summaryrefslogtreecommitdiff
path: root/sys
AgeCommit message (Collapse)Author
2010-02-19Make this truly MI:Miod Vallat
- remove (unused) MSR defines and x86 specific <machine/cpufunc.h> include - default sc_dmamap_flags to BUS_DMA_COHERENT
2010-02-19Move auglx(4) from i386-only land to MI land.Miod Vallat
(auglx.c is not modified in this commit yet, only moved around)
2010-02-19Add a datasheet URL for reference.Miod Vallat
2010-02-19even more USB; ok miod@Otto Moerbeek
2010-02-19do NOT install the loongson mbr on other architectures!Theo de Raadt
2010-02-19Add a driver for the M41T8x time-of-day clock chips, such as the M41T83Miod Vallat
found on the Gdium Liberty. These machines finally no longer complain about a missing todclock.
2010-02-18do not zero fields which are already zero; from cnstTheo de Raadt
2010-02-18Add a master driver for the VoyagerGX SM502 chip found on the Gdium, which isMiod Vallat
not only a framebuffer. Allow smfb to attach either at pci or at voyager. Add gpio@voyager and gdiumiic@gpio, a derivative of the MI gpioiic driver. This allows us to get lmtemp@iic on the Lemote; the next step being to write a driver for the i2c todclock chip found on the same bus.
2010-02-18improve startup messages rewording them in a more usual way; as aIgor Sobrado
side effect, this change makes displayed messages shorter. miod@ has observed that `Horse' is the codename of the chip, which gets paired with a `Saddle' companion chip. there is one hose (i.e., bus) per chip; others (`Tsunami') have two hoses per chip, or even four (`Typhoon'); so, `Horse' is not really a typo. written with clever advice from krw@ and miod@ miod@ has suggested the capitalization for `Saddle'. ok krw@, miod@
2010-02-18The PMON code is compiled as 64 bit code but still using the old ABI (actually,Miod Vallat
a 64-bit version of the old ABI), which explains why these wrappers are needed. Update comments to match reality, and make cpp write all the code of this file. (``survival of the laziest'' motto) Keep the misleading file name though (lazyness again)
2010-02-18fix a few capitalizations of QLogic.Igor Sobrado
2010-02-18improve startup messages displayed by isp(4) in such a way entriesIgor Sobrado
printed by this driver in dmesg are more standard (and short); while here, fix the capitalization of QLogic. ok krw@
2010-02-17Make sure we claim the interrupt as ours if the time out bit is set in theMark Kettenis
interrupt status register. Fixes a problem with spurious interrupts on the fuloong. ok miod@
2010-02-17Forgot to bump version number in comments; from Seth Wright via jmc@, thanks.Miod Vallat
2010-02-17If PMON has loaded an initrd binary, and this binary looks like a validMiod Vallat
ELF image, assume it's the kernel and try to boot it immediately. This allows a Gdium system with both the bootloader and the kernel image on an ext2fs partition, with `al' pointing to the bootblocks and `rd' pointing to the kernel in PMON environment, to boot a kernel with proper kernel symbols, for the first time. (please don't get me started on how reliable `load -k' is on the Gdium) Bump bootblocks version to 0.2.
2010-02-17descend into mbr; ok miod@Otto Moerbeek
2010-02-17proto mbr for loongsonOtto Moerbeek
2010-02-17Split boot/Makefile in case we need to build several boot.${VENDOR} compiledMiod Vallat
from the same sources with different CFLAGS; the need for this has not arisen yet but I fear it will hit the fan soon enough; no functional change in the meantime.
2010-02-17Do not always create a new node when a beacon or a probe responseDamien Bergamini
is received. The creation of a new node is already handled in ieee80211_recv_probe_resp() when necessary. This avoids creating empty nodes with wrong channels when beacons are received on the wrong channel (overlapping channels). Those empty nodes may also prevent the real node from being discovered because of ieee80211_find_node_for_beacon() filtering. This should prevent entries of the form: nwid "" chan 3 bssid 00:01:02:03:04:05 0dB 54M in "ifconfig if0 scan" output, like reported by Rivo Nurges.
2010-02-17read calibration version from ROM and set IWN_GP_DRIVER_CALIB_VER6Damien Bergamini
bit on 6x50 if version>=6.
