Merge tag 'net-next-6.3' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-nextgrafted

Pull networking updates from Jakub Kicinski:
 "Core:

   - Add dedicated kmem_cache for typical/small skb->head, avoid having
     to access struct page at kfree time, and improve memory use.

   - Introduce sysctl to set default RPS configuration for new netdevs.

   - Define Netlink protocol specification format which can be used to
     describe messages used by each family and auto-generate parsers.
     Add tools for generating kernel data structures and uAPI headers.

   - Expose all net/core sysctls inside netns.

   - Remove 4s sleep in netpoll if carrier is instantly detected on
     boot.

   - Add configurable limit of MDB entries per port, and port-vlan.

   - Continue populating drop reasons throughout the stack.

   - Retire a handful of legacy Qdiscs and classifiers.

  Protocols:

   - Support IPv4 big TCP (TSO frames larger than 64kB).

   - Add IP_LOCAL_PORT_RANGE socket option, to control local port range
     on socket by socket basis.

   - Track and report in procfs number of MPTCP sockets used.

   - Support mixing IPv4 and IPv6 flows in the in-kernel MPTCP path
     manager.

   - IPv6: don't check net.ipv6.route.max_size and rely on garbage
     collection to free memory (similarly to IPv4).

   - Support Penultimate Segment Pop (PSP) flavor in SRv6 (RFC8986).

   - ICMP: add per-rate limit counters.

   - Add support for user scanning requests in ieee802154.

   - Remove static WEP support.

   - Support minimal Wi-Fi 7 Extremely High Throughput (EHT) rate
     reporting.

   - WiFi 7 EHT channel puncturing support (client & AP).

  BPF:

   - Add a rbtree data structure following the "next-gen data structure"
     precedent set by recently added linked list, that is, by using
     kfunc + kptr instead of adding a new BPF map type.

   - Expose XDP hints via kfuncs with initial support for RX hash and
     timestamp metadata.

   - Add BPF_F_NO_TUNNEL_KEY extension to bpf_skb_set_tunnel_key to
     better support decap on GRE tunnel devices not operating in collect
     metadata.

   - Improve x86 JIT's codegen for PROBE_MEM runtime error checks.

   - Remove the need for trace_printk_lock for bpf_trace_printk and
     bpf_trace_vprintk helpers.

   - Extend libbpf's bpf_tracing.h support for tracing arguments of
     kprobes/uprobes and syscall as a special case.

   - Significantly reduce the search time for module symbols by
     livepatch and BPF.

   - Enable cpumasks to be used as kptrs, which is useful for tracing
     programs tracking which tasks end up running on which CPUs in
     different time intervals.

   - Add support for BPF trampoline on s390x and riscv64.

   - Add capability to export the XDP features supported by the NIC.

   - Add __bpf_kfunc tag for marking kernel functions as kfuncs.

   - Add cgroup.memory=nobpf kernel parameter option to disable BPF
     memory accounting for container environments.

  Netfilter:

   - Remove the CLUSTERIP target. It has been marked as obsolete for
     years, and we still have WARN splats wrt races of the out-of-band
     /proc interface installed by this target.

   - Add 'destroy' commands to nf_tables. They are identical to the
     existing 'delete' commands, but do not return an error if the
     referenced object (set, chain, rule...) did not exist.

  Driver API:

   - Improve cpumask_local_spread() locality to help NICs set the right
     IRQ affinity on AMD platforms.

   - Separate C22 and C45 MDIO bus transactions more clearly.

   - Introduce new DCB table to control DSCP rewrite on egress.

   - Support configuration of Physical Layer Collision Avoidance (PLCA)
     Reconciliation Sublayer (RS) (802.3cg-2019). Modern version of
     shared medium Ethernet.

   - Support for MAC Merge layer (IEEE 802.3-2018 clause 99). Allowing
     preemption of low priority frames by high priority frames.

   - Add support for controlling MACSec offload using netlink SET.

