summaryrefslogtreecommitdiffstats
path: root/arch/mips64
Commit message (Collapse)AuthorAgeFilesLines
* Add autofs ioctls.Ralf Baechle2000-09-081-16/+47
|
* Enable some more features now that the thing is getting usefull.Ralf Baechle2000-09-082-14/+24
|
* Reformat.Ralf Baechle2000-09-081-1/+2
|
* Erase all traces of the ioc3-eth driver irq hacks.Kanoj Sarcar2000-09-071-2/+3
|
* Delete usage of PG_skip flag from mips64-DISCONTIG code. It was onlyKanoj Sarcar2000-09-071-22/+6
| | | | | | | being used to flag mem_map entries corresponding to holes for counting and reporting during boot time. The current code reports more accurate figures, based on what is found at szmem() time (the old code was reporting more total memory than szmem() had probed).
* Compile fix: flush only L2 cache.Kanoj Sarcar2000-09-061-1/+1
|
* Set timeout values - this seems to be needed if we want to boot theKanoj Sarcar2000-09-061-0/+8
| | | | kernel on non-nasid0 systems.
* NTP fixes from Maciej.Ralf Baechle2000-09-062-6/+7
|
* - adjtimex(2) emulation.Ralf Baechle2000-09-053-50/+131
| | | | | | | | | | | | | | - set_rtc_mmss was crashing the machine. This went unnoticed so far only because it only happens one the machine is time synchronized. The machine now synchronizes with NTP quite well. - Better personality(2) emulation. This code is taken from Sparc64 which seems to do the right thing but I still wonder how it's supposed to work there. Uname(2) returning sparc64 or mips64 instead of sparc or mips is breaking alot of configuration scripts but on the other side returning the arch name without the 64 would prevent easy detection of the architecture by autoconf style scripts. I also don't see how the personality of 32-bit processes is ever being set to PER_LINUX32.
* Handle a ton more block device, IDE (yuck :-), partition handling andRalf Baechle2000-09-055-45/+285
| | | | | tape ioctls. Dump support for ancient variants of uname(2) syscall; Nothing is using the since Linux/MIPS history started with glibc 2.0.
* Slightly tweak how the ioc3 fixup interacts with serial driver. Also,Kanoj Sarcar2000-08-311-1/+2
| | | | | hardcode the 0xFF00 value into subsystem_vendor, instead of relying on PCI commands to ioc3 to do this.
* No more hardcoded serial port addresses. Get the serial driver probingKanoj Sarcar2000-08-311-0/+16
| | | | addresses for ports. Serial driver changes are being sent to Linus.
* Define prom_putchar/prom_getchar() for IP27s. The IP27 console printkKanoj Sarcar2000-08-292-10/+13
| | | | routine now uses prom_printf to correctly parse the format string.
* Merge with 2.4.0-test7.Ralf Baechle2000-08-287-28/+74
|
* More debris.Ralf Baechle2000-08-271-29/+0
|
* Desinfect console code from #ifdefmania.Ralf Baechle2000-08-276-39/+68
|
* Merge with Linux 2.4.0-test6-pre10.Ralf Baechle2000-08-251-2/+2
|
* Do the bare minimum pci fixups needed for the IOC3, not more ...Kanoj Sarcar2000-08-241-3/+0
|
* On an SMP system, every cpu should use the master node's IOC3 uart portKanoj Sarcar2000-08-241-3/+2
| | | | for printk output, instead of trying to access its own.
* Use prom_printf() as the primary method of doing printks() for IP27.Kanoj Sarcar2000-08-241-0/+24
|
* Make prom_printf() functional on IP27s. And prom_printf() is not anKanoj Sarcar2000-08-232-2/+28
| | | | init function, it needs to be around during regular system usage.
* Kill warning.Ralf Baechle2000-08-091-10/+13
|
* Faster memcpy / copy_{from,to}_user.Ralf Baechle2000-08-091-31/+76
|
* Kill BARRIER macro; it's not needed on R10000.Ralf Baechle2000-08-091-25/+0
|
* Merge with Linux 2.4.0-test6-pre8.Ralf Baechle2000-08-083-15/+16
|
* Merge with Linux 2.4.0-test6-pre4.Ralf Baechle2000-08-085-15/+7
|
* Port the recent cache changes forward and backward between mips andRalf Baechle2000-08-081-211/+70
| | | | mips64.
