summaryrefslogtreecommitdiff
path: root/sys
AgeCommit message (Collapse)Author
2000-02-25import latest version from kame.Jun-ichiro itojun Hagino
- update scoped address syntax to the latest - I-D/RFC # cleanups
2000-02-25sync with realityJun-ichiro itojun Hagino
2000-02-25clear m->m_pkthdr when changing mbuf type to M_PKTHDR.Jun-ichiro itojun Hagino
without it, junk pointer in m->m_pkthdr.tdbi may bite us.
2000-02-24hide IP6_EXTHDR_CHECK from the userland.Jun-ichiro itojun Hagino
2000-02-23if we discover we are on a v8 cpu, and thus have the multiply and divideTheo de Raadt
instructions, replace the .{u,}{mul,div,rem} functions at runtime with (much smaller and faster) blocks using the actual hardware instructions.
2000-02-22blah, blah, blah, msgbufMichael Shalayeff
2000-02-22enlarge msgbuf, somewhat line netbsd didTheo de Raadt
2000-02-22put C accessable labels on {,u}{rem,div}Theo de Raadt
2000-02-22Regen.Aaron Campbell
2000-02-22Add some more Texas Instruments PCI-CardBus bridges; from NetBSD.Aaron Campbell
2000-02-22Use an _BSD_FOO_T_ type declaration for off_t in machine/ansi.h to allowTodd C. Miller
off_t to be visible in stdio.h even in strict ansi mode. This is needed for ftello()/fseeko() and it allows us to remove the horrible fpos_t hack.
2000-02-22Add an __attribute__((unused))__ to the function declared by _MCOUNT_DECL,Artur Grabowski
sparc defines that function as static and gcc doesn't understand that it gets called from an assembler stub.
2000-02-21TCP SACK fixes via Tom Henderson (tomh@cs.berkeley.edu):Niels Provos
- tcp_sack_adjust() was completely rewritten, since it was erroneously referencing receiver side sequence numbers and comparing with sender side sequence numbers (thanks to Arun Desai (adesai@cisco.com) who discovered the problem) - in tcp_output(), moved assignment of sendalot=0 to the piece of code immediately following the search for sack-eligible retransmissions (bug identified by Arun Desai). - tcp_input() was not clearing t_dupacks if fewer than three dupacks arrived between acks of new data. (bug identified by Gaurav Banga (gaurav@netapp.com))
2000-02-21move fpproc into the cpuinfo structure.Artur Grabowski
2000-02-21From FreeBSD:Jason Wright
correct usage of m_pullup for small frames bzero() contents of mbuf when manually padding the frame
2000-02-21From FreeBSD: Correct use of NULL/0Jason Wright
2000-02-21dead code and symbol pollution.Artur Grabowski
2000-02-21zap unused polluting variable.Artur Grabowski
2000-02-21a few comment fixes from espie@Michael Shalayeff
2000-02-21Rework bridge_filter() so that it does not require a full packet copyJason Wright
Allow for reusing the current mbuf (rather than a copy) for the last interface in the list in bridge_output(). Replace some pointer derefs with on the stack copies of the data.
2000-02-21separate assembly thing into an own file to avoid compiler-deps.Michael Shalayeff
ignore suspend reqs while suspending. ignore suspend requests for 3 seconds holdoff period once awaken. protect event queue manipulations w/ splhigh's.
2000-02-21The last pieces of hypersparc support.Artur Grabowski
- Split get_faultstatus into get_syncflt and get_asyncflt. - Get the syncflt status before setting up the trap frame and put the results in cpuinfo.syncfltdump (related to the next change). - unlock the sfsr/sfva after reading from ASI_SRMMUFP (it gets locked if the translation fails). - + other interface changes and cleanups. (most work from NetBSD).
2000-02-21In cpumatch_hypersparc - zap the instruction cache if we are the main cpu.Artur Grabowski
2000-02-21remove the annoying "hypersparc support still under construction" warning.Artur Grabowski
2000-02-21hypersparc has a writeback cacheArtur Grabowski
2000-02-21When mapping something into iommu space hypersparc requires us to align itArtur Grabowski
so that cache_alias_bits match in the kernel mapping and the iommu mapping. (see code for better explaination).
2000-02-21Take the hypersparc cache size from prom, enable instruction cache andArtur Grabowski
disable "Unimplemented Flush Traps". From NetBSD.
2000-02-21add SES/SAF-TE SCSI devicemjacob
2000-02-21add T_ENCLOSURE name and NOLUN Photon SENA devicesmjacob
2000-02-21Add in Enclosure device type (0xd/13). Pad out inquiry data to 50 bytesmjacob
os that SAF-TE devices can be found.
2000-02-21add char 124 to be SES devicemjacob
2000-02-21recycle unused char 24 to be SES devicemjacob
2000-02-21add SES device init/decl macrosmjacob
2000-02-20Add support for CDBs > 12 bytes.mjacob
2000-02-20Add 12160 (Ultra3) support. Add files and changes for target mode support.mjacob
Do some SNS fabric suppor tchanges. Roll revision levels. Tested on GENERIC i386 && sparc.
2000-02-20add Qlogic 12160 microcodemjacob
2000-02-20Add 12160 (Ultra3) support. Handle case of Extended CMD CDBs (cdbs withmjacob
more than 12 byte) for parallel SCSI.
2000-02-19flush the caches that need flushing on context switch.Artur Grabowski
2000-02-19replacemul has been broken all the time. comment it out for now.Artur Grabowski
2000-02-19get a symbol for pure_vcache_flushArtur Grabowski
2000-02-19Add two new cache operations.Artur Grabowski
cache_flush_all - (pretty obvious). pure_vcache_flush - flush all VIVT caches (needed on context switch). (From NetBSD).
2000-02-19use more modern typesTheo de Raadt
2000-02-19If we got an xfer from our backlog queue, don't allow sleeping; we're inTodd C. Miller
interrupt context. thorpej@netbsd.org
2000-02-19From NetBSD:Todd C. Miller
Check for SCSI_ADW_WDTR_DISABLE flag in tidmask not tid
2000-02-19From NetBSD:Todd C. Miller
o Use newer pci attach method o Set chip_type corectly Fixes panic with the 3?940UW boards
2000-02-19link boot at a starting address 0x10000 higher than bootxx; avoids ↵Theo de Raadt
hypersparc cache aliasing
2000-02-19set interface fields to null when detaching substructuresNiklas Hallqvist
in preparation for softc retainment.
2000-02-18No need to call memerr4_4c through a pointer. We know when we are a 4/4c.Artur Grabowski
2000-02-18In vmapbuf expand uvm_km_valloc_wait into a direct call to uvm_mapArtur Grabowski
so that we can specify an offset. This allows uvm_map to use PMAP_PREFER and removes a big source of bad cache aliases. With this change I have not seen any bad cache aliases during normal use. (it is still possible to force them).
2000-02-18full support for a serial console.Michael Shalayeff
iface addition: pass serial speed from /boot. tested on 4 /boot+kernel permutations. also, niklas@ said he likes it bugs: any printfs before cninit() will go to the default console, which is mon/kbd, but are usually only sorts of debugging messages (like bios_memmap load).
generated by cgit v1.2.3 (git 2.25.1) at 2024-11-25 02:31:58 +0000