   - Rework devlink instance refcounts to allow registration and
     de-registration under the instance lock. Split the code into
     multiple files, drop some of the unnecessarily granular locks and
     factor out common parts of netlink operation handling.

   - Add TX frame aggregation parameters (for USB drivers).

   - Add a new attr TCA_EXT_WARN_MSG to report TC (offload) warning
     messages with notifications for debug.

   - Allow offloading of UDP NEW connections via act_ct.

   - Add support for per action HW stats in TC.

   - Support hardware miss to TC action (continue processing in SW from
     a specific point in the action chain).

   - Warn if old Wireless Extension user space interface is used with
     modern cfg80211/mac80211 drivers. Do not support Wireless
     Extensions for Wi-Fi 7 devices at all. Everyone should switch to
     using nl80211 interface instead.

   - Improve the CAN bit timing configuration. Use extack to return
     error messages directly to user space, update the SJW handling,
     including the definition of a new default value that will benefit
     CAN-FD controllers, by increasing their oscillator tolerance.

  New hardware / drivers:

   - Ethernet:
      - nVidia BlueField-3 support (control traffic driver)
      - Ethernet support for imx93 SoCs
      - Motorcomm yt8531 gigabit Ethernet PHY
      - onsemi NCN26000 10BASE-T1S PHY (with support for PLCA)
      - Microchip LAN8841 PHY (incl. cable diagnostics and PTP)
      - Amlogic gxl MDIO mux

   - WiFi:
      - RealTek RTL8188EU (rtl8xxxu)
      - Qualcomm Wi-Fi 7 devices (ath12k)

   - CAN:
      - Renesas R-Car V4H

  Drivers:

   - Bluetooth:
      - Set Per Platform Antenna Gain (PPAG) for Intel controllers.

   - Ethernet NICs:
      - Intel (1G, igc):
         - support TSN / Qbv / packet scheduling features of i226 model
      - Intel (100G, ice):
         - use GNSS subsystem instead of TTY
         - multi-buffer XDP support
         - extend support for GPIO pins to E823 devices
      - nVidia/Mellanox:
         - update the shared buffer configuration on PFC commands
         - implement PTP adjphase function for HW offset control
         - TC support for Geneve and GRE with VF tunnel offload
         - more efficient crypto key management method
         - multi-port eswitch support
      - Netronome/Corigine:
         - add DCB IEEE support
         - support IPsec offloading for NFP3800
      - Freescale/NXP (enetc):
         - support XDP_REDIRECT for XDP non-linear buffers
         - improve reconfig, avoid link flap and waiting for idle
         - support MAC Merge layer
      - Other NICs:
         - sfc/ef100: add basic devlink support for ef100
         - ionic: rx_push mode operation (writing descriptors via MMIO)
         - bnxt: use the auxiliary bus abstraction for RDMA
         - r8169: disable ASPM and reset bus in case of tx timeout
         - cpsw: support QSGMII mode for J721e CPSW9G
         - cpts: support pulse-per-second output
         - ngbe: add an mdio bus driver
         - usbnet: optimize usbnet_bh() by avoiding unnecessary queuing
         - r8152: handle devices with FW with NCM support
         - amd-xgbe: support 10Mbps, 2.5GbE speeds and rx-adaptation
         - virtio-net: support multi buffer XDP
         - virtio/vsock: replace virtio_vsock_pkt with sk_buff
         - tsnep: XDP support

   - Ethernet high-speed switches:
      - nVidia/Mellanox (mlxsw):
         - add support for latency TLV (in FW control messages)
      - Microchip (sparx5):
         - separate explicit and implicit traffic forwarding rules, make
           the implicit rules always active
         - add support for egress DSCP rewrite
         - IS0 VCAP support (Ingress Classification)
         - IS2 VCAP filters (protos, L3 addrs, L4 ports, flags, ToS
           etc.)
         - ES2 VCAP support (Egress Access Control)
         - support for Per-Stream Filtering and Policing (802.1Q,
           8.6.5.1)

