kconfig
 create mode 100644 drivers/gpu/drm/bochs/Makefile
 create mode 100644 drivers/gpu/drm/bochs/bochs.h
 create mode 100644 drivers/gpu/drm/bochs/bochs_drv.c
 create mode 100644 drivers/gpu/drm/bochs/bochs_fbdev.c
 create mode 100644 drivers/gpu/drm/bochs/bochs_hw.c
 create mode 100644 drivers/gpu/drm/bochs/bochs_kms.c
 create mode 100644 drivers/gpu/drm/bochs/bochs_mm.c
Signature · 1 line
1.8.3.1
author maintainer
#1 lore raw

PATCH 1/6 drm: add drm_set_preferred_mode

New helper function to set the preferred video mode. Can be called after drm_add_modes_noedid if you don't want the largest supported video mode be used by default.

---
 drivers/gpu/drm/drm_edid.c | 13 +++++++++++++
 include/drm/drm_crtc.h     |  2 ++
 2 files changed, 15 insertions(+)
drivers/gpu/drm/drm_edid.c +13 -0
index 1688ff5..a635034 100644
@@ -3286,6 +3286,19 @@ int drm_add_modes_noedid(struct drm_connector *connector,
 }
 EXPORT_SYMBOL(drm_add_modes_noedid);
 
+void drm_set_preferred_mode(struct drm_connector *connector,
+			   int hpref, int vpref)
+{
+	struct drm_display_mode *mode;
+
+	list_for_each_entry(mode, &connector->probed_modes, head) {
+		if (drm_mode_width(mode)  == hpref &&
+		    drm_mode_height(mode) == vpref)
+			mode->type |= DRM_MODE_TYPE_PREFERRED;
+	}
+}
+EXPORT_SYMBOL(drm_set_preferred_mode);
+
 /**
  * drm_hdmi_avi_infoframe_from_display_mode() - fill an HDMI AVI infoframe with
  *                                              data from a DRM display mode
include/drm/drm_crtc.h +2 -0
index 24f4995..ec5d737 100644
@@ -1108,6 +1108,8 @@ extern struct drm_display_mode *drm_gtf_mode_complex(struct drm_device *dev,
 				int GTF_2C, int GTF_K, int GTF_2J);
 extern int drm_add_modes_noedid(struct drm_connector *connector,
 				int hdisplay, int vdisplay);
+extern void drm_set_preferred_mode(struct drm_connector *connector,
+				   int hpref, int vpref);
 
 extern int drm_edid_header_is_valid(const u8 *raw_edid);
 extern bool drm_edid_block_valid(u8 *raw_edid, int block, bool print_bad_edid);
Signature · 1 line
1.8.3.1
author maintainer
#2 lore raw

PATCH 2/6 drm/cirrus: use drm_set_preferred_mode

In reply to Gerd Hoffmann

Explicitly set 1024x768 as default mode, so the display doesn't come up with the largest supported mode.

While being at it drop first three drm_add_modes_noedid calls. As drm_add_modes_noedid fills the mode list with modes from the database *up to* the specified size it is pretty pointless to call it multiple times with different sizes.

---
 drivers/gpu/drm/cirrus/cirrus_mode.c | 11 +++++------
 1 file changed, 5 insertions(+), 6 deletions(-)
drivers/gpu/drm/cirrus/cirrus_mode.c +5 -6
index 60685b2..adabc3d 100644
@@ -494,13 +494,12 @@ static struct drm_encoder *cirrus_encoder_init(struct drm_device *dev)
 
 int cirrus_vga_get_modes(struct drm_connector *connector)
 {
-	/* Just add a static list of modes */
-	drm_add_modes_noedid(connector, 640, 480);
-	drm_add_modes_noedid(connector, 800, 600);
-	drm_add_modes_noedid(connector, 1024, 768);
-	drm_add_modes_noedid(connector, 1280, 1024);
+	int count;
 
