Hi,
Following is v2 of the set. It contains some minor clean ups and the additions
from Dave, mostly related with DRM.
I just reorganized a little the ideas which everyone sent and summarized them
here. The basically TODO is:
- DRM hook to disable IRQ when arbiter is on action is done. What else we need
on DRM department (besides the hook usage that Dave is cooking)? That's a
preliminary here:
http://people.freedesktop.org/~airlied/vgaarb/0002-drm-add-support-to-drm-for-VGA-arbitration.patch
- make a PCI_COMMAND lock to deal with hotplug cards on drivers/pci. Also is
it need some spinlock to deal with bridges? (Waiting Jesse's opinion).
- know exactly if VGA on/off is enough and remove the just I/O or just memory
granularity. Alan states that it would reduce some deadlock.
- Change this hunk bellow:
+ init_waitqueue_entry(&wait, current);
+ add_wait_queue(&vga_wait_queue, &wait);
+ set_current_state(interruptible ?
+ TASK_INTERRUPTIBLE :
+ TASK_UNINTERRUPTIBLE);
+ if (signal_pending(current)) {
+ rc = -EINTR;
+ break;
+ }
+ schedule();
+ remove_wait_queue(&vga_wait_queue, &wait);
+ set_current_state(TASK_RUNNING);
by something like:
+ wait_event_interruptible(...)
it's Alan's comment. Does it seems doable to everyone?
For testing purposes, I'm using posttool:
http://cgit.freedesktop.org/~airlied/posttool/
which has the following dependencies:
http://cgit.freedesktop.org/~vignatti/libx86/
http://cgit.freedesktop.org/~vignatti/libpciaccess/ (vgaarb branch)
Please, comment on and send the feedbacks. Thanks,
Tiago
Dave Airlie (1):
drm: add support to drm for VGA arbitration.
Tiago Vignatti (2):
vga: implements VGA arbitration on Linux
vga: drops a documentation regarding the VGA arbiter
Documentation/vgaarbiter.txt | 197 ++++++
drivers/gpu/Makefile | 2 +-
drivers/gpu/drm/drm_irq.c | 24 +
drivers/gpu/drm/i915/i915_dma.c | 6 +
drivers/gpu/drm/i915/i915_reg.h | 1 +
drivers/gpu/drm/i915/intel_display.c | 17 +
drivers/gpu/drm/i915/intel_drv.h | 1 +
drivers/gpu/drm/radeon/r100.c | 9 +
drivers/gpu/drm/radeon/radeon.h | 2 +
drivers/gpu/drm/radeon/radeon_asic.h | 9 +
drivers/gpu/drm/radeon/radeon_device.c | 5 +-
drivers/gpu/vga/Kconfig | 10 +
drivers/gpu/vga/Makefile | 1 +
drivers/gpu/vga/vgaarb.c | 1116 ++++++++++++++++++++++++++++++++
drivers/video/Kconfig | 2 +
include/drm/drmP.h | 3 +
include/linux/vgaarb.h | 170 +++++
17 files changed, 1573 insertions(+), 2 deletions(-)
create mode 100644 Documentation/vgaarbiter.txt
create mode 100644 drivers/gpu/vga/Kconfig
create mode 100644 drivers/gpu/vga/Makefile
create mode 100644 drivers/gpu/vga/vgaarb.c
create mode 100644 include/linux/vgaarb.h
Signed-off-by: Tiago Vignatti <[email protected]>
---
drivers/gpu/Makefile | 2 +-
drivers/gpu/vga/Kconfig | 10 +
drivers/gpu/vga/Makefile | 1 +
drivers/gpu/vga/vgaarb.c | 1116 ++++++++++++++++++++++++++++++++++++++++++++++
drivers/video/Kconfig | 2 +
include/linux/vgaarb.h | 170 +++++++
6 files changed, 1300 insertions(+), 1 deletions(-)
create mode 100644 drivers/gpu/vga/Kconfig
create mode 100644 drivers/gpu/vga/Makefile
create mode 100644 drivers/gpu/vga/vgaarb.c
create mode 100644 include/linux/vgaarb.h
diff --git a/drivers/gpu/Makefile b/drivers/gpu/Makefile
index de566cf..30879df 100644
--- a/drivers/gpu/Makefile
+++ b/drivers/gpu/Makefile
@@ -1 +1 @@
-obj-y += drm/
+obj-y += drm/ vga/
diff --git a/drivers/gpu/vga/Kconfig b/drivers/gpu/vga/Kconfig
new file mode 100644
index 0000000..790e675
--- /dev/null
+++ b/drivers/gpu/vga/Kconfig
@@ -0,0 +1,10 @@
+config VGA_ARB
+ bool "VGA Arbitration" if EMBEDDED
+ default y
+ depends on PCI
+ help
+ Some "legacy" VGA devices implemented on PCI typically have the same
+ hard-decoded addresses as they did on ISA. When multiple PCI devices
+ are accessed at same time they need some kind of coordination. Please
+ see Documentation/vgaarbiter.txt for more details. Select this to
+ enable VGA arbiter.
diff --git a/drivers/gpu/vga/Makefile b/drivers/gpu/vga/Makefile
new file mode 100644
index 0000000..7cc8c1e
--- /dev/null
+++ b/drivers/gpu/vga/Makefile
@@ -0,0 +1 @@
+obj-$(CONFIG_VGA_ARB) += vgaarb.o
diff --git a/drivers/gpu/vga/vgaarb.c b/drivers/gpu/vga/vgaarb.c
new file mode 100644
index 0000000..79d1c36
--- /dev/null
+++ b/drivers/gpu/vga/vgaarb.c
@@ -0,0 +1,1116 @@
+/*
+ * vgaarb.c
+ *
+ * (C) Copyright 2005 Benjamin Herrenschmidt <[email protected]>
+ * (C) Copyright 2007 Paulo R. Zanoni <[email protected]>
+ * (C) Copyright 2007, 2009 Tiago Vignatti <[email protected]>
+ *
+ * Implements the VGA arbitration. For details refer to
+ * Documentation/vgaarbiter.txt
+ */
+
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/pci.h>
+#include <linux/errno.h>
+#include <linux/init.h>
+#include <linux/list.h>
+#include <linux/sched.h>
+#include <linux/wait.h>
+#include <linux/spinlock.h>
+#include <linux/poll.h>
+#include <linux/miscdevice.h>
+
+#include <linux/uaccess.h>
+
+#include <linux/vgaarb.h>
+
+/*
+ * We keep a list of all vga devices in the system to speed
+ * up the various operations of the arbiter
+ */
+struct vga_device {
+ struct list_head list;
+ struct pci_dev *pdev;
+ unsigned int decodes; /* what does it decodes */
+ unsigned int owns; /* what does it owns */
+ unsigned int locks; /* what does it locks */
+ unsigned int io_lock_cnt; /* legacy IO lock count */
+ unsigned int mem_lock_cnt; /* legacy MEM lock count */
+ unsigned int io_norm_cnt; /* normal IO count */
+ unsigned int mem_norm_cnt; /* normal MEM count */
+
+ /* allow IRQ enable/disable hook */
+ void *irq_cookie;
+ void (*irq_set_state)(void *cookie, bool enable);
+};
+
+static LIST_HEAD(vga_list);
+static DEFINE_SPINLOCK(vga_lock);
+static DECLARE_WAIT_QUEUE_HEAD(vga_wait_queue);
+
+#ifndef __ARCH_HAS_VGA_DEFAULT_DEVICE
+/* this is only used a cookie - it should not be dereferenced */
+static struct pci_dev *vga_default;
+#endif
+
+static void vga_arb_device_card_gone(struct pci_dev *pdev);
+
+/* Find somebody in our list */
+static struct vga_device *vgadev_find(struct pci_dev *pdev)
+{
+ struct vga_device *vgadev;
+
+ list_for_each_entry(vgadev, &vga_list, list)
+ if (pdev == vgadev->pdev)
+ return vgadev;
+ return NULL;
+}
+
+/* Returns the default VGA device (vgacon's babe) */
+#ifndef __ARCH_HAS_VGA_DEFAULT_DEVICE
+struct pci_dev *vga_default_device(void)
+{
+ return vga_default;
+}
+#endif
+
+static inline void vga_irq_set_state(struct vga_device *vgadev, bool state)
+{
+ if (vgadev->irq_set_state)
+ vgadev->irq_set_state(vgadev->irq_cookie, state);
+}
+/* Architecture can override enabling/disabling of a given
+ * device resources here
+ *
+ * TODO: <benh> you can try to be "smart" and if for example you have 2 cards
+ * that are behind separate bridges, you need only swapping the bridge controls
+ * and you can keep the card master IO and memory enable.