   - Ethernet embedded switches:
      - Marvell (mv88e6xxx):
         - add MAB (port auth) offload support
         - enable PTP receive for mv88e6390
      - NXP (ocelot):
         - support MAC Merge layer
         - support for the the vsc7512 internal copper phys
      - Microchip:
         - lan9303: convert to PHYLINK
         - lan966x: support TC flower filter statistics
         - lan937x: PTP support for KSZ9563/KSZ8563 and LAN937x
         - lan937x: support Credit Based Shaper configuration
         - ksz9477: support Energy Efficient Ethernet
      - other:
         - qca8k: convert to regmap read/write API, use bulk operations
         - rswitch: Improve TX timestamp accuracy

   - Intel WiFi (iwlwifi):
      - EHT (Wi-Fi 7) rate reporting
      - STEP equalizer support: transfer some STEP (connection to radio
        on platforms with integrated wifi) related parameters from the
        BIOS to the firmware.

   - Qualcomm 802.11ax WiFi (ath11k):
      - IPQ5018 support
      - Fine Timing Measurement (FTM) responder role support
      - channel 177 support

   - MediaTek WiFi (mt76):
      - per-PHY LED support
      - mt7996: EHT (Wi-Fi 7) support
      - Wireless Ethernet Dispatch (WED) reset support
      - switch to using page pool allocator

   - RealTek WiFi (rtw89):
      - support new version of Bluetooth co-existance

   - Mobile:
      - rmnet: support TX aggregation"

* tag 'net-next-6.3' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-next: (1872 commits)
  page_pool: add a comment explaining the fragment counter usage
  net: ethtool: fix __ethtool_dev_mm_supported() implementation
  ethtool: pse-pd: Fix double word in comments
  xsk: add linux/vmalloc.h to xsk.c
  sefltests: netdevsim: wait for devlink instance after netns removal
  selftest: fib_tests: Always cleanup before exit
  net/mlx5e: Align IPsec ASO result memory to be as required by hardware
  net/mlx5e: TC, Set CT miss to the specific ct action instance
  net/mlx5e: Rename CHAIN_TO_REG to MAPPED_OBJ_TO_REG
  net/mlx5: Refactor tc miss handling to a single function
  net/mlx5: Kconfig: Make tc offload depend on tc skb extension
  net/sched: flower: Support hardware miss to tc action
  net/sched: flower: Move filter handle initialization earlier
  net/sched: cls_api: Support hardware miss to tc action
  net/sched: Rename user cookie and act cookie
  sfc: fix builds without CONFIG_RTC_LIB
  sfc: clean up some inconsistent indentings
  net/mlx4_en: Introduce flexible array to silence overflow warning
  net: lan966x: Fix possible deadlock inside PTP
  net/ulp: Remove redundant ->clone() test in inet_clone_ulp().
  ...

