summaryrefslogtreecommitdiff
path: root/sys
AgeCommit message (Collapse)Author
2009-09-07Correct the PHY ID for IC+ IP1001. From Brad, confirmed with the datasheet.Stuart Henderson
2009-09-07Fix static-port, found by jmc@. ok henning@.Stuart Henderson
2009-09-06Don't pass uninitialized stack memory to setdisklabel() as the 'old'Kenneth R Westerback
label. Use the existing label, since we now check and reuse some values in it. And that's what all other drivers do. Bonus: we don't need that label on the stack anymore. ok deraadt@
2009-09-06Attach devices to obio(4) based on information from the flattened deviceMark Kettenis
tree. Since that information is also used for the PCI interrupts, devices in mini-PCI slots should work now as well, at least in the rb600.
2009-09-06Implement OF_getproplen().Mark Kettenis
2009-09-06Fix typo in comment, safe -> save.Marcus Glocker
Spotted by miod@
2009-09-06Fix udl_copycols overlay copy problem by copying area to off-screen firstMarcus Glocker
(same as in udl_copyrows).
2009-09-06Palm: cleanup kernel configuration filesMarek Vasut
2009-09-05Make use of the new wsdisplay ability, committed by miod previously, whichMarcus Glocker
allows our rasops functions to return EAGAIN if our usb command queue is full. This gets us rid of the ugly while { delay(); } loop and makes udl work also on MP kernels now (we faced a deadlock so far because spinning in that delay() loop with biglock active wasn't that much fun). With help from miod@
2009-09-05- don't try to shove a daddr64_t value into an int32_t.Jasper Lievisse Adriaanse
eventhough this won't really harm as ext2fs doesn't handle files that large, it was not correct. no objections from thib@
2009-09-05Buglet crept in.Miod Vallat
2009-09-05sync comment to reality, off-page page headers go intoThordur I. Bjornsson
an RB tree, not into a hashtable.
2009-09-05u_int32_t not u_int32_ts in debug code.Claudio Jeker
2009-09-05fix typosTheo de Raadt
2009-09-05Check the return value of all emulops in the emulation code, and abortMiod Vallat
tty output as soon as we hit a failure. Since the `output' of a character may cause several emulops to be called (e.g. if it causes scrollup or if this is the end of an escape sequence), all emulation code maintain a so-called `abort state', to be able to properly recover when the character is tentatively output later, and not reissue the emulops which did not fail the first time. With help from mglocker@
2009-09-05Make the output() wsemul_op return the number of characters processed andMiod Vallat
check it in wsdisplaystart() to suspend output if not all characters have been output; they will get reissued at the next tty rstrt_to timeout.
2009-09-05Change the wsdisplay_emulops return types from void to int; emulops will nowMiod Vallat
return zero on success and nonzero on failure. This commit only performs mechanical changes for the existing emulops to always return zero.
2009-09-05Rework internal interfaces in the wsdisplay emulation code to prepare forMiod Vallat
upcoming changes. No functional change.
2009-09-05scsi_done before COMPLETEDavid Gwynne
2009-09-05scsi_done before COMPLETE.David Gwynne
2009-09-05call scsi_done before returning COMPLETEDavid Gwynne
2009-09-05call scsi_done before returning COMPLETE in the stuffup case.David Gwynne
pointed out by miod
2009-09-05Fix various cases of stackgap_alloc() size arguments not being computedMiod Vallat
correctly, usually yielding the right value on 32 bit machines because sizeof int == sizeof pointer.
2009-09-05Make sure ehci_open() invokes ehci_device_setintr() at splusb; found withMiod Vallat
splassert.
2009-09-05Remove unnecessary assignments in sii311[24]_chip_map().Miod Vallat
2009-09-05Palm: add missing GPIO setup into palm_udc.cMarek Vasut
2009-09-05Palm: initial commit of sys/arch/palmMarek Vasut
OK deraadt@
2009-09-05Make DIOCSDINFO code consistant everywhere.Kenneth R Westerback
ok miod@ deraadt@
2009-09-04Add common framework for storing device listsJordan Hargrave
Useful for acpitz, acpipwrres, etc.