-	return 4;
+	/* Just add a static list of modes */
+	count = drm_add_modes_noedid(connector, 1280, 1024);
+	drm_set_preferred_mode(connector, 1024, 768);
+	return count;
 }
 
 static int cirrus_vga_mode_valid(struct drm_connector *connector,
Signature · 1 line
1.8.3.1
author maintainer
#3 lore raw

PATCH 3/6 drm/qxl: support 64bit surface bar

In reply to Gerd Hoffmann

qxl devices can have a 64bit surface bar, which is quite handy if you need a bit more surface memory. So try to use it if it is present. Note that this bar might be mapped above 4g.

QEMU command line to check that out:

    qemu-system-x86_64 -m 4g \
        -vga qxl -global qxl-vga.vram64_size_mb=512 \
        $otheroptions
---
 drivers/gpu/drm/qxl/qxl_kms.c | 32 +++++++++++++++++++++++++-------
 1 file changed, 25 insertions(+), 7 deletions(-)
drivers/gpu/drm/qxl/qxl_kms.c +25 -7
index 9e8da9e..e0ddd5b 100644
@@ -120,7 +120,7 @@ int qxl_device_init(struct qxl_device *qdev,
 		    struct pci_dev *pdev,
 		    unsigned long flags)
 {
-	int r;
+	int r, sb;
 
 	qdev->dev = &pdev->dev;
 	qdev->ddev = ddev;
@@ -136,21 +136,39 @@ int qxl_device_init(struct qxl_device *qdev,
 	qdev->rom_base = pci_resource_start(pdev, 2);
 	qdev->rom_size = pci_resource_len(pdev, 2);
 	qdev->vram_base = pci_resource_start(pdev, 0);
-	qdev->surfaceram_base = pci_resource_start(pdev, 1);
-	qdev->surfaceram_size = pci_resource_len(pdev, 1);
 	qdev->io_base = pci_resource_start(pdev, 3);
 
 	qdev->vram_mapping = io_mapping_create_wc(qdev->vram_base, pci_resource_len(pdev, 0));
-	qdev->surface_mapping = io_mapping_create_wc(qdev->surfaceram_base, qdev->surfaceram_size);
-	DRM_DEBUG_KMS("qxl: vram %llx-%llx(%dM %dk), surface %llx-%llx(%dM %dk)\n",
+
+	if (pci_resource_len(pdev, 4) > 0) {
+		/* 64bit surface bar present */
+		sb = 4;
+		qdev->surfaceram_base = pci_resource_start(pdev, sb);
+		qdev->surfaceram_size = pci_resource_len(pdev, sb);
+		qdev->surface_mapping =
+			io_mapping_create_wc(qdev->surfaceram_base,
+					     qdev->surfaceram_size);
+	}
+	if (qdev->surface_mapping == NULL) {
+		/* 64bit surface bar not present (or mapping failed) */
+		sb = 1;
+		qdev->surfaceram_base = pci_resource_start(pdev, sb);
+		qdev->surfaceram_size = pci_resource_len(pdev, sb);
+		qdev->surface_mapping =
+			io_mapping_create_wc(qdev->surfaceram_base,
+					     qdev->surfaceram_size);
+	}
+
+	DRM_DEBUG_KMS("qxl: vram %llx-%llx(%dM %dk), surface %llx-%llx(%dM %dk, %s)\n",
 		 (unsigned long long)qdev->vram_base,
 		 (unsigned long long)pci_resource_end(pdev, 0),
 		 (int)pci_resource_len(pdev, 0) / 1024 / 1024,
 		 (int)pci_resource_len(pdev, 0) / 1024,
 		 (unsigned long long)qdev->surfaceram_base,
-		 (unsigned long long)pci_resource_end(pdev, 1),
+		 (unsigned long long)pci_resource_end(pdev, sb),
 		 (int)qdev->surfaceram_size / 1024 / 1024,
-		 (int)qdev->surfaceram_size / 1024);
+		 (int)qdev->surfaceram_size / 1024,
+		 (sb == 4) ? "64bit" : "32bit");
 
 	qdev->rom = ioremap(qdev->rom_base, qdev->rom_size);
 	if (!qdev->rom) {
Signature · 1 line
1.8.3.1
author maintainer
#4 lore raw