+ */
+#ifndef __ARCH_HAS_VGA_DISABLE_RESOURCES
+static inline void vga_disable_resources(struct pci_dev *pdev,
+ unsigned int rsrc,
+ unsigned int change_bridge)
+{
+ struct pci_bus *bus;
+ struct pci_dev *bridge;
+ u16 cmd;
+
+ pr_devel("%s\n", __func__);
+
+ pci_read_config_word(pdev, PCI_COMMAND, &cmd);
+ if (rsrc & (VGA_RSRC_LEGACY_IO | VGA_RSRC_NORMAL_IO))
+ cmd &= ~PCI_COMMAND_IO;
+ if (rsrc & (VGA_RSRC_LEGACY_MEM | VGA_RSRC_NORMAL_MEM))
+ cmd &= ~PCI_COMMAND_MEMORY;
+ pci_write_config_word(pdev, PCI_COMMAND, cmd);
+
+ if (!change_bridge)
+ return;
+
+ bus = pdev->bus;
+ while (bus) {
+ bridge = bus->self;
+ if (bridge) {
+ pci_read_config_word(bridge, PCI_BRIDGE_CONTROL,
+ &cmd);
+ if (cmd & PCI_BRIDGE_CTL_VGA) {
+ cmd &= ~PCI_BRIDGE_CTL_VGA;
+ pci_write_config_word(bridge,
+ PCI_BRIDGE_CONTROL,
+ cmd);
+ }
+ }
+ bus = bus->parent;
+ }
+}
+#endif
+
+#ifndef __ARCH_HAS_VGA_ENABLE_RESOURCES
+static inline void vga_enable_resources(struct pci_dev *pdev,
+ unsigned int rsrc)
+{
+ struct pci_bus *bus;
+ struct pci_dev *bridge;
+ u16 cmd;
+
+ pr_devel("%s\n", __func__);
+
+ pci_read_config_word(pdev, PCI_COMMAND, &cmd);
+ if (rsrc & (VGA_RSRC_LEGACY_IO | VGA_RSRC_NORMAL_IO))
+ cmd |= PCI_COMMAND_IO;
+ if (rsrc & (VGA_RSRC_LEGACY_MEM | VGA_RSRC_NORMAL_MEM))
+ cmd |= PCI_COMMAND_MEMORY;
+ pci_write_config_word(pdev, PCI_COMMAND, cmd);
+
+ if (!(rsrc & VGA_RSRC_LEGACY_MASK))
+ return;
+
+ bus = pdev->bus;
+ while (bus) {
+ bridge = bus->self;
+ if (bridge) {
+ pci_read_config_word(bridge, PCI_BRIDGE_CONTROL,
+ &cmd);
+ if (!(cmd & PCI_BRIDGE_CTL_VGA)) {
+ cmd |= PCI_BRIDGE_CTL_VGA;
+ pci_write_config_word(bridge,
+ PCI_BRIDGE_CONTROL,
+ cmd);
+ }
+ }
+ bus = bus->parent;
+ }
+}
+#endif
+
+static struct vga_device *__vga_tryget(struct vga_device *vgadev,
+ unsigned int rsrc)
+{
+ unsigned int wants, legacy_wants, match;
+ struct vga_device *conflict;
+
+ /* Account for "normal" resources to lock. If we decode the legacy,
+ * counterpart, we need to request it as well
+ */
+ if ((rsrc & VGA_RSRC_NORMAL_IO) &&
+ (vgadev->decodes & VGA_RSRC_LEGACY_IO))
+ rsrc |= VGA_RSRC_LEGACY_IO;
+ if ((rsrc & VGA_RSRC_NORMAL_MEM) &&
+ (vgadev->decodes & VGA_RSRC_LEGACY_MEM))
+ rsrc |= VGA_RSRC_LEGACY_MEM;
+
+ pr_devel("%s: %d\n", __func__, rsrc);
+ pr_devel("%s: owns: %d\n", __func__, vgadev->owns);
+
+ /* Check what resources we need to acquire */
+ wants = rsrc & ~vgadev->owns;
+
+ /* We already own everything, just mark locked & bye bye */
+ if (wants == 0)
+ goto lock_them;
+
+ /* We don't need to request a legacy resource, we just enable
+ * appropriate decoding and go
+ */
+ legacy_wants = wants & VGA_RSRC_LEGACY_MASK;
+ if (legacy_wants == 0)
+ goto enable_them;
+
+ /* Ok, we don't, let's find out how we need to kick off */
+ list_for_each_entry(conflict, &vga_list, list) {
+ unsigned int lwants = legacy_wants;
+ unsigned int change_bridge = 0;
+
+ /* Don't conflict with myself */
+ if (vgadev == conflict)
+ continue;
+
+ /* Check if the architecture allows a conflict between those
+ * 2 devices or if they are on separate domains
+ */
+ if (!vga_conflicts(vgadev->pdev, conflict->pdev))
+ continue;
+
+ /* We have a possible conflict. before we go further, we must
+ * check if we sit on the same bus as the conflicting device.
+ * if we don't, then we must tie both IO and MEM resources
+ * together since there is only a single bit controlling
+ * VGA forwarding on P2P bridges
+ */
+ if (vgadev->pdev->bus != conflict->pdev->bus) {
+ change_bridge = 1;
+ lwants = VGA_RSRC_LEGACY_IO | VGA_RSRC_LEGACY_MEM;
+ }
+
+ /* Check if the guy has a lock on the resource. If he does,
+ * return the conflicting entry
+ */
+ if (conflict->locks & lwants)
+ return conflict;
+
+ /* Ok, now check if he owns the resource we want. We don't need
+ * to check "decodes" since it should be impossible to own
+ * own legacy resources you don't decode unless I have a bug
+ * in this code...
+ */
+ WARN_ON(conflict->owns & ~conflict->decodes);
+ match = lwants & conflict->owns;
+ if (!match)
+ continue;
+
+ /* looks like he doesn't have a lock, we can steal
+ * them from him
+ */
+ vga_irq_set_state(conflict, false);
+ vga_disable_resources(conflict->pdev, lwants,
+ change_bridge);
+ conflict->owns &= ~lwants;
+ /* If he also owned non-legacy, that is no longer the case */
+ if (lwants & VGA_RSRC_LEGACY_MEM)
+ conflict->owns &= ~VGA_RSRC_NORMAL_MEM;
+ if (lwants & VGA_RSRC_LEGACY_IO)
+ conflict->owns &= ~VGA_RSRC_NORMAL_IO;
+ }
+
+enable_them:
+ /* ok dude, we got it, everybody conflicting has been disabled, let's
+ * enable us. Make sure we don't mark a bit in "owns" that we don't
+ * also have in "decodes". We can lock resources we don't decode but
+ * not own them.
+ */
+ vga_enable_resources(vgadev->pdev, wants);
+ vga_irq_set_state(vgadev, true);
+ vgadev->owns |= (wants & vgadev->decodes);
+lock_them:
+ vgadev->locks |= (rsrc & VGA_RSRC_LEGACY_MASK);
+ if (rsrc & VGA_RSRC_LEGACY_IO)
+ vgadev->io_lock_cnt++;
+ if (rsrc & VGA_RSRC_LEGACY_MEM)
+ vgadev->mem_lock_cnt++;
+ if (rsrc & VGA_RSRC_NORMAL_IO)
+ vgadev->io_norm_cnt++;
+ if (rsrc & VGA_RSRC_NORMAL_MEM)
+ vgadev->mem_norm_cnt++;
+
+ return NULL;
+}
+
+static void __vga_put(struct vga_device *vgadev, unsigned int rsrc)
+{
+ unsigned int old_locks = vgadev->locks;
+
+ pr_devel("%s\n", __func__);
+
+ /* Update our counters, and account for equivalent legacy resources
+ * if we decode them
+ */
+ if ((rsrc & VGA_RSRC_NORMAL_IO) && vgadev->io_norm_cnt > 0) {
+ vgadev->io_norm_cnt--;
+ if (vgadev->decodes & VGA_RSRC_LEGACY_IO)
+ rsrc |= VGA_RSRC_LEGACY_IO;
+ }
+ if ((rsrc & VGA_RSRC_NORMAL_MEM) && vgadev->mem_norm_cnt > 0) {
+ vgadev->mem_norm_cnt--;
+ if (vgadev->decodes & VGA_RSRC_LEGACY_MEM)
+ rsrc |= VGA_RSRC_LEGACY_MEM;
+ }
+ if ((rsrc & VGA_RSRC_LEGACY_IO) && vgadev->io_lock_cnt > 0)
+ vgadev->io_lock_cnt--;
+ if ((rsrc & VGA_RSRC_LEGACY_MEM) && vgadev->mem_lock_cnt > 0)
+ vgadev->mem_lock_cnt--;
+
+ /* Just clear lock bits, we do lazy operations so we don't really
+ * have to bother about anything else at this point
+ */
+ if (vgadev->io_lock_cnt == 0)
+ vgadev->locks &= ~VGA_RSRC_LEGACY_IO;
+ if (vgadev->mem_lock_cnt == 0)
+ vgadev->locks &= ~VGA_RSRC_LEGACY_MEM;
+
+ /* Kick the wait queue in case somebody was waiting if we actually
+ * released something
+ */
+ if (old_locks != vgadev->locks)
+ wake_up_all(&vga_wait_queue);
+}
+
+int vga_get(struct pci_dev *pdev, unsigned int rsrc, int interruptible)
+{
+ struct vga_device *vgadev, *conflict;
+ unsigned long flags;
+ wait_queue_t wait;
+ int rc = 0;
+
+ /* The one who calls us should check for this, but lets be sure... */
+ if (pdev == NULL)
+ pdev = vga_default_device();
+ if (pdev == NULL)
+ return 0;
+
+ for (;;) {
+ spin_lock_irqsave(&vga_lock, flags);
+ vgadev = vgadev_find(pdev);
+ if (vgadev == NULL) {
+ spin_unlock_irqrestore(&vga_lock, flags);
+ rc = -ENODEV;
+ break;
+ }
+ conflict = __vga_tryget(vgadev, rsrc);
+ spin_unlock_irqrestore(&vga_lock, flags);
+ if (conflict == NULL)
+ break;
+
+
+ /* We have a conflict, we wait until somebody kicks the
+ * work queue. Currently we have one work queue that we
+ * kick each time some resources are released, but it would
+ * be fairly easy to have a per device one so that we only
+ * need to attach to the conflicting device
+ */
+ init_waitqueue_entry(&wait, current);
+ add_wait_queue(&vga_wait_queue, &wait);
+ set_current_state(interruptible ?
+ TASK_INTERRUPTIBLE :
+ TASK_UNINTERRUPTIBLE);
+ if (signal_pending(current)) {
+ rc = -EINTR;
+ break;
+ }
+ schedule();
+ remove_wait_queue(&vga_wait_queue, &wait);
+ set_current_state(TASK_RUNNING);
+ }
+ return rc;
+}
+
+int vga_tryget(struct pci_dev *pdev, unsigned int rsrc)
+{
+ struct vga_device *vgadev;
+ unsigned long flags;
+ int rc = 0;
+
+ /* The one who calls us should check for this, but lets be sure... */
+ if (pdev == NULL)
+ pdev = vga_default_device();
+ if (pdev == NULL)
+ return 0;
+ spin_lock_irqsave(&vga_lock, flags);
+ vgadev = vgadev_find(pdev);
+ if (vgadev == NULL) {
+ rc = -ENODEV;
+ goto bail;
+ }
+ if (__vga_tryget(vgadev, rsrc))
+ rc = -EBUSY;
+bail:
+ spin_unlock_irqrestore(&vga_lock, flags);
+ return rc;
+}
+
+void vga_put(struct pci_dev *pdev, unsigned int rsrc)
+{
+ struct vga_device *vgadev;
+ unsigned long flags;
+
+ /* The one who calls us should check for this, but lets be sure... */
+ if (pdev == NULL)
+ pdev = vga_default_device();
+ if (pdev == NULL)
+ return;
+ spin_lock_irqsave(&vga_lock, flags);
+ vgadev = vgadev_find(pdev);
+ if (vgadev == NULL)
+ goto bail;
+ __vga_put(vgadev, rsrc);
+bail:
+ spin_unlock_irqrestore(&vga_lock, flags);
+}
+
+/*
+ * Currently, we assume that the "initial" setup of the system is
+ * not sane, that is we come up with conflicting devices and let
+ * the arbiter's client decides if devices decodes or not legacy
+ * things.