1 files changed, 441 insertions, 0 deletions

diff --git a/Documentation/fb/matroxfb.rst b/Documentation/fb/matroxfb.rst
new file mode 100644
index 000000000..6158c49c8
--- /dev/null
+++ b/Documentation/fb/matroxfb.rst
@@ -0,0 +1,441 @@
+=================
+What is matroxfb?
+=================
+
+.. [This file is cloned from VesaFB. Thanks go to Gerd Knorr]
+
+
+This is a driver for a graphic framebuffer for Matrox devices on
+Alpha, Intel and PPC boxes.
+
+Advantages:
+
+ * It provides a nice large console (128 cols + 48 lines with 1024x768)
+   without using tiny, unreadable fonts.
+ * You can run XF{68,86}_FBDev or XFree86 fbdev driver on top of /dev/fb0
+ * Most important: boot logo :-)
+
+Disadvantages:
+
+ * graphic mode is slower than text mode... but you should not notice
+   if you use same resolution as you used in textmode.
+
+
+How to use it?
+==============
+
+Switching modes is done using the video=matroxfb:vesa:... boot parameter
+or using `fbset` program.
+
+If you want, for example, enable a resolution of 1280x1024x24bpp you should
+pass to the kernel this command line: "video=matroxfb:vesa:0x1BB".
+
+You should compile in both vgacon (to boot if you remove you Matrox from
+box) and matroxfb (for graphics mode). You should not compile-in vesafb
+unless you have primary display on non-Matrox VBE2.0 device (see
+Documentation/fb/vesafb.rst for details).
+
+Currently supported video modes are (through vesa:... interface, PowerMac
+has [as addon] compatibility code):
+
+
+Graphic modes
+-------------
+
+===  =======  =======  =======  =======  =======
+bpp  640x400  640x480  768x576  800x600  960x720
+===  =======  =======  =======  =======  =======
+  4             0x12             0x102
+  8   0x100    0x101    0x180    0x103    0x188
+ 15            0x110    0x181    0x113    0x189
+ 16            0x111    0x182    0x114    0x18A
+ 24            0x1B2    0x184    0x1B5    0x18C
+ 32            0x112    0x183    0x115    0x18B
+===  =======  =======  =======  =======  =======
+
+
+Graphic modes (continued)
+-------------------------
+
+===  ======== ======== ========= ========= =========
+bpp  1024x768 1152x864 1280x1024 1408x1056 1600x1200
+===  ======== ======== ========= ========= =========
+  4    0x104             0x106
+  8    0x105    0x190    0x107     0x198     0x11C
+ 15    0x116    0x191    0x119     0x199     0x11D
+ 16    0x117    0x192    0x11A     0x19A     0x11E
+ 24    0x1B8    0x194    0x1BB     0x19C     0x1BF
+ 32    0x118    0x193    0x11B     0x19B
+===  ======== ======== ========= ========= =========
+
+
+Text modes
+----------
+
+==== =======  =======  ========  ========  ========
+text 640x400  640x480  1056x344  1056x400  1056x480
+==== =======  =======  ========  ========  ========
+ 8x8   0x1C0    0x108     0x10A     0x10B     0x10C
+8x16 2, 3, 7                        0x109
+==== =======  =======  ========  ========  ========
+
+You can enter these number either hexadecimal (leading `0x`) or decimal
+(0x100 = 256). You can also use value + 512 to achieve compatibility
+with your old number passed to vesafb.
+
+Non-listed number can be achieved by more complicated command-line, for
+example 1600x1200x32bpp can be specified by `video=matroxfb:vesa:0x11C,depth:32`.
+
+
+X11
+===
+
+XF{68,86}_FBDev should work just fine, but it is non-accelerated. On non-intel
+architectures there are some glitches for 24bpp videomodes. 8, 16 and 32bpp
+works fine.
+
+Running another (accelerated) X-Server like XF86_SVGA works too. But (at least)
+XFree servers have big troubles in multihead configurations (even on first
+head, not even talking about second). Running XFree86 4.x accelerated mga
+driver is possible, but you must not enable DRI - if you do, resolution and
+color depth of your X desktop must match resolution and color depths of your
+virtual consoles, otherwise X will corrupt accelerator settings.
+
+
+SVGALib
+=======
+
+Driver contains SVGALib compatibility code. It is turned on by choosing textual
+mode for console. You can do it at boot time by using videomode
+2,3,7,0x108-0x10C or 0x1C0. At runtime, `fbset -depth 0` does this work.