2009-09-04Bring back support for iCH10 based chips. This time support for fiber cardsDariusz Swiderski
seems to be fixed, thanks to tests done by mpf at mailq dot de. Also support for older fiber cards that have no PHY seems to be working, thanks to claudio@ The code includes all the changes that i backed out, plus two tweaks: 1. em_detect_gig_phy() gets called in em_setup_link() instead of em_copper_link_preconfig(), this enables phy detection on fiber cards. 2. em_detect_gig_phy() gets a condition to look for old fiber cards, that have no PHY. ok by claudio@, prodded by deraadt@
2009-09-04Make sure the MAC address is correct on big-endian platforms if it isn't readMark Kettenis
from the EEPROM. ok miod@
2009-09-04Missing scsi_done() in some error path returning COMPLETE; ok dlg@Miod Vallat
2009-09-03pxa2x0_mmc.c: allow passing detect GPIO from platformMarek Vasut
OK drahn@
2009-09-03Rearrange the interrupt handler so that it handlers "error status bits"Theo de Raadt
first instead of trying to push data transfers forward. Also, ensure that DONE interrupts are only un-masked when we are on the last sub-block, and that otherwise we process based on RD_REQ and WR_REQ interrupts coming in. Done with drahn and marex (starting from bizzare workarounds by marex which we could not explain) This appears to make Zaurus SD reads & writes completely reliable, even at very low cpu speeds. ok drahn marex
2009-09-03pxa2x0_mmc.c: Dont read MMC_STAT in intr_done if not neededMarek Vasut
It is not needed to read MMC_STAT if DPRINTF is empty. OK deraadt
2009-09-03Missing header.Joel Sing
2009-09-03The sparc64 boot block currently compares a memory address to the ELF magicJoel Sing
and complains if they do not match. Instead, load the start of the ELF header from memory and complain if this does not match the ELF magic. Tested by kettenis@ ok miod@
2009-09-03Check the return value of copyin().Joel Sing
ok marco@
2009-09-03Don't call VOP_CLOSE() with a null pointer, if the VOP_OPEN() call failed.Joel Sing
ok marco@
2009-09-03call scsi_done before returning COMPLETEDavid Gwynne
2009-09-03must call scsi_done before returning complete.David Gwynne
2009-09-03- fix two more format strings wrt daddr64_t valuesJasper Lievisse Adriaanse
"sure" deraadt@, ok otto@
2009-09-03crypto hardware (eg, hifn) establishes its interrupt handler atDavid Gwynne
IPL_NET. when the hardware finishes some work for the crypto subsystem and therefore something in the kernel that wanted crypto done, it calls crypto_done from that interrupt handler. one of the things that uses crypto is ipsec. when crypto is done for ipsec it then pushes the packet along the network stack. the problem is that all the structures inside the network stack are only protected at splsoftnet. we could be in the middle of modifications to the pf state table or the pfsync queues when we get a hifn interrupt and then go stomp on the same structures. the solution is to defer the completions so they can do the right spl protections. this basically reverts r1.46 of src/sys/crypto/crypto.c. found by naddy@
2009-09-02Attach devices to mainbus(4) based on information from the flattened deviceMark Kettenis
tree.
2009-09-02Add a minimal set of OpenFirmware compatibility interfaces.Mark Kettenis
ok dms@
2009-09-02check if we have access to the vnode before checking if we can write toFederico G. Schwindt
it. thib@ ok'd the idea and an earlier diff.
2009-09-02Backout the asyncio/aiod change, as it causes buf's to get hung.Thordur I. Bjornsson
problem noticed by deraadt@ ok beck@
2009-09-02Remove some out-of-date comments that are no longer present in the NetBSDMark Kettenis
tree, and s/NetBSD/OpenBSD/ in a debug printf.
2009-09-02Bring over some more glue from isp_pci.c; makes ISP2200 SBus FC cards actuallyMark Kettenis
work. Tested by deraadt@.
2009-09-02take advantage of workq_queue_task.David Gwynne