+ */
+static void vga_arbiter_add_pci_device(struct pci_dev *pdev)
+{
+ struct vga_device *vgadev;
+ unsigned long flags;
+ struct pci_bus *bus;
+ struct pci_dev *bridge;
+ u16 cmd;
+
+ /* Only deal with VGA class devices */
+ if ((pdev->class >> 8) != PCI_CLASS_DISPLAY_VGA)
+ return;
+
+ /* Allocate structure */
+ vgadev = kmalloc(sizeof(struct vga_device), GFP_KERNEL);
+ if (vgadev == NULL) {
+ pr_err("vgaarb: failed to allocate pci device\n");
+ /* What to do on allocation failure ? For now, let's
+ * just do nothing, I'm not sure there is anything saner
+ * to be done
+ */
+ return;
+ }
+
+ memset(vgadev, 0, sizeof(*vgadev));
+
+ /* Take lock & check for duplicates */
+ spin_lock_irqsave(&vga_lock, flags);
+ if (vgadev_find(pdev) != NULL) {
+ BUG_ON(1);
+ goto fail;
+ }
+ vgadev->pdev = pdev;
+
+ /* By default, assume we decode everything */
+ vgadev->decodes = VGA_RSRC_LEGACY_IO | VGA_RSRC_LEGACY_MEM |
+ VGA_RSRC_NORMAL_IO | VGA_RSRC_NORMAL_MEM;
+
+ /* Mark that we "own" resources based on our enables, we will
+ * clear that below if the bridge isn't forwarding
+ */
+ pci_read_config_word(pdev, PCI_COMMAND, &cmd);
+ if (cmd & PCI_COMMAND_IO)
+ vgadev->owns |= VGA_RSRC_LEGACY_IO;
+ if (cmd & PCI_COMMAND_MEMORY)
+ vgadev->owns |= VGA_RSRC_LEGACY_MEM;
+
+ /* Check if VGA cycles can get down to us */
+ bus = pdev->bus;
+ while (bus) {
+ bridge = bus->self;
+ if (bridge) {
+ u16 l;
+ pci_read_config_word(bridge, PCI_BRIDGE_CONTROL,
+ &l);
+ if (!(l & PCI_BRIDGE_CTL_VGA)) {
+ vgadev->owns = 0;
+ break;
+ }
+ }
+ bus = bus->parent;
+ }
+
+ /* Deal with VGA default device. Use first enabled one
+ * by default if arch doesn't have it's own hook
+ */
+#ifndef __ARCH_HAS_VGA_DEFAULT_DEVICE
+ if (vga_default == NULL && (vgadev->owns & VGA_RSRC_LEGACY_MEM) &&
+ (vgadev->owns & VGA_RSRC_LEGACY_IO))
+ vga_default = pdev;
+
+#endif
+
+ /* Add to the list */
+ list_add(&vgadev->list, &vga_list);
+ pr_info("vgaarb: device added: %d %d %d\n",
+ pdev->device, pdev->vendor, pdev->devfn);
+ spin_unlock_irqrestore(&vga_lock, flags);
+ return;
+fail:
+ spin_unlock_irqrestore(&vga_lock, flags);
+ kfree(vgadev);
+}
+
+static void vga_arbiter_del_pci_device(struct pci_dev *pdev)
+{
+ struct vga_device *vgadev;
+ unsigned long flags;
+
+ spin_lock_irqsave(&vga_lock, flags);
+ vgadev = vgadev_find(pdev);
+ if (vgadev == NULL)
+ goto bail;
+
+ if (vga_default == pdev)
+ vga_default = NULL;
+
+ /* Remove entry from list */
+ list_del(&vgadev->list);
+
+ /* Notify userland driver that the device is gone so it discards
+ * it's copies of the pci_dev pointer
+ */
+ vga_arb_device_card_gone(pdev);
+
+ /* Wake up all possible waiters */
+ wake_up_all(&vga_wait_queue);
+bail:
+ spin_unlock_irqrestore(&vga_lock, flags);
+ kfree(vgadev);
+}
+
+void vga_set_legacy_decoding(struct pci_dev *pdev, unsigned int decodes)
+{
+ struct vga_device *vgadev;
+ unsigned long flags;
+
+ decodes &= VGA_RSRC_LEGACY_MASK;
+
+ spin_lock_irqsave(&vga_lock, flags);
+ vgadev = vgadev_find(pdev);
+ if (vgadev == NULL)
+ goto bail;
+
+ vgadev->decodes = decodes;
+ vgadev->owns &= decodes;
+
+ /* XXX if somebody is going from "doesn't decode" to "decodes" state
+ * here, additional care must be taken as we may have pending owner
+ * ship of non-legacy region ...
+ */
+bail:
+ spin_unlock_irqrestore(&vga_lock, flags);
+}
+EXPORT_SYMBOL(vga_set_legacy_decoding);
+
+void vga_set_irq_callback(struct pci_dev *pdev, void *cookie, void (*irq_func)(void *cookie, bool enable))
+{
+ struct vga_device *vgadev;
+ unsigned long flags;
+
+ spin_lock_irqsave(&vga_lock, flags);
+ vgadev = vgadev_find(pdev);
+ if (!vgadev)
+ goto bail;
+
+ vgadev->irq_cookie = cookie;
+ vgadev->irq_set_state = irq_func;
+bail:
+ spin_unlock_irqrestore(&vga_lock, flags);
+
+}
+EXPORT_SYMBOL(vga_set_irq_callback);
+
+/*
+ * Char driver implementation
+ *
+ * Semantics is:
+ *
+ * open : open user instance of the arbitrer. by default, it's
+ * attached to the default VGA device of the system.
+ *
+ * close : close user instance, release locks
+ *
+ * read : return a string indicating the status of the target.
+ * an IO state string is of the form {io,mem,io+mem,none},
+ * mc and ic are respectively mem and io lock counts (for
+ * debugging/diagnostic only). "decodes" indicate what the
+ * card currently decodes, "owns" indicates what is currently
+ * enabled on it, and "locks" indicates what is locked by this
+ * card. If the card is unplugged, we get "invalid" then for
+ * card_ID and an -ENODEV error is returned for any command
+ * until a new card is targeted
+ *
+ * "<card_ID>,decodes=<io_state>,owns=<io_state>,locks=<io_state> (ic,mc)"
+ *
+ * write : write a command to the arbiter. List of commands is:
+ *
+ * target <card_ID> : switch target to card <card_ID> (see below)
+ * lock <io_state> : acquires locks on target ("none" is invalid io_state)
+ * trylock <io_state> : non-blocking acquire locks on target
+ * unlock <io_state> : release locks on target
+ * unlock all : release all locks on target held by this user
+ * decodes <io_state> : set the legacy decoding attributes for the card
+ *
+ * poll : event if something change on any card (not just the target)
+ *
+ * card_ID is of the form "PCI:domain:bus:dev.fn". It can be set to "default"
+ * to go back to the system default card (TODO: not implemented yet).
+ * Currently, only PCI is supported as a prefix, but the userland API may
+ * support other bus types in the future, even if the current kernel
+ * implementation doesn't.
+ *
+ * Note about locks:
+ *
+ * The driver keeps track of which user has what locks on which card. It
+ * supports stacking, like the kernel one. This complexifies the implementation
+ * a bit, but makes the arbiter more tolerant to userspace problems and able
+ * to properly cleanup in all cases when a process dies.
+ * Currently, a max of 16 cards simultaneously can have locks issued from
+ * userspace for a given user (file descriptor instance) of the arbiter.
+ *
+ * If the device is hot-unplugged, there is a hook inside the module to notify
+ * they being added/removed in the system and automatically added/removed in
+ * the arbiter.
+ */
+
+#define MAX_USER_CARDS 16
+#define PCI_INVALID_CARD ((struct pci_dev *)-1UL)
+
+/*
+ * Each user has an array of these, tracking which cards have locks
+ */
+struct vga_arb_user_card {
+ struct pci_dev *pdev;
+ unsigned int mem_cnt;
+ unsigned int io_cnt;
+};
+
+struct vga_arb_private {
+ struct list_head list;
+ struct pci_dev *target;
+ struct vga_arb_user_card cards[MAX_USER_CARDS];
+ spinlock_t lock;
+};
+
+static LIST_HEAD(vga_user_list);
+static DEFINE_SPINLOCK(vga_user_lock);
+
+static const char *vga_iostate_to_str(unsigned int iostate)
+{
+ /* Ignore VGA_RSRC_IO and VGA_RSRC_MEM */
+ iostate &= VGA_RSRC_LEGACY_IO | VGA_RSRC_LEGACY_MEM;
+ switch (iostate) {
+ case VGA_RSRC_LEGACY_IO | VGA_RSRC_LEGACY_MEM:
+ return "io+mem";
+ case VGA_RSRC_LEGACY_IO:
+ return "io";
+ case VGA_RSRC_LEGACY_MEM:
+ return "mem";
+ }
+ return "none";
+}
+
+static int vga_str_to_iostate(char *buf, int str_size, int *io_state)
+{
+ /* XXX We're not chekcing the str_size! */
+ if (strncmp(buf, "io+mem", 6) == 0)
+ *io_state = VGA_RSRC_LEGACY_IO | VGA_RSRC_LEGACY_MEM;
+ else if (strncmp(buf, "io", 2) == 0)
+ *io_state = VGA_RSRC_LEGACY_IO;
+ else if (strncmp(buf, "mem", 3) == 0)
+ *io_state = VGA_RSRC_LEGACY_MEM;
+ else if (strncmp(buf, "none", 4) == 0)
+ *io_state = VGA_RSRC_NONE;
+ else
+ return 0;
+ return 1;
+}
+
+/*
+ * This function gets a string in the format: "PCI:domain:bus:dev.fn" and
+ * returns the respective values. If the string is not in this format,
+ * it returns 0.