+Unfortunately, after SVGALib application exits, screen contents is corrupted.
+Switching to another console and back fixes it. I hope that it is SVGALib's
+problem and not mine, but I'm not sure.
+
+
+Configuration
+=============
+
+You can pass kernel command line options to matroxfb with
+`video=matroxfb:option1,option2:value2,option3` (multiple options should be
+separated by comma, values are separated from options by `:`).
+Accepted options:
+
+============ ===================================================================
+mem:X        size of memory (X can be in megabytes, kilobytes or bytes)
+	     You can only decrease value determined by driver because of
+	     it always probe for memory. Default is to use whole detected
+	     memory usable for on-screen display (i.e. max. 8 MB).
+disabled     do not load driver; you can use also `off`, but `disabled`
+	     is here too.
+enabled      load driver, if you have `video=matroxfb:disabled` in LILO
+	     configuration, you can override it by this (you cannot override
+	     `off`). It is default.
+noaccel      do not use acceleration engine. It does not work on Alphas.
+accel        use acceleration engine. It is default.
+nopan        create initial consoles with vyres = yres, thus disabling virtual
+	     scrolling.
+pan          create initial consoles as tall as possible (vyres = memory/vxres).
+	     It is default.
+nopciretry   disable PCI retries. It is needed for some broken chipsets,
+	     it is autodetected for intel's 82437. In this case device does
+	     not comply to PCI 2.1 specs (it will not guarantee that every
+	     transaction terminate with success or retry in 32 PCLK).
+pciretry     enable PCI retries. It is default, except for intel's 82437.
+novga        disables VGA I/O ports. It is default if BIOS did not enable
+	     device. You should not use this option, some boards then do not
+	     restart without power off.
+vga          preserve state of VGA I/O ports. It is default. Driver does not
+	     enable VGA I/O if BIOS did not it (it is not safe to enable it in
+	     most cases).
+nobios       disables BIOS ROM. It is default if BIOS did not enable BIOS
+	     itself. You should not use this option, some boards then do not
+	     restart without power off.
+bios         preserve state of BIOS ROM. It is default. Driver does not enable
+	     BIOS if BIOS was not enabled before.
+noinit       tells driver, that devices were already initialized. You should use
+	     it if you have G100 and/or if driver cannot detect memory, you see
+	     strange pattern on screen and so on. Devices not enabled by BIOS
+	     are still initialized. It is default.
+init         driver initializes every device it knows about.
+memtype      specifies memory type, implies 'init'. This is valid only for G200
+	     and G400 and has following meaning:
+
+	       G200:
+		 -  0 -> 2x128Kx32 chips, 2MB onboard, probably sgram
+		 -  1 -> 2x128Kx32 chips, 4MB onboard, probably sgram
+		 -  2 -> 2x256Kx32 chips, 4MB onboard, probably sgram
+		 -  3 -> 2x256Kx32 chips, 8MB onboard, probably sgram
+		 -  4 -> 2x512Kx16 chips, 8/16MB onboard, probably sdram only
+		 -  5 -> same as above
+		 -  6 -> 4x128Kx32 chips, 4MB onboard, probably sgram
+		 -  7 -> 4x128Kx32 chips, 8MB onboard, probably sgram
+	       G400:
+		 -  0 -> 2x512Kx16 SDRAM, 16/32MB
+		 -	 2x512Kx32 SGRAM, 16/32MB
+		 -  1 -> 2x256Kx32 SGRAM, 8/16MB
+		 -  2 -> 4x128Kx32 SGRAM, 8/16MB
+		 -  3 -> 4x512Kx32 SDRAM, 32MB
+		 -  4 -> 4x256Kx32 SGRAM, 16/32MB
+		 -  5 -> 2x1Mx32 SDRAM, 32MB
+		 -  6 -> reserved
+		 -  7 -> reserved
+
+	     You should use sdram or sgram parameter in addition to memtype
+	     parameter.
+nomtrr       disables write combining on frame buffer. This slows down driver
+	     but there is reported minor incompatibility between GUS DMA and
+	     XFree under high loads if write combining is enabled (sound
+	     dropouts).
+mtrr         enables write combining on frame buffer. It speeds up video
+	     accesses much. It is default. You must have MTRR support enabled
+	     in kernel and your CPU must have MTRR (f.e. Pentium II have them).
