- 23 Aug, 2004 34 commits
-
-
Paul Mackerras authored
The EEH code currently can end up doing an extra pci_dev_put() in the case where we hot-unplug a card for which we are ignoring EEH errors (e.g. a graphics card). This patch fixes that problem by only maintaining a reference to the PCI device if we have entered any of its resource addresses into our address -> PCI device cache. This patch is based on an earlier patch by Linas Vepstas. Signed-off-by:
Paul Mackerras <paulus@samba.org> Signed-off-by:
Andrew Morton <akpm@osdl.org> Signed-off-by:
Linus Torvalds <torvalds@osdl.org>
-
Paul Mackerras authored
Firmware can report errors at any time, and not atypically during boot. However, these reports were being discarded until th rtasd comes up, which occurs fairly late in the boot cycle. As a result, firmware errors during boot were being silently ignored. Signed-off-by:
Linas Vepstas <linas@linas.org> Signed-off-by:
-
David Gibson authored
Fairly trivial PPC64 cleanup. This patch makes the ppc64 INIT_THREAD #define use C99 initializers, which will make it less likely to get broken if we need to change thread_struct. Signed-off-by:
David Gibson <dwg@au1.ibm.com> Signed-off-by:
-
Paul Mackerras authored
In the default_idle and dedicated_idle loops, there are some inner loops out of which we should break if the cpu is marked offline. Otherwise, it is possible for the cpu to get stuck and never actually go offline. shared_idle is unaffected. Signed-off-by:
Nathan Lynch <nathanl@austin.ibm.com> Signed-off-by:
-
Paul Mackerras authored
When taking a cpu offline, once the cpu has been removed from cpu_online_map, it is not supposed to service any more interrupts. This presents a problem on ppc64 because we cannot truly disable the decrementer. There used to be cpu_is_offline() checks in several scheduler functions (e.g. rebalance_tick()) which papered over this issue, but these checks were removed recently. So with recent 2.6 kernels, an attempt to offline a cpu can result in a crash in find_busiest_group(). This patch prevents such crashes. Signed-off-by:
-
Paul Mackerras authored
Here is a patch that sets struct iommu_table->it_type to TCE_PCI in pSeries_iommu.c. This is just for code completeness (and it is updated in iSeries_iommu.c, but was somehow missed in pSeries_iommu.c). Signed-off-by:
Ananth N Mavinakayanahalli <ananth@in.ibm.com> Signed-off-by:
-
Anton Blanchard authored
Clean up an oprofile error message, it was missing a newline. Signed-off-by:
Anton Blanchard <anton@samba.org> Signed-off-by:
-
Anton Blanchard authored
Add some missing exports, required for oprofile to be compiled as a module. Signed-off-by:
-
Anton Blanchard authored
Allow the oprofile module to be unloaded, before we never removed the oprofile specific interrupt handler. Handle the pending exception case in the dummy interrupt handler instead. Signed-off-by:
-
Anton Blanchard authored
Disable oprofile debug messages. They arent much use now things are working reliably. Signed-off-by:
-
Anton Blanchard authored
POWER4 oprofile updates from Carl Love. - Create mmcr0, mmcr1, mmcra oprofilefs files. - Use kernel and user profile disable bits. (Some modifications by me) Signed-off-by:
-
Paul Mackerras authored
With cpu_present_map, we don't need these any longer. Signed-off-by:
-
Paul Mackerras authored
Our (ab)use of cpu_possible_map in setup_system to start secondary SMT threads bothers me. Mark such threads in cpu_possible_map during early boot; let RTAS tell us which present cpus are still offline later so we can start them. I'm not totally sure about this one, it might be better to set up cpu_sibling_map in prom_hold_cpus and use that in setup_system. Signed-off-by:
-
Paul Mackerras authored
Adopt the "standard" cpu_present_map for describing cpus which are present in the system, but not necessarily online. cpu_present_map is meant to be a superset of cpu_online_map and a subset of cpu_possible_map. Signed-off-by:
-
Paul Mackerras authored
We were using Linux's cpu numbering for cpu-related hypervisor calls (e.g. vpa registration, H_CONFER). It happened to work most of the time because Linux and the hypervisor usually, but not always, have the same numbering for cpus. Signed-off-by:
-
Dave Hansen authored