+ */
+static int vga_pci_str_to_vars(char *buf, int count, unsigned int *domain,
+ unsigned int *bus, unsigned int *devfn)
+{
+ int n;
+ unsigned int slot, func;
+
+ n = sscanf(buf, "PCI:%d:%d:%d.%d", domain, bus, &slot, &func);
+ if (n != 4)
+ return 0;
+
+ *devfn = PCI_DEVFN(slot, func);
+
+ return 1;
+}
+
+static ssize_t vga_arb_read(struct file *file, char __user * buf,
+ size_t count, loff_t *ppos)
+{
+ struct vga_arb_private *priv = file->private_data;
+ struct vga_device *vgadev;
+ struct pci_dev *pdev;
+ unsigned long flags;
+ size_t len;
+ int rc;
+ char *lbuf;
+
+ lbuf = kmalloc(1024, GFP_KERNEL);
+ if (lbuf == NULL)
+ return -ENOMEM;
+
+ /* Shields against vga_arb_device_card_gone (pci_dev going
+ * away), and allows access to vga list
+ */
+ spin_lock_irqsave(&vga_lock, flags);
+
+ /* If we are targetting the default, use it */
+ pdev = priv->target;
+ if (pdev == NULL || pdev == PCI_INVALID_CARD) {
+ spin_unlock_irqrestore(&vga_lock, flags);
+ len = sprintf(lbuf, "invalid");
+ goto done;
+ }
+
+ /* Find card vgadev structure */
+ vgadev = vgadev_find(pdev);
+ if (vgadev == NULL) {
+ /* Wow, it's not in the list, that shouldn't happen,
+ * let's fix us up and return invalid card
+ */
+ if (pdev == priv->target)
+ vga_arb_device_card_gone(pdev);
+ spin_unlock_irqrestore(&vga_lock, flags);
+ len = sprintf(lbuf, "invalid");
+ goto done;
+ }
+
+ /* Fill the buffer with infos */
+ len = snprintf(lbuf, 1024,
+ "PCI:%s,decodes=%s,owns=%s,locks=%s (%d,%d)\n",
+ pci_name(pdev),
+ vga_iostate_to_str(vgadev->decodes),
+ vga_iostate_to_str(vgadev->owns),
+ vga_iostate_to_str(vgadev->locks),
+ vgadev->io_lock_cnt, vgadev->mem_lock_cnt);
+
+ spin_unlock_irqrestore(&vga_lock, flags);
+done:
+
+ /* Copy that to user */
+ if (len > count)
+ len = count;
+ rc = copy_to_user(buf, lbuf, len);
+ kfree(lbuf);
+ if (rc)
+ return -EFAULT;
+ return len;
+}
+
+/*
+ * TODO: To avoid parsing inside kernel and to improve the speed we may
+ * consider use ioctl here
+ */
+static ssize_t vga_arb_write(struct file *file, const char __user * buf,
+ size_t count, loff_t *ppos)
+{
+ struct vga_arb_private *priv = file->private_data;
+ struct pci_dev *pdev;
+
+ unsigned int io_state;
+
+ char *kbuf, *curr_pos;
+ size_t remaining = count;
+
+ int ret_val;
+ int i;
+
+
+ kbuf = kmalloc(count + 1, GFP_KERNEL);
+ if (!kbuf)
+ return -ENOMEM;
+
+ if (copy_from_user(kbuf, buf, count)) {
+ kfree(kbuf);
+ return -EFAULT;
+ }
+ curr_pos = kbuf;
+ kbuf[count] = '\0'; /* Just to make sure... */
+
+ if (strncmp(curr_pos, "lock ", 5) == 0) {
+ curr_pos += 5;
+ remaining -= 5;
+
+ pr_devel("client 0x%X called 'lock'\n", (int)priv);
+
+ if (!vga_str_to_iostate(curr_pos, remaining, &io_state)) {
+ ret_val = -EPROTO;
+ goto done;
+ }
+ if (io_state == VGA_RSRC_NONE) {
+ ret_val = -EPROTO;
+ goto done;
+ }
+
+ pdev = priv->target;
+ if (priv->target == NULL) {
+ ret_val = -ENODEV;
+ goto done;
+ }
+
+ vga_get_uninterruptible(pdev, io_state);
+
+ /* Update the client's locks lists... */
+ for (i = 0; i < MAX_USER_CARDS; i++) {
+ if (priv->cards[i].pdev == pdev) {
+ if (io_state & VGA_RSRC_LEGACY_IO)
+ priv->cards[i].io_cnt++;
+ if (io_state & VGA_RSRC_LEGACY_MEM)
+ priv->cards[i].mem_cnt++;
+ break;
+ }
+ }
+
+ ret_val = count;
+ goto done;
+ } else if (strncmp(curr_pos, "unlock ", 7) == 0) {
+ curr_pos += 7;
+ remaining -= 7;
+
+ pr_devel("client 0x%X called 'unlock'\n", (int)priv);
+
+ if (strncmp(curr_pos, "all", 3) == 0)
+ io_state = VGA_RSRC_LEGACY_IO | VGA_RSRC_LEGACY_MEM;
+ else {
+ if (!vga_str_to_iostate
+ (curr_pos, remaining, &io_state)) {
+ ret_val = -EPROTO;
+ goto done;
+ }
+ /* TODO: Add this?
+ if (io_state == VGA_RSRC_NONE) {
+ ret_val = -EPROTO;
+ goto done;
+ }
+ */
+ }
+
+ pdev = priv->target;
+ if (priv->target == NULL) {
+ ret_val = -ENODEV;
+ goto done;
+ }
+
+ vga_put(pdev, io_state);
+ for (i = 0; i < MAX_USER_CARDS; i++) {
+ if (priv->cards[i].pdev == pdev) {
+ if (io_state & VGA_RSRC_LEGACY_IO)
+ priv->cards[i].io_cnt--;
+ if (io_state & VGA_RSRC_LEGACY_MEM)
+ priv->cards[i].mem_cnt--;
+ break;
+ }
+ }
+
+ ret_val = count;
+ goto done;
+ } else if (strncmp(curr_pos, "trylock ", 8) == 0) {
+ curr_pos += 8;
+ remaining -= 8;
+
+ pr_devel("client 0x%X called 'trylock'\n", (int)priv);
+
+ if (!vga_str_to_iostate(curr_pos, remaining, &io_state)) {
+ ret_val = -EPROTO;
+ goto done;
+ }
+ /* TODO: Add this?
+ if (io_state == VGA_RSRC_NONE) {
+ ret_val = -EPROTO;
+ goto done;
+ }
+ */
+
+ pdev = priv->target;
+ if (priv->target == NULL) {
+ ret_val = -ENODEV;
+ goto done;
+ }
+
+ if (vga_tryget(pdev, io_state)) {
+ /* Update the client's locks lists... */
+ for (i = 0; i < MAX_USER_CARDS; i++) {
+ if (priv->cards[i].pdev == pdev) {
+ if (io_state & VGA_RSRC_LEGACY_IO)
+ priv->cards[i].io_cnt++;
+ if (io_state & VGA_RSRC_LEGACY_MEM)
+ priv->cards[i].mem_cnt++;
+ break;
+ }
+ }
+ ret_val = count;
+ goto done;
+ } else {
+ ret_val = -EBUSY;
+ goto done;
+ }
+
+ } else if (strncmp(curr_pos, "target ", 7) == 0) {
+ unsigned int domain, bus, devfn;
+ struct vga_device *vgadev;
+
+ curr_pos += 7;
+ remaining -= 7;
+ pr_devel("client 0x%X called 'target'\n", (int)priv);
+ if (!vga_pci_str_to_vars(curr_pos, remaining,
+ &domain, &bus, &devfn)) {
+ ret_val = -EPROTO;
+ goto done;
+ }
+
+ pdev = pci_get_bus_and_slot(bus, devfn);
+ if (!pdev) {
+ pr_info("invalid PCI address\n");
+ ret_val = -ENODEV;
+ goto done;
+ }
+
+ vgadev = vgadev_find(pdev);
+ if (vgadev == NULL) {
+ pr_info("this PCI is not a vga device\n");
+ ret_val = -ENODEV;
+ goto done;
+ }
+
+ priv->target = pdev;
+ for (i = 0; i < MAX_USER_CARDS; i++) {
+ if (priv->cards[i].pdev == pdev)
+ break;
+ if (priv->cards[i].pdev == NULL) {
+ priv->cards[i].pdev = pdev;
+ priv->cards[i].io_cnt = 0;
+ priv->cards[i].mem_cnt = 0;
+ break;
+ }
+ }
+ if (i == MAX_USER_CARDS) {
+ pr_err("maximum user cards number reached!\n");
+ /* XXX: which value to return? */
+ ret_val = -ENOMEM;
+ goto done;
+ }
+
+ ret_val = count;
+ goto done;
+
+
+ } else if (strncmp(curr_pos, "decodes ", 8) == 0) {
+ curr_pos += 8;
+ remaining -= 8;
+ pr_devel("client 0x%X called 'decodes'\n", (int)priv);
+
+ if (!vga_str_to_iostate(curr_pos, remaining, &io_state)) {
+ ret_val = -EPROTO;
+ goto done;
+ }
+ pdev = priv->target;
+ if (priv->target == NULL) {
+ ret_val = -ENODEV;
+ goto done;
+ }
+
+ vga_set_legacy_decoding(pdev, io_state);
+ ret_val = count;
+ goto done;
+ }
+ /* If we got here, the message written is not part of the protocol! */
+ kfree(kbuf);
+ return -EPROTO;
+
+done:
+ kfree(kbuf);
+ return ret_val;
+}
+
+static unsigned int vga_arb_fpoll(struct file *file, poll_table * wait)
+{
+ struct vga_arb_private *priv = file->private_data;
+
+ pr_devel("%s\n", __func__);
+
+ if (priv == NULL)
+ return -ENODEV;
+ poll_wait(file, &vga_wait_queue, wait);
+ return POLLIN;
+}
+
+static int vga_arb_open(struct inode *inode, struct file *file)
+{
+ struct vga_arb_private *priv;
+ unsigned long flags;
+
+ pr_devel("%s\n", __func__);
+
+ priv = kmalloc(sizeof(struct vga_arb_private), GFP_KERNEL);
+ if (priv == NULL)
+ return -ENOMEM;
+ memset(priv, 0, sizeof(*priv));
+ spin_lock_init(&priv->lock);
+ file->private_data = priv;
+
+ spin_lock_irqsave(&vga_user_lock, flags);
+ list_add(&priv->list, &vga_user_list);
+ spin_unlock_irqrestore(&vga_user_lock, flags);
+
+ /* Set the client' lists of locks */
+ priv->target = vga_default_device(); /* Maybe this is still null! */
+ priv->cards[0].pdev = priv->target;
+ priv->cards[0].io_cnt = 0;
+ priv->cards[0].mem_cnt = 0;
+
+
+ return 0;
+}
+
+static int vga_arb_release(struct inode *inode, struct file *file)
+{
+ struct vga_arb_private *priv = file->private_data;
+ struct vga_arb_user_card *uc;
+ unsigned long flags;
+ int i;
+
+ pr_devel("%s\n", __func__);
+
+ if (priv == NULL)
+ return -ENODEV;
+
+ spin_lock_irqsave(&vga_user_lock, flags);
+ list_del(&priv->list);
+ for (i = 0; i < MAX_USER_CARDS; i++) {
+ uc = &priv->cards[i];
+ if (uc->pdev == NULL)
+ continue;
+ pr_devel("uc->io_cnt == %d, uc->mem_cnt == %d\n",
+ uc->io_cnt, uc->mem_cnt);
+ while (uc->io_cnt--)
+ vga_put(uc->pdev, VGA_RSRC_LEGACY_IO);
+ while (uc->mem_cnt--)
+ vga_put(uc->pdev, VGA_RSRC_LEGACY_MEM);
+ }
+ spin_unlock_irqrestore(&vga_user_lock, flags);
+
+ kfree(priv);
+
+ return 0;
+}
+
+static void vga_arb_device_card_gone(struct pci_dev *pdev)
+{
+}
+
+static int pci_notify(struct notifier_block *nb, unsigned long action,
+ void *data)
+{
+ struct device *dev = data;
+ struct pci_dev *pdev = to_pci_dev(dev);
+
+ pr_devel("%s\n", __func__);
+
+ /* For now we're only intereted in devices added and removed. I didn't
+ * test this thing here, so someone needs to double check for the