PATCH 4/6 drm/qxl: add some surface memory logging

In reply to Gerd Hoffmann
---
 drivers/gpu/drm/qxl/qxl_kms.c | 10 +++++++---
 drivers/gpu/drm/qxl/qxl_ttm.c |  2 ++
 2 files changed, 9 insertions(+), 3 deletions(-)
drivers/gpu/drm/qxl/qxl_kms.c +7 -3
index e0ddd5b..e5ca498 100644
@@ -248,9 +248,13 @@ int qxl_device_init(struct qxl_device *qdev,
 	qdev->surfaces_mem_slot = setup_slot(qdev, 1,
 		(unsigned long)qdev->surfaceram_base,
 		(unsigned long)qdev->surfaceram_base + qdev->surfaceram_size);
-	DRM_INFO("main mem slot %d [%lx,%x)\n",
-		qdev->main_mem_slot,
-		(unsigned long)qdev->vram_base, qdev->rom->ram_header_offset);
+	DRM_INFO("main mem slot %d [%lx,%x]\n",
+		 qdev->main_mem_slot,
+		 (unsigned long)qdev->vram_base, qdev->rom->ram_header_offset);
+	DRM_INFO("surface mem slot %d [%lx,%lx]\n",
+		 qdev->surfaces_mem_slot,
+		 (unsigned long)qdev->surfaceram_base,
+		 (unsigned long)qdev->surfaceram_size);
 
 
 	qdev->gc_queue = create_singlethread_workqueue("qxl_gc");
drivers/gpu/drm/qxl/qxl_ttm.c +2 -0
index 037786d..c7e7e65 100644
@@ -516,6 +516,8 @@ int qxl_ttm_init(struct qxl_device *qdev)
 		 (unsigned)qdev->vram_size / (1024 * 1024));
 	DRM_INFO("qxl: %luM of IO pages memory ready (VRAM domain)\n",
 		 ((unsigned)num_io_pages * PAGE_SIZE) / (1024 * 1024));
+	DRM_INFO("qxl: %uM of Surface memory size\n",
+		 (unsigned)qdev->surfaceram_size / (1024 * 1024));
 	if (unlikely(qdev->mman.bdev.dev_mapping == NULL))
 		qdev->mman.bdev.dev_mapping = qdev->ddev->dev_mapping;
 	r = qxl_ttm_debugfs_init(qdev);
Signature · 1 line
1.8.3.1
author maintainer
#5 lore raw

PATCH 5/6 drm/qxl: request regions

In reply to Gerd Hoffmann

So they show up in /proc/{iomem,ioports}.

No error checking (yet?). Doesn't make things
worse than they are now, but still not nice.

Known issue: Doesn't work for qxl-vram (vesafb conflict?).

---
 drivers/gpu/drm/qxl/qxl_kms.c | 6 ++++++
 1 file changed, 6 insertions(+)
drivers/gpu/drm/qxl/qxl_kms.c +6 -0
index e5ca498..d430f3c 100644
@@ -264,6 +264,11 @@ int qxl_device_init(struct qxl_device *qdev,
 	if (r)
 		return r;
 
+	r = pci_request_region(pdev, 0, "qxl-vram");
+	r = pci_request_region(pdev, 2, "qxl-rom");
+	r = pci_request_region(pdev, 3, "qxl-io");
+	r = pci_request_region(pdev, sb, "qxl-surface");
+
 	return 0;
 }
 
@@ -289,6 +294,7 @@ static void qxl_device_fini(struct qxl_device *qdev)
 	qdev->mode_info.modes = NULL;
 	qdev->mode_info.num_modes = 0;
 	qxl_debugfs_remove_files(qdev);
+	pci_release_regions(qdev->pdev);
 }
 
 int qxl_driver_unload(struct drm_device *dev)
Signature · 1 line
1.8.3.1
author maintainer
#6 lore raw

PATCH 6/6 drm/bochs: new driver

In reply to Gerd Hoffmann

DRM driver for (virtual) vga cards using the bochs dispi interface, such as the qemu standard vga (qemu -vga std).