+sgram        tells to driver that you have Gxx0 with SGRAM memory. It has no
+	     effect without `init`.
+sdram        tells to driver that you have Gxx0 with SDRAM memory.
+	     It is a default.
+inv24        change timings parameters for 24bpp modes on Millennium and
+	     Millennium II. Specify this if you see strange color shadows
+	     around  characters.
+noinv24      use standard timings. It is the default.
+inverse      invert colors on screen (for LCD displays)
+noinverse    show true colors on screen. It is default.
+dev:X        bind driver to device X. Driver numbers device from 0 up to N,
+	     where device 0 is first `known` device found, 1 second and so on.
+	     lspci lists devices in this order.
+	     Default is `every` known device.
+nohwcursor   disables hardware cursor (use software cursor instead).
+hwcursor     enables hardware cursor. It is default. If you are using
+	     non-accelerated mode (`noaccel` or `fbset -accel false`), software
+	     cursor is used (except for text mode).
+noblink      disables cursor blinking. Cursor in text mode always blinks (hw
+	     limitation).
+blink        enables cursor blinking. It is default.
+nofastfont   disables fastfont feature. It is default.
+fastfont:X   enables fastfont feature. X specifies size of memory reserved for
+	     font data, it must be >= (fontwidth*fontheight*chars_in_font)/8.
+	     It is faster on Gx00 series, but slower on older cards.
+grayscale    enable grayscale summing. It works in PSEUDOCOLOR modes (text,
+	     4bpp, 8bpp). In DIRECTCOLOR modes it is limited to characters
+	     displayed through putc/putcs. Direct accesses to framebuffer
+	     can paint colors.
+nograyscale  disable grayscale summing. It is default.
+cross4MB     enables that pixel line can cross 4MB boundary. It is default for
+	     non-Millennium.
+nocross4MB   pixel line must not cross 4MB boundary. It is default for
+	     Millennium I or II, because of these devices have hardware
+	     limitations which do not allow this. But this option is
+	     incompatible with some (if not all yet released) versions of
+	     XF86_FBDev.
+dfp          enables digital flat panel interface. This option is incompatible
+	     with secondary (TV) output - if DFP is active, TV output must be
+	     inactive and vice versa. DFP always uses same timing as primary
+	     (monitor) output.
+dfp:X        use settings X for digital flat panel interface. X is number from
+	     0 to 0xFF, and meaning of each individual bit is described in
+	     G400 manual, in description of DAC register 0x1F. For normal
+	     operation you should set all bits to zero, except lowest bit. This
+	     lowest bit selects who is source of display clocks, whether G400,
+	     or panel. Default value is now read back from hardware - so you
+	     should specify this value only if you are also using `init`
+	     parameter.
+outputs:XYZ  set mapping between CRTC and outputs. Each letter can have value
+	     of 0 (for no CRTC), 1 (CRTC1) or 2 (CRTC2), and first letter
+	     corresponds to primary analog output, second letter to the
+	     secondary analog output and third letter to the DVI output.
+	     Default setting is 100 for cards below G400 or G400 without DFP,
+	     101 for G400 with DFP, and 111 for G450 and G550. You can set
+	     mapping only on first card, use matroxset for setting up other
+	     devices.
+vesa:X       selects startup videomode. X is number from 0 to 0x1FF, see table
+	     above for detailed explanation. Default is 640x480x8bpp if driver
+	     has 8bpp support. Otherwise first available of 640x350x4bpp,
+	     640x480x15bpp, 640x480x24bpp, 640x480x32bpp or 80x25 text
+	     (80x25 text is always available).
+============ ===================================================================
+
+If you are not satisfied with videomode selected by `vesa` option, you
+can modify it with these options:
+
+============ ===================================================================
+xres:X       horizontal resolution, in pixels. Default is derived from `vesa`
+	     option.