+ * cases of hotplugable vga cards. */
+ if (action == BUS_NOTIFY_ADD_DEVICE)
+ vga_arbiter_add_pci_device(pdev);
+ else if (action == BUS_NOTIFY_DEL_DEVICE)
+ vga_arbiter_del_pci_device(pdev);
+
+ return 0;
+}
+
+static struct notifier_block pci_notifier = {
+ .notifier_call = pci_notify,
+};
+
+static const struct file_operations vga_arb_device_fops = {
+ .read = vga_arb_read,
+ .write = vga_arb_write,
+ .poll = vga_arb_fpoll,
+ .open = vga_arb_open,
+ .release = vga_arb_release,
+};
+
+static struct miscdevice vga_arb_device = {
+ MISC_DYNAMIC_MINOR, "vga_arbiter", &vga_arb_device_fops
+};
+
+static int __init vga_arb_device_init(void)
+{
+ int rc;
+ struct pci_dev *pdev;
+
+ rc = misc_register(&vga_arb_device);
+ if (rc < 0)
+ pr_err("vgaarb: error %d registering device\n", rc);
+
+ bus_register_notifier(&pci_bus_type, &pci_notifier);
+
+ /* We add all pci devices satisfying vga class in the arbiter by
+ * default */
+ pdev = NULL;
+ while ((pdev =
+ pci_get_subsys(PCI_ANY_ID, PCI_ANY_ID, PCI_ANY_ID,
+ PCI_ANY_ID, pdev)) != NULL)
+ vga_arbiter_add_pci_device(pdev);
+
+ pr_info("vgaarb: loaded\n");
+ return rc;
+}
+subsys_initcall(vga_arb_device_init);
diff --git a/drivers/video/Kconfig b/drivers/video/Kconfig
index 8afcf08..f4ed145 100644
--- a/drivers/video/Kconfig
+++ b/drivers/video/Kconfig
@@ -7,6 +7,8 @@ menu "Graphics support"
source "drivers/char/agp/Kconfig"
+source "drivers/gpu/vga/Kconfig"
+
source "drivers/gpu/drm/Kconfig"
config VGASTATE
diff --git a/include/linux/vgaarb.h b/include/linux/vgaarb.h
new file mode 100644
index 0000000..cdc4ebb
--- /dev/null
+++ b/include/linux/vgaarb.h
@@ -0,0 +1,170 @@
+/*
+ * vgaarb.c
+ *
+ * (C) Copyright 2005 Benjamin Herrenschmidt <[email protected]>
+ * (C) Copyright 2007 Paulo R. Zanoni <[email protected]>
+ * (C) Copyright 2007, 2009 Tiago Vignatti <[email protected]>
+ */
+
+#ifndef LINUX_VGA_H
+
+#include <asm/vga.h>
+
+/* Legacy VGA regions */
+#define VGA_RSRC_NONE 0x00
+#define VGA_RSRC_LEGACY_IO 0x01
+#define VGA_RSRC_LEGACY_MEM 0x02
+#define VGA_RSRC_LEGACY_MASK (VGA_RSRC_LEGACY_IO | VGA_RSRC_LEGACY_MEM)
+/* Non-legacy access */
+#define VGA_RSRC_NORMAL_IO 0x04
+#define VGA_RSRC_NORMAL_MEM 0x08
+
+/* Passing that instead of a pci_dev to use the system "default"
+ * device, that is the one used by vgacon. Archs will probably
+ * have to provide their own vga_default_device();
+ */
+#define VGA_DEFAULT_DEVICE (NULL)
+
+/* For use by clients */
+
+/**
+ * vga_set_legacy_decoding
+ *
+ * @pdev: pci device of the VGA card
+ * @decodes: bit mask of what legacy regions the card decodes
+ *
+ * Indicates to the arbiter if the card decodes legacy VGA IOs,
+ * legacy VGA Memory, both, or none. All cards default to both,
+ * the card driver (fbdev for example) should tell the arbiter
+ * if it has disabled legacy decoding, so the card can be left
+ * out of the arbitration process (and can be safe to take
+ * interrupts at any time.
+ */
+extern void vga_set_legacy_decoding(struct pci_dev *pdev,
+ unsigned int decodes);
+
+/**
+ * vga_get - acquire & locks VGA resources
+ *
+ * pdev: pci device of the VGA card or NULL for the system default
+ * rsrc: bit mask of resources to acquire and lock
+ * interruptible: blocking should be interruptible by signals ?
+ *
+ * This function acquires VGA resources for the given
+ * card and mark those resources locked. If the resource requested
+ * are "normal" (and not legacy) resources, the arbiter will first check
+ * wether the card is doing legacy decoding for that type of resource. If
+ * yes, the lock is "converted" into a legacy resource lock.
+ * The arbiter will first look for all VGA cards that might conflict
+ * and disable their IOs and/or Memory access, inlcuding VGA forwarding
+ * on P2P bridges if necessary, so that the requested resources can
+ * be used. Then, the card is marked as locking these resources and
+ * the IO and/or Memory accesse are enabled on the card (including
+ * VGA forwarding on parent P2P bridges if any).
+ * This function will block if some conflicting card is already locking
+ * one of the required resources (or any resource on a different bus
+ * segment, since P2P bridges don't differenciate VGA memory and IO
+ * afaik). You can indicate wether this blocking should be interruptible
+ * by a signal (for userland interface) or not.
+ * Must not be called at interrupt time or in atomic context.
+ * If the card already owns the resources, the function succeeds.
+ * Nested calls are supported (a per-resource counter is maintained)
+ */
+
+extern int vga_get(struct pci_dev *pdev, unsigned int rsrc,
+ int interruptible);
+
+/**
+ * vga_get_interruptible
+ *
+ * Shortcut to vga_get
+ */
+
+static inline int vga_get_interruptible(struct pci_dev *pdev,
+ unsigned int rsrc)
+{
+ return vga_get(pdev, rsrc, 1);
+}
+
+/**
+ * vga_get_interruptible
+ *
+ * Shortcut to vga_get
+ */
+
+static inline int vga_get_uninterruptible(struct pci_dev *pdev,
+ unsigned int rsrc)
+{
+ return vga_get(pdev, rsrc, 0);
+}
+
+/**
+ * vga_tryget - try to acquire & lock legacy VGA resources
+ *
+ * @pdev: pci devivce of VGA card or NULL for system default
+ * @rsrc: bit mask of resources to acquire and lock
+ *
+ * This function performs the same operation as vga_get(), but
+ * will return an error (-EBUSY) instead of blocking if the resources
+ * are already locked by another card. It can be called in any context
+ */
+
+extern int vga_tryget(struct pci_dev *pdev, unsigned int rsrc);
+
+/**
+ * vga_put - release lock on legacy VGA resources
+ *
+ * @pdev: pci device of VGA card or NULL for system default
+ * @rsrc: but mask of resource to release
+ *
+ * This function releases resources previously locked by vga_get()
+ * or vga_tryget(). The resources aren't disabled right away, so
+ * that a subsequence vga_get() on the same card will succeed
+ * immediately. Resources have a counter, so locks are only
+ * released if the counter reaches 0.
+ */
+
+extern void vga_put(struct pci_dev *pdev, unsigned int rsrc);
+
+
+/**
+ * vga_default_device
+ *
+ * This can be defined by the platform. The default implementation
+ * is rather dumb and will probably only work properly on single
+ * vga card setups and/or x86 platforms.
+ *
+ * If your VGA default device is not PCI, you'll have to return
+ * NULL here. In this case, I assume it will not conflict with
+ * any PCI card. If this is not true, I'll have to define two archs
+ * hooks for enabling/disabling the VGA default device if that is
+ * possible. This may be a problem with real _ISA_ VGA cards, in
+ * addition to a PCI one. I don't know at this point how to deal
+ * with that card. Can theirs IOs be disabled at all ? If not, then
+ * I suppose it's a matter of having the proper arch hook telling
+ * us about it, so we basically never allow anybody to succeed a
+ * vga_get()...
+ */
+
+#ifndef __ARCH_HAS_VGA_DEFAULT_DEVICE
+extern struct pci_dev *vga_default_device(void);
+#endif
+
+/**
+ * vga_conflicts
+ *
+ * Architectures should define this if they have several
+ * independant PCI domains that can afford concurrent VGA
+ * decoding
+ */
+
+#ifndef __ARCH_HAS_VGA_CONFLICT
+static inline int vga_conflicts(struct pci_dev *p1, struct pci_dev *p2)
+{
+ return 1;
+}
+#endif
+
+void vga_set_irq_callback(struct pci_dev *pdev, void *cookie, void (*irq_func)(void *cookie, bool enable));
+
+#endif /* LINUX_VGA_H */
--
1.5.6.3
From: Dave Airlie <[email protected]>
This adds 3 things to the drm,
for non-modesetting drivers, it adds an irq hook
for the vga arb to disable/enable irqs around vga arb times.
radeon/kms: adds support to disable VGA decoding on r100->r500 under KMS
intel/kms: add support to disable VGA decoding on hopefully all Intel GPUs
Signed-off-by: Dave Airlie <[email protected]>
Acked-by: Tiago Vignatti <[email protected]>
---
drivers/gpu/drm/drm_irq.c | 24 ++++++++++++++++++++++++
drivers/gpu/drm/i915/i915_dma.c | 6 ++++++
drivers/gpu/drm/i915/i915_reg.h | 1 +
drivers/gpu/drm/i915/intel_display.c | 17 +++++++++++++++++
drivers/gpu/drm/i915/intel_drv.h | 1 +
drivers/gpu/drm/radeon/r100.c | 9 +++++++++
drivers/gpu/drm/radeon/radeon.h | 2 ++
drivers/gpu/drm/radeon/radeon_asic.h | 9 +++++++++
drivers/gpu/drm/radeon/radeon_device.c | 5 ++++-
include/drm/drmP.h | 3 +++
10 files changed, 76 insertions(+), 1 deletions(-)
diff --git a/drivers/gpu/drm/drm_irq.c b/drivers/gpu/drm/drm_irq.c
index b4a3dbc..436889e 100644
--- a/drivers/gpu/drm/drm_irq.c
+++ b/drivers/gpu/drm/drm_irq.c
@@ -37,6 +37,7 @@
#include <linux/interrupt.h> /* For task queue support */
+#include <linux/vgaarb.h>
/**
* Get interrupt from bus id.