arch/ppc64/kernel/irq.c: In function `init_irq_proc': arch/ppc64/kernel/irq.c:797: warning: implicit declaration of function `create_prof_cpu_mask' Signed-off-by:
Dave Hansen <haveblue@us.ibm.com> Signed-off-by:
-
Paul Mackerras authored
Somewhere along the line we lost the code that updates some fields of the systemcfg structure that are used for translating timebase values to time of day. I want to get rid of the systemcfg structure eventually, but applications are using it (and in particular these fields) and I don't want to break the ABI in a stable kernel series. Signed-off-by:
-
Anton Blanchard authored
Remove some unused things in asm-offsets.c Signed-off-by:
-
Anton Blanchard authored
Reduce the stack overflow warning from 4kB to 2kB now that its been in and tested for a while. Signed-off-by:
-
Matt Porter authored
This patch removes warnings associated with Ebony MTD related defines. Please apply. Signed-off-by:
Eugene Surovegin <ebs@ebshome.net> Signed-off-by:
Matt Porter <mporter@kernel.crashing.org> Signed-off-by:
-
Paul Mackerras authored
This patch fixes a bug in the kernel emulation of altivec instructions with denormalized operands. The emulation of the vmaddfp and vmnsubfp instructions was giving the wrong answer because I had the wrong order of operands to the fmadds and fnmsubs instructions. This patch fixes it for both ppc32 and ppc64. Signed-off-by:
-
Eugene Surovegin authored
This patch adds missing exports for __dma_sync and __dma_sync_page (DMA API helpers for non-coherent cache PPCs). Signed-off-by:
-
Matt Porter authored
Adds documentation of the PPC noltlbs and nobats kernel cmdline parameters. noltlbs is a new option and nobats never had an entry. Signed-off-by:
-
Paul Mackerras authored
This patch adds emulation in the illegal instruction handler for a couple of old instructions that are no longer implemented in the PPC970 and later chips. This patch adds the code for both ppc32 and ppc64, and cleans up the ppc64 traps.c a bit, along the lines of the ppc32 code. It also makes sure that the ppc64 code generates a SIGTRAP after emulating an instruction if single-stepping is enabled. Signed-off-by:
-
Paul Mackerras authored
This patch adds code to the ppc32 alignment exception handler to make it handle the load/store string and load/store multiple word instructions. This is an issue for older CPUs such as the PPC601, which traps on load/store string instructions which cross a page boundary (newer CPUs handle this in hardware). I have a little test program which exercises this code, so I am reasonably confident it's correct. Signed-off-by:
-
Matt Porter authored
This makes the PPC40x lowmem large tlb mapping selectable via a cmdline option. This allows use of the normal page-sized mapping so that kernel text can be read only if desired. Signed-off-by:
Josh Boyer <jwboyer@charter.net> Signed-off-by:
-
Matt Porter authored
The following patch fixes the situation where the loop condition could generate a next_dec of zero while exiting the loop. This is suboptimal on Classic PPC because it forces another interrupt to occur and reenter the handler. It is fatal on Book E cores, because their decrementer is stopped when writing a zero (Classic interrupts on a 0->-1 transition, Book E interrupts on a 1->0 transition). Instead, stay in the loop on a next_dec==0. Signed-off-by:
-
Benjamin Herrenschmidt authored
This patch by Vincent Hanquez removes some hard coded offsets for accessing thread info fields from assembly, uses the normal offset generation mecanism that we already have for other things instead. Signed-off-by:
Vincent Hanquez <tab@snarc.org> Signed-off-by:
Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by:
-
Keith Owens authored
Make i386 die() more resilient against recursive errors, almost a cut and paste of the ia64 die() routine. Much of the patch is indentation changes. Mainly to make it easier to add crash, lcrash, kmsgdump or other RAS patches. They are invoked from die() and if they crash themselves, we have to avoid recursive loops in die(). Signed-off-by:
Keith Owens <kaos@sgi.com> Signed-off-by:
-
Akiyama Nobuyuki authored
