5 Graphic devices are accessed through ranges in I/O or memory space. While most
6 modern devices allow relocation of such ranges, some "Legacy" VGA devices
7 implemented on PCI will typically have the same "hard-decoded" addresses as
8 they did on ISA. For more details see "PCI Bus Binding to IEEE Std 1275-1994
9 Standard for Boot (Initialization Configuration) Firmware Revision 2.1"
10 Section 7, Legacy Devices.
12 The Resource Access Control (RAC) module inside the X server [0] existed for
13 the legacy VGA arbitration task (besides other bus management tasks) when more
14 than one legacy device co-exists on the same machine. But the problem happens
15 when these devices are trying to be accessed by different userspace clients
16 (e.g. two server in parallel). Their address assignments conflict. Moreover,
17 ideally, being an userspace application, it is not the role of the the X
18 server to control bus resources. Therefore an arbitration scheme outside of
19 the X server is needed to control the sharing of these resources. This
20 document introduces the operation of the VGA arbiter implemented for Linux
23 ----------------------------------------------------------------------------
25 I. Details and Theory of Operation
28 I.3 xf86VGAArbiter (X server implementation)
33 I. Details and Theory of Operation
34 ==================================
39 The vgaarb is a module of the Linux Kernel. When it is initially loaded, it
40 scans all PCI devices and adds the VGA ones inside the arbitration. The
41 arbiter then enables/disables the decoding on different devices of the VGA
42 legacy instructions. Device which do not want/need to use the arbiter may
43 explicitly tell it by calling vga_set_legacy_decoding().
45 The kernel exports a char device interface (/dev/vga_arbiter) to the clients,
46 which has the following semantics:
48 open : open user instance of the arbiter. By default, it's attached to
49 the default VGA device of the system.
51 close : close user instance. Release locks made by the user
53 read : return a string indicating the status of the target like:
55 "<card_ID>,decodes=<io_state>,owns=<io_state>,locks=<io_state> (ic,mc)"
57 An IO state string is of the form {io,mem,io+mem,none}, mc and
58 ic are respectively mem and io lock counts (for debugging/
59 diagnostic only). "decodes" indicate what the card currently
60 decodes, "owns" indicates what is currently enabled on it, and
61 "locks" indicates what is locked by this card. If the card is
62 unplugged, we get "invalid" then for card_ID and an -ENODEV
63 error is returned for any command until a new card is targeted.
66 write : write a command to the arbiter. List of commands:
68 target <card_ID> : switch target to card <card_ID> (see below)
69 lock <io_state> : acquires locks on target ("none" is an invalid io_state)
70 trylock <io_state> : non-blocking acquire locks on target (returns EBUSY if
72 unlock <io_state> : release locks on target
73 unlock all : release all locks on target held by this user (not
75 decodes <io_state> : set the legacy decoding attributes for the card
77 poll : event if something changes on any card (not just the
80 card_ID is of the form "PCI:domain:bus:dev.fn". It can be set to "default"
81 to go back to the system default card (TODO: not implemented yet). Currently,
82 only PCI is supported as a prefix, but the userland API may support other bus
83 types in the future, even if the current kernel implementation doesn't.
87 The driver keeps track of which user has which locks on which card. It
88 supports stacking, like the kernel one. This complexifies the implementation
89 a bit, but makes the arbiter more tolerant to user space problems and able
90 to properly cleanup in all cases when a process dies.
91 Currently, a max of 16 cards can have locks simultaneously issued from
92 user space for a given user (file descriptor instance) of the arbiter.
94 In the case of devices hot-{un,}plugged, there is a hook - pci_notify() - to
95 notify them being added/removed in the system and automatically added/removed
98 There's also a in-kernel API of the arbiter in the case of DRM, vgacon and
99 others which may use the arbiter.
105 To use the vga arbiter char device it was implemented an API inside the
106 libpciaccess library. One field was added to struct pci_device (each device
109 /* the type of resource decoded by the device */
112 Besides it, in pci_system were added:
116 struct pci_device *vga_target;
117 struct pci_device *vga_default_dev;
120 The vga_count is usually need to keep informed how many cards are being
121 arbitrated, so for instance if there's only one then it can totally escape the
125 These functions below acquire VGA resources for the given card and mark those
126 resources as locked. If the resources requested are "normal" (and not legacy)
127 resources, the arbiter will first check whether the card is doing legacy
128 decoding for that type of resource. If yes, the lock is "converted" into a
129 legacy resource lock. The arbiter will first look for all VGA cards that
130 might conflict and disable their IOs and/or Memory access, including VGA
131 forwarding on P2P bridges if necessary, so that the requested resources can
132 be used. Then, the card is marked as locking these resources and the IO and/or
133 Memory access is enabled on the card (including VGA forwarding on parent
134 P2P bridges if any). In the case of vga_arb_lock(), the function will block
135 if some conflicting card is already locking one of the required resources (or
136 any resource on a different bus segment, since P2P bridges don't differentiate
137 VGA memory and IO afaik). If the card already owns the resources, the function
138 succeeds. vga_arb_trylock() will return (-EBUSY) instead of blocking. Nested
139 calls are supported (a per-resource counter is maintained).
142 Set the target device of this client.
143 int pci_device_vgaarb_set_target (struct pci_device *dev);
146 For instance, in x86 if two devices on the same bus want to lock different
147 resources, both will succeed (lock). If devices are in different buses and
148 trying to lock different resources, only the first who tried succeeds.
149 int pci_device_vgaarb_lock (void);
150 int pci_device_vgaarb_trylock (void);
152 Unlock resources of device.
153 int pci_device_vgaarb_unlock (void);
155 Indicates to the arbiter if the card decodes legacy VGA IOs, legacy VGA
156 Memory, both, or none. All cards default to both, the card driver (fbdev for
157 example) should tell the arbiter if it has disabled legacy decoding, so the
158 card can be left out of the arbitration process (and can be safe to take
159 interrupts at any time.
160 int pci_device_vgaarb_decodes (int new_vgaarb_rsrc);
162 Connects to the arbiter device, allocates the struct
163 int pci_device_vgaarb_init (void);
166 void pci_device_vgaarb_fini (void);
169 I.3 xf86VGAArbiter (X server implementation)
170 --------------------------------------------
174 X server basically wraps all the functions that touch VGA registers somehow.
180 Benjamin Herrenschmidt (IBM?) started this work when he discussed such design
181 with the Xorg community in 2005 [1, 2]. In the end of 2007, Paulo Zanoni and
182 Tiago Vignatti (both of C3SL/Federal University of ParanĂ¡) proceeded his work
183 enhancing the kernel code to adapt as a kernel module and also did the
184 implementation of the user space side [3]. Now (2009) Tiago Vignatti and Dave
185 Airlie finally put this work in shape and queued to Jesse Barnes' PCI tree.
191 [0] http://cgit.freedesktop.org/xorg/xserver/commit/?id=4b42448a2388d40f257774fbffdccaea87bd0347
192 [1] http://lists.freedesktop.org/archives/xorg/2005-March/006663.html
193 [2] http://lists.freedesktop.org/archives/xorg/2005-March/006745.html
194 [3] http://lists.freedesktop.org/archives/xorg/2007-October/029507.html