Don't bother supporting anything but 32bpp for now. Maybe add 16bpp later on.

Known issue: mmap(/dev/fb0) doesn't work.

---
 drivers/gpu/drm/Kconfig             |   2 +
 drivers/gpu/drm/Makefile            |   1 +
 drivers/gpu/drm/bochs/Kconfig       |  11 +
 drivers/gpu/drm/bochs/Makefile      |   4 +
 drivers/gpu/drm/bochs/bochs.h       | 167 +++++++++++
 drivers/gpu/drm/bochs/bochs_drv.c   | 179 ++++++++++++
 drivers/gpu/drm/bochs/bochs_fbdev.c | 215 ++++++++++++++
 drivers/gpu/drm/bochs/bochs_hw.c    | 162 +++++++++++
 drivers/gpu/drm/bochs/bochs_kms.c   | 294 +++++++++++++++++++
 drivers/gpu/drm/bochs/bochs_mm.c    | 552 ++++++++++++++++++++++++++++++++++++
 10 files changed, 1587 insertions(+)
 create mode 100644 drivers/gpu/drm/bochs/Kconfig
 create mode 100644 drivers/gpu/drm/bochs/Makefile
 create mode 100644 drivers/gpu/drm/bochs/bochs.h
 create mode 100644 drivers/gpu/drm/bochs/bochs_drv.c
 create mode 100644 drivers/gpu/drm/bochs/bochs_fbdev.c
 create mode 100644 drivers/gpu/drm/bochs/bochs_hw.c
 create mode 100644 drivers/gpu/drm/bochs/bochs_kms.c
 create mode 100644 drivers/gpu/drm/bochs/bochs_mm.c
drivers/gpu/drm/Kconfig +2 -0
index 955555d..b9d1795 100644
@@ -235,4 +235,6 @@ source "drivers/gpu/drm/tilcdc/Kconfig"
 
 source "drivers/gpu/drm/qxl/Kconfig"
 
+source "drivers/gpu/drm/bochs/Kconfig"
+
 source "drivers/gpu/drm/msm/Kconfig"
drivers/gpu/drm/Makefile +1 -0
index f089adf..e4bb63a 100644
@@ -54,5 +54,6 @@ obj-$(CONFIG_DRM_SHMOBILE) +=shmobile/
 obj-$(CONFIG_DRM_OMAP)	+= omapdrm/
 obj-$(CONFIG_DRM_TILCDC)	+= tilcdc/
 obj-$(CONFIG_DRM_QXL) += qxl/
+obj-$(CONFIG_DRM_BOCHS) += bochs/
 obj-$(CONFIG_DRM_MSM) += msm/
 obj-y			+= i2c/
drivers/gpu/drm/bochs/Kconfig new file +11 -0
new file mode 100644
index 0000000..c8fcf12
@@ -0,0 +1,11 @@
+config DRM_BOCHS
+	tristate "DRM Support for bochs dispi vga interface (qemu stdvga)"
+	depends on DRM && PCI
+	select DRM_KMS_HELPER
+	select FB_SYS_FILLRECT
+	select FB_SYS_COPYAREA
+	select FB_SYS_IMAGEBLIT
+	select DRM_TTM
+	help
+	  Choose this option for qemu.
+	  If M is selected the module will be called bochs-drm.
drivers/gpu/drm/bochs/Makefile new file +4 -0
new file mode 100644
index 0000000..844a556
@@ -0,0 +1,4 @@
+ccflags-y := -Iinclude/drm
+bochs-drm-y := bochs_drv.o bochs_mm.o bochs_kms.o bochs_fbdev.o bochs_hw.o
+
+obj-$(CONFIG_DRM_BOCHS)	+= bochs-drm.o
drivers/gpu/drm/bochs/bochs.h new file +167 -0