*
@@ -171,6 +172,23 @@ err:
}
EXPORT_SYMBOL(drm_vblank_init);
+static void drm_irq_vgaarb_nokms(void *cookie, bool state)
+{
+ struct drm_device *dev = cookie;
+
+ if (dev->driver->vgaarb_irq) {
+ dev->driver->vgaarb_irq(dev, state);
+ return;
+ }
+
+ if (state)
+ dev->driver->irq_uninstall(dev);
+ else {
+ dev->driver->irq_preinstall(dev);
+ dev->driver->irq_postinstall(dev);
+ }
+}
+
/**
* Install IRQ handler.
*
@@ -231,6 +249,9 @@ int drm_irq_install(struct drm_device *dev)
return ret;
}
+ if (!drm_core_check_feature(dev, DRIVER_MODESET))
+ vga_set_irq_callback(dev->pdev, (void *)dev, drm_irq_vgaarb_nokms);
+
/* After installing handler */
ret = dev->driver->irq_postinstall(dev);
if (ret < 0) {
@@ -279,6 +300,9 @@ int drm_irq_uninstall(struct drm_device * dev)
DRM_DEBUG("irq=%d\n", dev->pdev->irq);
+ if (!drm_core_check_feature(dev, DRIVER_MODESET))
+ vga_set_irq_callback(dev->pdev, NULL, NULL);
+
dev->driver->irq_uninstall(dev);
free_irq(dev->pdev->irq, dev);
diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c
index 8c47831..391e496 100644
--- a/drivers/gpu/drm/i915/i915_dma.c
+++ b/drivers/gpu/drm/i915/i915_dma.c
@@ -33,6 +33,7 @@
#include "i915_drm.h"
#include "i915_drv.h"
+#include <linux/vgaarb.h>
#define I915_DRV "i915_drv"
/* Really want an OS-independent resettable timer. Would like to have
@@ -1029,6 +1030,11 @@ static int i915_load_modeset_init(struct drm_device *dev,
if (ret)
DRM_INFO("failed to find VBIOS tables\n");
+ /* tell the VGA arbiter to set off */
+ ret = intel_modeset_vga_disable(dev);
+ if (ret == 0)
+ vga_set_legacy_decoding(dev->pdev, VGA_RSRC_NONE);
+
ret = drm_irq_install(dev);
if (ret)
goto destroy_ringbuffer;
diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
index 6c08584..3e3145e 100644
--- a/drivers/gpu/drm/i915/i915_reg.h
+++ b/drivers/gpu/drm/i915/i915_reg.h
@@ -30,6 +30,7 @@
* fb aperture size and the amount of pre-reserved memory.
*/
#define INTEL_GMCH_CTRL 0x52
+#define INTEL_GMCH_VGA_DISABLE (1 << 1)
#define INTEL_GMCH_ENABLED 0x4
#define INTEL_GMCH_MEM_MASK 0x1
#define INTEL_GMCH_MEM_64M 0x1
diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
index 508838e..ad7ef01 100644
--- a/drivers/gpu/drm/i915/intel_display.c
+++ b/drivers/gpu/drm/i915/intel_display.c
@@ -3188,3 +3188,20 @@ struct drm_encoder *intel_best_encoder(struct drm_connector *connector)
return &intel_output->enc;
}
+
+int intel_modeset_vga_disable(struct drm_device *dev)
+{
+ struct pci_dev *bridge_dev;
+ u16 gmch_ctrl;
+
+ bridge_dev = pci_get_bus_and_slot(0, PCI_DEVFN(0, 0));
+ if (!bridge_dev) {
+ DRM_ERROR("Can't disable VGA, no bridge\n");
+ return -1;
+ }
+
+ pci_read_config_word(bridge_dev, INTEL_GMCH_CTRL, &gmch_ctrl);
+ gmch_ctrl |= INTEL_GMCH_VGA_DISABLE;
+ pci_write_config_word(bridge_dev, INTEL_GMCH_CTRL, gmch_ctrl);
+ return 0;
+}
diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h
index 004541c..ab5be5b 100644
--- a/drivers/gpu/drm/i915/intel_drv.h
+++ b/drivers/gpu/drm/i915/intel_drv.h
@@ -154,4 +154,5 @@ extern int intel_framebuffer_create(struct drm_device *dev,
struct drm_mode_fb_cmd *mode_cmd,
struct drm_framebuffer **fb,
struct drm_gem_object *obj);
+extern int intel_modeset_vga_disable(struct drm_device *dev);
#endif /* __INTEL_DRV_H__ */
diff --git a/drivers/gpu/drm/radeon/r100.c b/drivers/gpu/drm/radeon/r100.c
index c550932..1fa6723 100644
--- a/drivers/gpu/drm/radeon/r100.c
+++ b/drivers/gpu/drm/radeon/r100.c
@@ -1256,6 +1256,15 @@ static void r100_vram_get_type(struct radeon_device *rdev)
}
}
+void r100_vga_disable(struct radeon_device *rdev)
+{
+ uint32_t temp;
+ temp = RREG32(RADEON_CONFIG_CNTL);
+ temp &= ~(1<<8);
+ temp |= (1<<9);
+ WREG32(RADEON_CONFIG_CNTL, temp);
+}
+
void r100_vram_info(struct radeon_device *rdev)
{
r100_vram_get_type(rdev);
diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h
index d61f2fc..583bfb2 100644
--- a/drivers/gpu/drm/radeon/radeon.h
+++ b/drivers/gpu/drm/radeon/radeon.h
@@ -499,6 +499,7 @@ struct radeon_asic {
int (*init)(struct radeon_device *rdev);
void (*errata)(struct radeon_device *rdev);
void (*vram_info)(struct radeon_device *rdev);
+ void (*vga_disable)(struct radeon_device *rdev);
int (*gpu_reset)(struct radeon_device *rdev);
int (*mc_init)(struct radeon_device *rdev);
void (*mc_fini)(struct radeon_device *rdev);
@@ -773,6 +774,7 @@ static inline void radeon_ring_write(struct radeon_device *rdev, uint32_t v)
#define radeon_cs_parse(p) rdev->asic->cs_parse((p))
#define radeon_errata(rdev) (rdev)->asic->errata((rdev))
#define radeon_vram_info(rdev) (rdev)->asic->vram_info((rdev))
+#define radeon_vga_disable(rdev) (rdev)->asic->vga_disable((rdev))
#define radeon_gpu_reset(rdev) (rdev)->asic->gpu_reset((rdev))
#define radeon_mc_init(rdev) (rdev)->asic->mc_init((rdev))
#define radeon_mc_fini(rdev) (rdev)->asic->mc_fini((rdev))
diff --git a/drivers/gpu/drm/radeon/radeon_asic.h b/drivers/gpu/drm/radeon/radeon_asic.h
index e2e5673..4586819 100644
--- a/drivers/gpu/drm/radeon/radeon_asic.h
+++ b/drivers/gpu/drm/radeon/radeon_asic.h
@@ -46,6 +46,7 @@ uint32_t r100_mm_rreg(struct radeon_device *rdev, uint32_t reg);
void r100_mm_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v);
void r100_errata(struct radeon_device *rdev);
void r100_vram_info(struct radeon_device *rdev);
+void r100_vga_disable(struct radeon_device *rdev);
int r100_gpu_reset(struct radeon_device *rdev);
int r100_mc_init(struct radeon_device *rdev);
void r100_mc_fini(struct radeon_device *rdev);
@@ -76,6 +77,7 @@ static struct radeon_asic r100_asic = {
.init = &r100_init,
.errata = &r100_errata,
.vram_info = &r100_vram_info,
+ .vga_disable = &r100_vga_disable,
.gpu_reset = &r100_gpu_reset,
.mc_init = &r100_mc_init,
.mc_fini = &r100_mc_fini,
@@ -132,6 +134,7 @@ static struct radeon_asic r300_asic = {
.init = &r300_init,
.errata = &r300_errata,
.vram_info = &r300_vram_info,
+ .vga_disable = &r100_vga_disable,
.gpu_reset = &r300_gpu_reset,
.mc_init = &r300_mc_init,
.mc_fini = &r300_mc_fini,
@@ -169,6 +172,7 @@ static struct radeon_asic r420_asic = {
.init = &r300_init,
.errata = &r420_errata,
.vram_info = &r420_vram_info,
+ .vga_disable = &r100_vga_disable,
.gpu_reset = &r300_gpu_reset,
.mc_init = &r420_mc_init,
.mc_fini = &r420_mc_fini,
@@ -213,6 +217,7 @@ static struct radeon_asic rs400_asic = {
.init = &r300_init,
.errata = &rs400_errata,
.vram_info = &rs400_vram_info,
+ .vga_disable = &r100_vga_disable,
.gpu_reset = &r300_gpu_reset,
.mc_init = &rs400_mc_init,
.mc_fini = &rs400_mc_fini,
@@ -258,6 +263,7 @@ static struct radeon_asic rs600_asic = {
.init = &r300_init,
.errata = &rs600_errata,
.vram_info = &rs600_vram_info,
+ .vga_disable = &r100_vga_disable,
.gpu_reset = &r300_gpu_reset,
.mc_init = &rs600_mc_init,
.mc_fini = &rs600_mc_fini,
@@ -298,6 +304,7 @@ static struct radeon_asic rs690_asic = {
.init = &r300_init,
.errata = &rs690_errata,
.vram_info = &rs690_vram_info,
+ .vga_disable = &r100_vga_disable,
.gpu_reset = &r300_gpu_reset,
.mc_init = &rs690_mc_init,
.mc_fini = &rs690_mc_fini,
@@ -343,6 +350,7 @@ static struct radeon_asic rv515_asic = {
.init = &rv515_init,
.errata = &rv515_errata,
.vram_info = &rv515_vram_info,
+ .vga_disable = &r100_vga_disable,
.gpu_reset = &rv515_gpu_reset,
.mc_init = &rv515_mc_init,
.mc_fini = &rv515_mc_fini,
@@ -381,6 +389,7 @@ static struct radeon_asic r520_asic = {
.init = &rv515_init,
.errata = &r520_errata,
.vram_info = &r520_vram_info,
+ .vga_disable = &r100_vga_disable,
.gpu_reset = &rv515_gpu_reset,
.mc_init = &r520_mc_init,
.mc_fini = &r520_mc_fini,
diff --git a/drivers/gpu/drm/radeon/radeon_device.c b/drivers/gpu/drm/radeon/radeon_device.c
index f97563d..c02f655 100644
--- a/drivers/gpu/drm/radeon/radeon_device.c
+++ b/drivers/gpu/drm/radeon/radeon_device.c
@@ -29,6 +29,7 @@
#include <drm/drmP.h>
#include <drm/drm_crtc_helper.h>
#include <drm/radeon_drm.h>
+#include <linux/vgaarb.h>
#include "radeon_reg.h"
#include "radeon.h"
#include "radeon_asic.h"
@@ -516,7 +517,9 @@ int radeon_device_init(struct radeon_device *rdev,
/* Initialize surface registers */
radeon_surface_init(rdev);
- /* TODO: disable VGA need to use VGA request */
+ radeon_vga_disable(rdev);
+ vga_set_legacy_decoding(rdev->pdev, VGA_RSRC_NONE);
+
/* BIOS*/
if (!radeon_get_bios(rdev)) {
if (ASIC_IS_AVIVO(rdev))
diff --git a/include/drm/drmP.h b/include/drm/drmP.h
index 45b67d9..95106c7 100644
--- a/include/drm/drmP.h
+++ b/include/drm/drmP.h
@@ -786,6 +786,9 @@ struct drm_driver {
int (*gem_init_object) (struct drm_gem_object *obj);
void (*gem_free_object) (struct drm_gem_object *obj);
+ /* vga arb irq handler */
+ void (*vgaarb_irq)(struct drm_device *dev, bool state);
+
/* Driver private ops for this object */
struct vm_operations_struct *gem_vm_ops;
--
1.5.6.3
Signed-off-by: Tiago Vignatti <[email protected]>
---
Documentation/vgaarbiter.txt | 197 ++++++++++++++++++++++++++++++++++++++++++
1 files changed, 197 insertions(+), 0 deletions(-)
create mode 100644 Documentation/vgaarbiter.txt
diff --git a/Documentation/vgaarbiter.txt b/Documentation/vgaarbiter.txt
new file mode 100644
index 0000000..37d3126
--- /dev/null
+++ b/Documentation/vgaarbiter.txt
@@ -0,0 +1,197 @@
+
+VGA Arbiter
+===========
+
+Graphic devices are accessed through ranges in I/O or memory space. While most
+modern devices allow relocation of such ranges, some "Legacy" VGA devices
+implemented on PCI will typically have the same "hard-decoded" addresses as
+they did on ISA. For more details see "PCI Bus Binding to IEEE Std 1275-1994
+Standard for Boot (Initialization Configuration) Firmware Revision 2.1"
+Section 7, Legacy Devices.