+yres:X       vertical resolution, in pixel lines. Default is derived from `vesa`
+	     option.
+upper:X      top boundary: lines between end of VSYNC pulse and start of first
+	     pixel line of picture. Default is derived from `vesa` option.
+lower:X      bottom boundary: lines between end of picture and start of VSYNC
+	     pulse. Default is derived from `vesa` option.
+vslen:X      length of VSYNC pulse, in lines. Default is derived from `vesa`
+	     option.
+left:X       left boundary: pixels between end of HSYNC pulse and first pixel.
+	     Default is derived from `vesa` option.
+right:X      right boundary: pixels between end of picture and start of HSYNC
+	     pulse. Default is derived from `vesa` option.
+hslen:X      length of HSYNC pulse, in pixels. Default is derived from `vesa`
+	     option.
+pixclock:X   dotclocks, in ps (picoseconds). Default is derived from `vesa`
+	     option and from `fh` and `fv` options.
+sync:X       sync. pulse - bit 0 inverts HSYNC polarity, bit 1 VSYNC polarity.
+	     If bit 3 (value 0x08) is set, composite sync instead of HSYNC is
+	     generated. If bit 5 (value 0x20) is set, sync on green is turned
+	     on. Do not forget that if you want sync on green, you also probably
+	     want composite sync.
+	     Default depends on `vesa`.
+depth:X      Bits per pixel: 0=text, 4,8,15,16,24 or 32. Default depends on
+	     `vesa`.
+============ ===================================================================
+
+If you know capabilities of your monitor, you can specify some (or all) of
+`maxclk`, `fh` and `fv`. In this case, `pixclock` is computed so that
+pixclock <= maxclk, real_fh <= fh and real_fv <= fv.
+
+============ ==================================================================
+maxclk:X     maximum dotclock. X can be specified in MHz, kHz or Hz. Default is
+	     `don`t care`.
+fh:X         maximum horizontal synchronization frequency. X can be specified
+	     in kHz or Hz. Default is `don't care`.
+fv:X         maximum vertical frequency. X must be specified in Hz. Default is
+	     70 for modes derived from `vesa` with yres <= 400, 60Hz for
+	     yres > 400.
+============ ==================================================================
+
+
+Limitations
+===========
+
+There are known and unknown bugs, features and misfeatures.
+Currently there are following known bugs:
+
+ - SVGALib does not restore screen on exit
+ - generic fbcon-cfbX procedures do not work on Alphas. Due to this,
+   `noaccel` (and cfb4 accel) driver does not work on Alpha. So everyone
+   with access to `/dev/fb*` on Alpha can hang machine (you should restrict
+   access to `/dev/fb*` - everyone with access to this device can destroy
+   your monitor, believe me...).
+ - 24bpp does not support correctly XF-FBDev on big-endian architectures.
+ - interlaced text mode is not supported; it looks like hardware limitation,
+   but I'm not sure.
+ - Gxx0 SGRAM/SDRAM is not autodetected.
+ - maybe more...
+
+And following misfeatures:
+
+ - SVGALib does not restore screen on exit.
+ - pixclock for text modes is limited by hardware to
+
+    - 83 MHz on G200
+    - 66 MHz on Millennium I
+    - 60 MHz on Millennium II
+
+   Because I have no access to other devices, I do not know specific
+   frequencies for them. So driver does not check this and allows you to
+   set frequency higher that this. It causes sparks, black holes and other
+   pretty effects on screen. Device was not destroyed during tests. :-)
+ - my Millennium G200 oscillator has frequency range from 35 MHz to 380 MHz
+   (and it works with 8bpp on about 320 MHz dotclocks (and changed mclk)).
+   But Matrox says on product sheet that VCO limit is 50-250 MHz, so I believe
+   them (maybe that chip overheats, but it has a very big cooler (G100 has
+   none), so it should work).
+ - special mixed video/graphics videomodes of Mystique and Gx00 - 2G8V16 and
+   G16V16 are not supported
+ - color keying is not supported
+ - feature connector of Mystique and Gx00 is set to VGA mode (it is disabled
+   by BIOS)
+ - DDC (monitor detection) is supported through dualhead driver