* Tweak the initialization macros so that they work even on a systemKanoj Sarcar2000-08-051-3/+1
| | | | that has no physical address 0 (non nasid 0 systems).
* sys_sysctl implementationUlf Carlsson2000-08-052-1/+89
|
* Fix over enthusiastic tlbflush optimizations.Kanoj Sarcar2000-08-031-3/+6
|
* ARC console is not just for IP22.Keith M Wesolowski2000-08-011-1/+1
| | | | Timer dead code cleanup.
* Compile fix: get the ip27-timer.c code that updates process times lookingKanoj Sarcar2000-08-011-25/+10
| | | | like its i386 counterpart.
* Shave of 50% of lat_mmap. Our cache routines were plain stupid.Ralf Baechle2000-07-311-1/+1
|
* Smarter and more scalable tlb flushing routines. No need to interruptKanoj Sarcar2000-07-301-12/+43
| | | | | | other cpus when flushing tlbs for single threaded mm's. Rather, just make sure the mm will have to do a new context allocation if it runs on the other cpus.
* Merge with 2.4.0-test5 final.Ralf Baechle2000-07-283-3/+3
|
* Optimized cache flushing on r10k/o200s, assuming processor handlesKanoj Sarcar2000-07-271-103/+20
| | | | | VCEs in hardwire, and system guarantees io coherency. Only need to do cache flushes for icache coherency.
* Optimize andes_clear_page() and andes_copy_page() with prefetchRalf Baechle2000-07-261-16/+23
| | | | | | | operations. While those routines are now 30% - 60% faster it turned out that their influence on realworld applications and benchmars is fairly low; I saw an improvment of ~ 3% for lmbench exec sh and even less for other benches.
* When a debugger faults in a page for a debugee, we do not need to updateKanoj Sarcar2000-07-252-0/+12
| | | | | | the mmu cache/tlbs, since the debugger will use a kernel address to access the page, and not the user address. This should fix the strace warning messages that Ralf and Ulf have seen. Fix should be backported into MIPS.
* Name change: the generic call flush_cache_all() does not do anythingKanoj Sarcar2000-07-247-24/+24
| | | | | anymore. All the flush_cache_all() calls in MIPS code is changed to call flush_cache_l1(), and ends up flushing the L1 i/d caches.
* Introduce a new cpu specific routine to flush the L2 cache. This helpsKanoj Sarcar2000-07-245-21/+36
| | | | | | the stability of DISCONTIGMEM kernels. The L2 flushing is needed during bootup. With this, the initialization hacks of scanning the node memories at boot up time is not needed any more.
* Fix syscall tracing for mips64, too.Ralf Baechle2000-07-243-10/+10
|
* Merge with Linux 2.4.0-test5-pre4.Ralf Baechle2000-07-233-3/+3
|
* 32 bit sys ipc, stolen from ia64 ...Kanoj Sarcar2000-07-212-1/+415
|
* The 32 bit version of exec (picked from ia64) was writing into userKanoj Sarcar2000-07-201-10/+21
| | | | | | | space from kernel without protection. Under tight memory conditions when the page fault could not allocate a page, irritating messages were being printed by the kernel. Write to user space with proper protection.
* Fix a (harmless) typo in the SMP save_and_cli definition. Also, markKanoj Sarcar2000-07-171-1/+0
| | | | | the irqlock as not held by anyone initially. This allows SMP kernels to boot up.
* Merge with Linux 2.4.0-test5-pre1. This works again on Origin UP.Ralf Baechle2000-07-152-1/+3
| | | | The IP22 cache bugs which are plaguing some machines are still unfixed.
* Fixup the recent ELF loader changes which were toast ...Ralf Baechle2000-07-131-3/+3
|
* Revert last change. Bug noticed by Linus.Ralf Baechle2000-07-121-1/+1
|
* Use --trap for optimization.Ralf Baechle2000-07-121-2/+1
|
* Merge with 2.4.0-test4-pre2.Ralf Baechle2000-07-121-0/+1
|
generated by cgit v1.2.3 (git 2.25.1) at 2024-09-20 19:29:36 +0000