+
+The Resource Access Control (RAC) module inside the X server currently does
+the task of arbitration when more than one legacy device co-exists on the same
+machine. But the problem happens when these devices are trying to be accessed
+by different userspace clients (e.g. two server in parallel). Their address
+assignments conflict. Therefore an arbitration scheme _outside_ of the X
+server is needed to control the sharing of these resources. This document
+introduces the operation of the VGA arbiter implemented for Linux kernel.
+
+----------------------------------------------------------------------------
+
+I. Details and Theory of Operation
+ I.1 vgaarb
+ I.2 libpciaccess
+ I.3 xf86VGAArbiter (X server implementation)
+II. Open Issues / Bugs
+III.Credits
+VI. References
+
+
+I. Details and Theory of Operation
+==================================
+
+I.1 vgaarb
+----------
+
+The vgaarb is a module of the Linux Kernel. When it is initially loaded, it
+scans all PCI devices and add the VGA ones inside the arbitration. The arbiter
+then enables/disables the decoding of VGA legacy instructions by different
+devices which are trying to use them on the same machine. The device which
+does not want/need to use the arbiter may explicitly tell it by calling
+vga_arb_decodes() from the VGA library function. Driver must take a special
+care here.
+
+Basically the kernel exports a char device interface (/dev/vga_arbiter) to the
+clients. It has the usual semantics:
+
+ open : open user instance of the arbiter. By default, it's attached to
+ the default VGA device of the system.
+
+ close : close user instance. Release locks made by the user
+
+ read : return a string indicating the status of the target like:
+
+ "<card_ID>,decodes=<io_state>,owns=<io_state>,locks=<io_state> (ic,mc)"
+
+ An IO state string is of the form {io,mem,io+mem,none}, mc and
+ ic are respectively mem and io lock counts (for debugging/
+ diagnostic only). "decodes" indicate what the card currently
+ decodes, "owns" indicates what is currently enabled on it, and
+ "locks" indicates what is locked by this card. If the card is
+ unplugged, we get "invalid" then for card_ID and an -ENODEV
+ error is returned for any command until a new card is targeted.
+
+
+ write : write a command to the arbiter. List of commands:
+
+ target <card_ID> : switch target to card <card_ID> (see below)
+ lock <io_state> : acquires locks on target ("none" is an invalid io_state)
+ trylock <io_state> : non-blocking acquire locks on target (returns EBUSY if
+ unsuccessful)
+ unlock <io_state> : release locks on target
+ unlock all : release all locks on target held by this user (not
+ implemented yet)
+ decodes <io_state> : set the legacy decoding attributes for the card
+
+ poll : event if something changes on any card (not just the
+ target)
+
+ card_ID is of the form "PCI:domain:bus:dev.fn". It can be set to "default"
+ to go back to the system default card (TODO: not implemented yet). Currently,
+ only PCI is supported as a prefix, but the userland API may support other bus
+ types in the future, even if the current kernel implementation doesn't.
+
+Note about locks:
+
+The driver keeps track of which user has which locks on which card. It
+supports stacking, like the kernel one. This complexifies the implementation
+a bit, but makes the arbiter more tolerant to user space problems and able
+to properly cleanup in all cases when a process dies.
+Currently, a max of 16 cards can have locks simultaneously issued from
+user space for a given user (file descriptor instance) of the arbiter.
+
+If the device is hot-{un,}plugged, there is a hook inside the module to notify
+them being added/removed in the system and automatically added/removed in
+the arbiter.
+
+There's also a in-kernel API of the arbiter in the case of DRM, vgacon and
+others which may use the arbiter.
+
+
+I.2 libpciaccess
+----------------
+
+To use the vga arbiter char device it was implemented a serie of functions
+inside the pciaccess library. Two fields were added to struct pci_device for
+this be possible:
+
+ /* the type of resource decoded by the device */
+ int vgaarb_rsrc;
+ /* the file descriptor (/dev/vga_arbiter) */
+ int vgaarb_fd;
+
+
+and the functions:
+
+These functions below acquire VGA resources for the given card and mark those
+resources as locked. If the resources requested are "normal" (and not legacy)
+resources, the arbiter will first check whether the card is doing legacy
+decoding for that type of resource. If yes, the lock is "converted" into a
+legacy resource lock. The arbiter will first look for all VGA cards that
+might conflict and disable their IOs and/or Memory access, including VGA
+forwarding on P2P bridges if necessary, so that the requested resources can
+be used. Then, the card is marked as locking these resources and the IO and/or
+Memory access is enabled on the card (including VGA forwarding on parent
+P2P bridges if any). In the case of vga_arb_lock(), the function will block
+if some conflicting card is already locking one of the required resources (or
+any resource on a different bus segment, since P2P bridges don't differentiate
+VGA memory and IO afaik). If the card already owns the resources, the function
+succeeds. vga_arb_trylock() will return (-EBUSY) instead of blocking. Nested
+calls are supported (a per-resource counter is maintained).
+
+
+Set the target device of this client.
+ int pci_device_vgaarb_set_target (struct pci_device *dev);
+
+
+For instance, in x86 if two devices on the same bus want to lock different
+resources, both will succeed (lock). If devices are in different buses and
+trying to lock different resources, only the first who tried succeeds.
+ int pci_device_vgaarb_lock (struct pci_device *dev);
+ int pci_device_vgaarb_trylock (struct pci_device *dev);
+
+Unlock resources of device.
+ int pci_device_vgaarb_unlock (struct pci_device *dev);
+
+Indicates to the arbiter if the card decodes legacy VGA IOs, legacy VGA
+Memory, both, or none. All cards default to both, the card driver (fbdev for
+example) should tell the arbiter if it has disabled legacy decoding, so the
+card can be left out of the arbitration process (and can be safe to take
+interrupts at any time.
+ int pci_device_vgaarb_decodes (struct pci_device *dev);
+
+Connects to the arbiter device, allocates the struct
+ int pci_device_vgaarb_init (struct pci_device *dev);
+
+Close the connection
+ void pci_device_vgaarb_fini (struct pci_device *dev);
+
+
+I.3 xf86VGAArbiter (X server implementation)
+--------------------------------------------
+
+(TODO)
+
+This work will probably remove the RAC (Resource Access Control) code inside
+Xorg DDX. Complain: RAC is non-OS dependent. VGA Arbiter is.
+
+
+II. Open Issues / Bugs
+======================
+
+Current status:
+ - two X servers is parallel works (multiseat style)
+ - non-bottable card works
+ - secondary card works
+ - xinemara-like _doesn't_ works (need to trace more paths paths that touch
+ the registers and bracket with lock/unlock)
+
+
+III. Credits
+===========
+
+Benjamin Herrenschmidt (IBM?) started this work when he discussed such design
+with the Xorg community in 2005 [1, 2]. In the end of 2007, Paulo Zanoni and
+Tiago Vignatti (both of C3SL/Federal University of Paraná) proceeded his work
+enhancing the kernel code to adapt as a kernel module and also did the
+implementation of the user space side [3]. Now (2009) Tiago Vignatti is
+finally trying to push such work upstream.
+
+
+VI. References
+==============
+
+[1] http://lists.freedesktop.org/archives/xorg/2005-March/006663.html
+[2] http://lists.freedesktop.org/archives/xorg/2005-March/006745.html
+[3] http://lists.freedesktop.org/archives/xorg/2007-October/029507.html
--
1.5.6.3
On Thursday 16 July 2009 09:48:30 am Tiago Vignatti wrote:
> Signed-off-by: Tiago Vignatti <[email protected]>
> ---
> Documentation/vgaarbiter.txt | 197 ++++++++++++++++++++++++++++++++++++++++++
> 1 files changed, 197 insertions(+), 0 deletions(-)
> create mode 100644 Documentation/vgaarbiter.txt
I think the documentation update should be part of the same patch
that adds the functionality. Then the docs will always match
the code.