+ - some check for input values are not so strict how it should be (you can
+   specify vslen=4000 and so on).
+ - maybe more...
+
+And following features:
+
+ - 4bpp is available only on Millennium I and Millennium II. It is hardware
+   limitation.
+ - selection between 1:5:5:5 and 5:6:5 16bpp videomode is done by -rgba
+   option of fbset: "fbset -depth 16 -rgba 5,5,5" selects 1:5:5:5, anything
+   else selects 5:6:5 mode.
+ - text mode uses 6 bit VGA palette instead of 8 bit (one of 262144 colors
+   instead of one of 16M colors). It is due to hardware limitation of
+   Millennium I/II and SVGALib compatibility.
+
+
+Benchmarks
+==========
+It is time to redraw whole screen 1000 times in 1024x768, 60Hz. It is
+time for draw 6144000 characters on screen through /dev/vcsa
+(for 32bpp it is about 3GB of data (exactly 3000 MB); for 8x16 font in
+16 seconds, i.e. 187 MBps).
+Times were obtained from one older version of driver, now they are about 3%
+faster, it is kernel-space only time on P-II/350 MHz, Millennium I in 33 MHz
+PCI slot, G200 in AGP 2x slot. I did not test vgacon::
+
+  NOACCEL
+	8x16                 12x22
+	Millennium I  G200   Millennium I  G200
+  8bpp    16.42         9.54   12.33         9.13
+  16bpp   21.00        15.70   19.11        15.02
+  24bpp   36.66        36.66   35.00        35.00
+  32bpp   35.00        30.00   33.85        28.66
+
+  ACCEL, nofastfont
+	8x16                 12x22                6x11
+	Millennium I  G200   Millennium I  G200   Millennium I  G200
+  8bpp     7.79         7.24   13.55         7.78   30.00        21.01
+  16bpp    9.13         7.78   16.16         7.78   30.00        21.01
+  24bpp   14.17        10.72   18.69        10.24   34.99        21.01
+  32bpp   16.15	     16.16   18.73        13.09   34.99        21.01
+
+  ACCEL, fastfont
+	8x16                 12x22                6x11
+	Millennium I  G200   Millennium I  G200   Millennium I  G200
+  8bpp     8.41         6.01    6.54         4.37   16.00        10.51
+  16bpp    9.54         9.12    8.76         6.17   17.52        14.01
+  24bpp   15.00        12.36   11.67        10.00   22.01        18.32
+  32bpp   16.18        18.29*  12.71        12.74   24.44        21.00
+
+  TEXT
+	8x16
+	Millennium I  G200
+  TEXT     3.29         1.50
+
+  * Yes, it is slower than Millennium I.
+
+
+Dualhead G400
+=============
+Driver supports dualhead G400 with some limitations:
+ + secondary head shares videomemory with primary head. It is not problem
+   if you have 32MB of videoram, but if you have only 16MB, you may have
+   to think twice before choosing videomode (for example twice 1880x1440x32bpp
+   is not possible).
+ + due to hardware limitation, secondary head can use only 16 and 32bpp
+   videomodes.
+ + secondary head is not accelerated. There were bad problems with accelerated
+   XFree when secondary head used to use acceleration.
+ + secondary head always powerups in 640x480@60-32 videomode. You have to use
+   fbset to change this mode.
+ + secondary head always powerups in monitor mode. You have to use fbmatroxset
+   to change it to TV mode. Also, you must select at least 525 lines for
+   NTSC output and 625 lines for PAL output.
+ + kernel is not fully multihead ready. So some things are impossible to do.
+ + if you compiled it as module, you must insert i2c-matroxfb, matroxfb_maven
+   and matroxfb_crtc2 into kernel.
+
+
+Dualhead G450
+=============
+Driver supports dualhead G450 with some limitations:
+ + secondary head shares videomemory with primary head. It is not problem
+   if you have 32MB of videoram, but if you have only 16MB, you may have
+   to think twice before choosing videomode.
+ + due to hardware limitation, secondary head can use only 16 and 32bpp
+   videomodes.
+ + secondary head is not accelerated.
+ + secondary head always powerups in 640x480@60-32 videomode. You have to use
+   fbset to change this mode.
+ + TV output is not supported
+ + kernel is not fully multihead ready, so some things are impossible to do.
+ + if you compiled it as module, you must insert matroxfb_g450 and matroxfb_crtc2
+   into kernel.
+
+Petr Vandrovec <vandrove@vc.cvut.cz>