2010-02-17Build and install the boot blocks, and put it as part of the snapshots files.Miod Vallat
2010-02-17- prevent "anti-memory" from showing up, this makes systems with 2GB ofJasper Lievisse Adriaanse
RAM work. help and ok miod@
2010-02-17KNFPaul Irofti
2010-02-17change constant of convert formulra of SHT1X sensor to fitYojiro Uo
TEMPerHUM hardware. (It uses 3.38-3.48V as SHT1X's VDD). mentiond by Michael Fuckner
2010-02-17dont defer broadcast or multicast packets.David Gwynne
ok sthen@ henning@
2010-02-16Move part of the device_register() logic into platform-specific code. AddMiod Vallat
correct root device identification for Gdium (assuming it is booted from the G-Key), although it will not attach as sd0 if there is anything connected to the usb port next to the Ethernet port.
2010-02-16Use printf() in there, not pmon_printf().Miod Vallat
2010-02-16Display the system type on the mainbus0 line as well.Miod Vallat
2010-02-16Actually, both the Gdium and the Lemote PMON reset the USB controllersMiod Vallat
before transferring control to any binary, so trying to boot anything on the Gdium without bringing half the kernel code is a no-go. So, only keep `wd' as a possible boot device, for the Lemote systems, and complain loudly if run on a Gdium.
2010-02-16fix AR_DTIM_PERIODDamien Bergamini
2010-02-16Recognize the BCM57780 PHY, tested by bruce on tech@.Stuart Henderson
From Brad, ok jsg@ kettenis@
2010-02-16syncStuart Henderson
2010-02-16Add BCM57780, from Brad. ok jsg@ kettenis@Stuart Henderson
2010-02-16add some defines needed for the upcoming newfs_ext2fs; ok miod@Otto Moerbeek
2010-02-16Use i2c_addr_t as the type for the init_xfer function because thisMichael Knudsen
is what the function pointer type uses. ok grange
2010-02-15Simplify MBR and install it as /usr/mdec/mbr instead of /usr/mdec/mbr.miniMark Kettenis
since it makes a fairly decent template for fdisk -i. Change the procedure that makes the rb600 installation miniroot image to use fdisk -i instead of dd(1) to create the MBR for the image. Adjust the miniroot disk type in /etc/disktab to the slightly different layout that fdisk -i creates.
2010-02-15regenDamien Bergamini
2010-02-15add MSI (Micro Star International) vendor and RT3090 product idDamien Bergamini
2010-02-15Properly detect when the band changes from 2GHz to 5GHz andDamien Bergamini
vice-versa, and do a full reset of the chip when it happens. This does not affect the AR9280 since a full reset is always necessary when the channel changes. Should fix dual-band AR5416 devices (problem reported by Rivo Nurges).
2010-02-15- handle SIOCS80211CHANNEL in monitor mode without doing a full resetDamien Bergamini
- call athn_set_multi on SIOCADDMULTI/SIOCDELMULTI
2010-02-15Since strip doesn't know how to strip a U-boot image, don't attempt toMark Kettenis
do so during install.
2010-02-15Descend into mbr subdirectory.Mark Kettenis
2010-02-15MBR for rb600 installation miniroot image.Mark Kettenis
2010-02-14Work in progress standalone boot blocks for OpenBSD/loongson systems. WillMiod Vallat
load a kernel from the OpenBSD partition, with symbols. However: - does not work on gdium (usb i/o errors) - no countdown due to PMON limitations, so either `bsd' in the PMON environment contains a kernel filename to load, and it will get loaded immediately, or it doesn't (e.g. the variable does no exist) and you get a prompt and can pass options to the kernel. Note that /etc/boot.conf commands can still be used in both cases. - does not work for me when loaded automatically from PMON (put in the `al' environment variable). However, `boot $al' works fine. Need to investigate.
2010-02-14Bump version number (to 1.0).Mark Kettenis
2010-02-14Enable loading kernels compressed by gzip(1).Mark Kettenis
2010-02-14The ASIC does not like PCI bursts on registers because of a silicon bug.Damien Bergamini
To prevent PCI read or write bursts, we issue a read to a non-contiguous register before accessing a register. This problem does not show up on the x86 architecture since the memory model makes it almost impossible to generate PCI bursts without doing DMA. This bug is "documented" in the Ralink Linux driver as: "//Patch for ASIC turst read/write bug, needs to remove after metel fix" This fix makes the RT2561S chip on the Gdium (loongson) work. It fails to load the firmware without it. Tested by miod@ of the Gdium (slightly different diff though)
2010-02-14Don't use absolute paths for includes.Michael Knudsen
ok drahn
2010-02-13Force chunks into correct order once metadata is attached. This preventsJoel Sing
chunks from being reordered on an existing volume. ok marco@
2010-02-13Unbreak automatic rebuilds from hotspares.Joel Sing
ok marco@
generated by cgit v1.2.3 (git 2.25.1) at 2025-01-07 00:35:46 +0000