Bjorn
> diff --git a/Documentation/vgaarbiter.txt b/Documentation/vgaarbiter.txt
> new file mode 100644
> index 0000000..37d3126
> --- /dev/null
> +++ b/Documentation/vgaarbiter.txt
> @@ -0,0 +1,197 @@
> +
> +VGA Arbiter
> +===========
> +
> +Graphic devices are accessed through ranges in I/O or memory space. While most
> +modern devices allow relocation of such ranges, some "Legacy" VGA devices
> +implemented on PCI will typically have the same "hard-decoded" addresses as
> +they did on ISA. For more details see "PCI Bus Binding to IEEE Std 1275-1994
> +Standard for Boot (Initialization Configuration) Firmware Revision 2.1"
> +Section 7, Legacy Devices.
> +
> +The Resource Access Control (RAC) module inside the X server currently does
> +the task of arbitration when more than one legacy device co-exists on the same
> +machine. But the problem happens when these devices are trying to be accessed
> +by different userspace clients (e.g. two server in parallel). Their address
> +assignments conflict. Therefore an arbitration scheme _outside_ of the X
> +server is needed to control the sharing of these resources. This document
> +introduces the operation of the VGA arbiter implemented for Linux kernel.
> +
> +----------------------------------------------------------------------------
> +
> +I. Details and Theory of Operation
> + I.1 vgaarb
> + I.2 libpciaccess
> + I.3 xf86VGAArbiter (X server implementation)
> +II. Open Issues / Bugs
> +III.Credits
> +VI. References
> +
> +
> +I. Details and Theory of Operation
> +==================================
> +
> +I.1 vgaarb
> +----------
> +
> +The vgaarb is a module of the Linux Kernel. When it is initially loaded, it
> +scans all PCI devices and add the VGA ones inside the arbitration. The arbiter
> +then enables/disables the decoding of VGA legacy instructions by different
> +devices which are trying to use them on the same machine. The device which
> +does not want/need to use the arbiter may explicitly tell it by calling
> +vga_arb_decodes() from the VGA library function. Driver must take a special
> +care here.
> +
> +Basically the kernel exports a char device interface (/dev/vga_arbiter) to the
> +clients. It has the usual semantics:
> +
> + open : open user instance of the arbiter. By default, it's attached to
> + the default VGA device of the system.
> +
> + close : close user instance. Release locks made by the user
> +
> + read : return a string indicating the status of the target like:
> +
> + "<card_ID>,decodes=<io_state>,owns=<io_state>,locks=<io_state> (ic,mc)"
> +
> + An IO state string is of the form {io,mem,io+mem,none}, mc and
> + ic are respectively mem and io lock counts (for debugging/
> + diagnostic only). "decodes" indicate what the card currently
> + decodes, "owns" indicates what is currently enabled on it, and
> + "locks" indicates what is locked by this card. If the card is
> + unplugged, we get "invalid" then for card_ID and an -ENODEV
> + error is returned for any command until a new card is targeted.
> +
> +
> + write : write a command to the arbiter. List of commands:
> +
> + target <card_ID> : switch target to card <card_ID> (see below)
> + lock <io_state> : acquires locks on target ("none" is an invalid io_state)
> + trylock <io_state> : non-blocking acquire locks on target (returns EBUSY if
> + unsuccessful)
> + unlock <io_state> : release locks on target
> + unlock all : release all locks on target held by this user (not
> + implemented yet)
> + decodes <io_state> : set the legacy decoding attributes for the card
> +
> + poll : event if something changes on any card (not just the
> + target)
> +
> + card_ID is of the form "PCI:domain:bus:dev.fn". It can be set to "default"
> + to go back to the system default card (TODO: not implemented yet). Currently,
> + only PCI is supported as a prefix, but the userland API may support other bus
> + types in the future, even if the current kernel implementation doesn't.
> +
> +Note about locks:
> +
> +The driver keeps track of which user has which locks on which card. It
> +supports stacking, like the kernel one. This complexifies the implementation
> +a bit, but makes the arbiter more tolerant to user space problems and able
> +to properly cleanup in all cases when a process dies.
> +Currently, a max of 16 cards can have locks simultaneously issued from
> +user space for a given user (file descriptor instance) of the arbiter.
> +
> +If the device is hot-{un,}plugged, there is a hook inside the module to notify
> +them being added/removed in the system and automatically added/removed in
> +the arbiter.
> +
> +There's also a in-kernel API of the arbiter in the case of DRM, vgacon and
> +others which may use the arbiter.
> +
> +
> +I.2 libpciaccess
> +----------------
> +
> +To use the vga arbiter char device it was implemented a serie of functions
> +inside the pciaccess library. Two fields were added to struct pci_device for
> +this be possible:
> +
> + /* the type of resource decoded by the device */
> + int vgaarb_rsrc;
> + /* the file descriptor (/dev/vga_arbiter) */
> + int vgaarb_fd;
> +
> +
> +and the functions:
> +
> +These functions below acquire VGA resources for the given card and mark those
> +resources as locked. If the resources requested are "normal" (and not legacy)
> +resources, the arbiter will first check whether the card is doing legacy
> +decoding for that type of resource. If yes, the lock is "converted" into a
> +legacy resource lock. The arbiter will first look for all VGA cards that
> +might conflict and disable their IOs and/or Memory access, including VGA
> +forwarding on P2P bridges if necessary, so that the requested resources can
> +be used. Then, the card is marked as locking these resources and the IO and/or
> +Memory access is enabled on the card (including VGA forwarding on parent
> +P2P bridges if any). In the case of vga_arb_lock(), the function will block
> +if some conflicting card is already locking one of the required resources (or
> +any resource on a different bus segment, since P2P bridges don't differentiate
> +VGA memory and IO afaik). If the card already owns the resources, the function
> +succeeds. vga_arb_trylock() will return (-EBUSY) instead of blocking. Nested
> +calls are supported (a per-resource counter is maintained).
> +
> +
> +Set the target device of this client.
> + int pci_device_vgaarb_set_target (struct pci_device *dev);
> +
> +
> +For instance, in x86 if two devices on the same bus want to lock different
> +resources, both will succeed (lock). If devices are in different buses and
> +trying to lock different resources, only the first who tried succeeds.
> + int pci_device_vgaarb_lock (struct pci_device *dev);
> + int pci_device_vgaarb_trylock (struct pci_device *dev);
> +
> +Unlock resources of device.
> + int pci_device_vgaarb_unlock (struct pci_device *dev);
> +
> +Indicates to the arbiter if the card decodes legacy VGA IOs, legacy VGA
> +Memory, both, or none. All cards default to both, the card driver (fbdev for
> +example) should tell the arbiter if it has disabled legacy decoding, so the
> +card can be left out of the arbitration process (and can be safe to take
> +interrupts at any time.
> + int pci_device_vgaarb_decodes (struct pci_device *dev);
> +
> +Connects to the arbiter device, allocates the struct
> + int pci_device_vgaarb_init (struct pci_device *dev);
> +
> +Close the connection
> + void pci_device_vgaarb_fini (struct pci_device *dev);
> +
> +
> +I.3 xf86VGAArbiter (X server implementation)
> +--------------------------------------------
> +
> +(TODO)
> +
> +This work will probably remove the RAC (Resource Access Control) code inside
> +Xorg DDX. Complain: RAC is non-OS dependent. VGA Arbiter is.
> +
> +
> +II. Open Issues / Bugs
> +======================
> +
> +Current status:
> + - two X servers is parallel works (multiseat style)
> + - non-bottable card works
> + - secondary card works
> + - xinemara-like _doesn't_ works (need to trace more paths paths that touch
> + the registers and bracket with lock/unlock)
> +
> +
> +III. Credits
> +===========
> +
> +Benjamin Herrenschmidt (IBM?) started this work when he discussed such design
> +with the Xorg community in 2005 [1, 2]. In the end of 2007, Paulo Zanoni and
> +Tiago Vignatti (both of C3SL/Federal University of Paraná) proceeded his work
> +enhancing the kernel code to adapt as a kernel module and also did the
> +implementation of the user space side [3]. Now (2009) Tiago Vignatti is
> +finally trying to push such work upstream.
> +
> +
> +VI. References
> +==============
> +
> +[1] http://lists.freedesktop.org/archives/xorg/2005-March/006663.html
> +[2] http://lists.freedesktop.org/archives/xorg/2005-March/006745.html
> +[3] http://lists.freedesktop.org/archives/xorg/2007-October/029507.html
On Tue, Jul 28, 2009 at 09:25:50PM +0200, Bjorn Helgaas wrote:
> On Thursday 16 July 2009 09:48:30 am Tiago Vignatti wrote:
> > Signed-off-by: Tiago Vignatti <[email protected]>
> > ---
> > Documentation/vgaarbiter.txt | 197 ++++++++++++++++++++++++++++++++++++++++++
> > 1 files changed, 197 insertions(+), 0 deletions(-)
> > create mode 100644 Documentation/vgaarbiter.txt
>
> I think the documentation update should be part of the same patch
> that adds the functionality. Then the docs will always match
> the code.
I don't think some "normal" kernel developer will need such document.
Looks like we're doing nice cuts (well, Dave is) on this patch and coming in a
state that people won't need to understand in deep things related with VGA
arbiter before doing hacks with DRM/PCI drivers. Moreover, the vgaarb.h is
pretty well readable :)
Cheers,
Tiago
On Wednesday 29 July 2009 02:52:29 am Vignatti Tiago (Nokia-D/Helsinki) wrote:
> On Tue, Jul 28, 2009 at 09:25:50PM +0200, Bjorn Helgaas wrote:
> > On Thursday 16 July 2009 09:48:30 am Tiago Vignatti wrote:
> > > Signed-off-by: Tiago Vignatti <[email protected]>
> > > ---
> > > Documentation/vgaarbiter.txt | 197 ++++++++++++++++++++++++++++++++++++++++++
> > > 1 files changed, 197 insertions(+), 0 deletions(-)
> > > create mode 100644 Documentation/vgaarbiter.txt
> >
> > I think the documentation update should be part of the same patch
> > that adds the functionality. Then the docs will always match
> > the code.
>
> I don't think some "normal" kernel developer will need such document.
>
> Looks like we're doing nice cuts (well, Dave is) on this patch and coming in a
> state that people won't need to understand in deep things related with VGA
> arbiter before doing hacks with DRM/PCI drivers. Moreover, the vgaarb.h is
> pretty well readable :)
I guess I didn't make my point very well :-) I haven't even looked at
the code, so I'll take your word for the fact that it's readable without
extra documentation.
My point was that if you add a feature "X" and you add documentation
for "X", both should be in a single patch so the logically related things
are all in one commit.
Bjorn