167 lines not on this page · raw message

drivers/gpu/drm/bochs/bochs_drv.c new file +179 -0

179 lines not on this page · raw message

drivers/gpu/drm/bochs/bochs_fbdev.c new file +215 -0

215 lines not on this page · raw message

drivers/gpu/drm/bochs/bochs_hw.c new file +162 -0

162 lines not on this page · raw message

drivers/gpu/drm/bochs/bochs_kms.c new file +294 -0

294 lines not on this page · raw message

drivers/gpu/drm/bochs/bochs_mm.c new file +552 -0

552 lines not on this page · raw message

Signature · 1 line
1.8.3.1
maintainer Applied
#7 lore raw
In reply to Gerd Hoffmann
Quoted 6 lines

Here comes a collection of drm patches for qemu emulated virtual graphics cards. Small improvements for cirrus and qxl. A new kms driver for the qemu standard vga.

Patches 1-4 can be queued up for merge, unless someone finds bugs / problems in review of course.

Hi Gerd,

I've merged 1-4, thanks for those.

Quoted 3 lines

Patches 5+6 have known issues and are still work-in-progress. I'm posing them for early review nevertheless.

Yeah I'm not really sure I like request_regions for drm drivers, yes its nice to see things, but the fact
that we always get in trouble with efifb and vesafb and that people seem to insist on later coming along
and supplying unhelpful patches to fix the lack of return checking, I generally just steer clear.

I'll try and find some time to look at the bochs driver, though is it really much more than we have with the
simple drm layer that David Herrmann is producing?

Dave.

#8 lore raw
Quoted 14 lines · Dave Airlie
Quoted 6 lines

Here comes a collection of drm patches for qemu emulated virtual graphics cards. Small improvements for cirrus and qxl. A new kms driver for the qemu standard vga.

Patches 1-4 can be queued up for merge, unless someone finds bugs / problems in review of course.

Hi Gerd,

I've merged 1-4, thanks for those.

Quoted 3 lines

Patches 5+6 have known issues and are still work-in-progress. I'm posing them for early review nevertheless.

Quoted 6 lines

Yeah I'm not really sure I like request_regions for drm drivers, yes its nice to see things, but the fact
that we always get in trouble with efifb and vesafb and that people seem to insist on later coming along
and supplying unhelpful patches to fix the lack of return checking, I generally just steer clear.

We've recently started to request_region the intel graphics stolen block. Surprising how much gunk the core pci resource layer loves to put in there, so caught a few real bugs. Also it's a major pain to actually get it to work (some random platform has an off-by-one apparently and fun stuff like that). But I agree that for the mmio bars it's probably not worth the trouble due to vesafb/efifb bonghits.
-Daniel

Signature · 3 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
author maintainer
#9 lore raw
In reply to Dave Airlie

Hi,

Quoted 3 lines

I'll try and find some time to look at the bochs driver, though is it really much more than we have with the
simple drm layer that David Herrmann is producing?

simpledrm is fixed resolution and depends on the firmware setting things up (from what I've seen a few weeks back), correct?

bochs doesn't need firmware help and can switch resolutions at runtime via xrandr just fine. It's more like cirrus-drm, except that there are not the nasty constrains dictated by real cirrus cards, such as the very limited amount of video memory and supporting 24 bpp only.

bochs-drm doesn't need to hop through the loops to shuffle around the framebuffer bo's. They are in vram no matter what, we have 16MB by default and qemu can be configured to make it larger if needed. Want a FullHD display? No problem. bochs-drm operates with 32bpp. The virtual hardware can do 8bpp and 16bpp too, but I didn't go the extra mile to support that as it looks pointless these days.

cheers,
Gerd