I made a patch for debugging with the help of NMI trigger switch. When kernel hangs severely, keyboard operation(e.g.Ctrl-Alt-Del) doesn't work properly. This patch enables debugging information to be displayed on console in this case. I think this feature is necessary as standard functionality. Please feel free to use this patch and let me know if you have any comments. Background: When a trouble occurs in kernel, we usually begin to investigate with following information: - panic >> panic message. - oops >> CPU registers and stack trace. - hang >> **NONE** no standard method established. How it works: Most IA32 servers have a NMI switch that fires NMI interrupt up. The NMI interrupt can interrupt even if kernel is serious state, for example deadlock under the interrupt disabled. When the NMI switch is pressed after this feature is activated, CPU registers and stack trace are displayed on console and then panic occurs. This feature is activated or deactivated with sysctl. On IA32 architecture, only the following are defined as reason of NMI interrupt: - memory parity error - I/O check error The reason code of NMI switch is not defined, so this patch assumes that all undefined NMI interrupts are fired by MNI switch. However, oprofile and NMI watchdog also use undefined NMI interrupt. Therefore this feature cannot be used at the same time with oprofile and NMI watchdog. This feature hands NMI interrupt over to oprofile and NMI watchdog. So, when they have been activated, this feature doesn't work even if it is activated. Supported architecture: IA32 Setup: Set up the system control parameter as follows: # sysctl -w kernel.unknown_nmi_panic=1 kernel.unknown_nmi_panic = 1 If the NMI switch is pressed, CPU registers and stack trace will be displayed on console and then panic occurs. Signed-off-by:
-
Arnd Bergmann authored
Reading the contents of a module_param_string through sysfs currently oopses because the param_get_charp() function cannot operate on a kparam_string struct. This introduces the required param_get_string. Signed-off-by:
Arnd Bergmann <arnd@arndb.de> Signed-off-by:
-
Mike Kravetz authored
Races have been observed between excec-time overwriting of task->comm and /proc accesses to the same data. This causes environment string information to appear in /proc. Fix that up by taking task_lock() around updates to and accesses to task->comm. Signed-off-by:
-
Randy Dunlap authored
add_pin_to_irq() should not be __init; it is used after init code. Signed-off-by:
Randy Dunlap <rddunlap@osdl.org> Signed-off-by:
-
Ingo Molnar authored
while debugging/improving scheduling latencies i got the following strange latency report from Lee Revell: http://krustophenia.net/testresults.php?dataset=2.6.8.1-P6#/var/www/2.6.8.1-P6 this trace shows a 120 usec latency caused by XFree86, on a 600 MHz x86 system. Looking closer reveals: 00000002 0.006ms (+0.003ms): __switch_to (schedule) 00000002 0.088ms (+0.082ms): finish_task_switch (schedule) it took more than 80 usecs for XFree86 to do a context-switch! it turns out that the reason for this (massive) context-switching overhead is the following change in 2.6.8: [PATCH] larger IO bitmaps To demonstrate the effect of this change i've written ioperm-latency.c (attached), which gives the following on vanilla 2.6.8.1: # ./ioperm-latency default no ioperm: scheduling latency: 2528 cycles turning on port 80 ioperm: scheduling latency: 10563 cycles turning on port 65535 ioperm: scheduling latency: 10517 cycles the ChangeSet says: Now, with the lazy bitmap allocation and per-CPU TSS, this will really not drain any resources I think. this is plain wrong. An increase in the IO bitmap size introduces per-context-switch overhead as well: we now have to copy an 8K bitmap every time XFree86 context-switches - even though XFree86 never uses ports higher than 1024! I've straced XFree86 on a number of x86 systems and in every instance ioperm() was used - so i'd say the majority of x86 Linux systems running 2.6.8.1 are affected by this problem. This not only causes lots of overhead, it also trashes ~16K out of the L1 and L2 caches, on every context-switch. It's as if XFree86 did a L1 cache flush on every context-switch ... the simple solution would be to revert IO_BITMAP_BITS back to 1024 and release 2.6.8.2? I've implemented another solution as well, which tracks the highest-enabled port # for every task and does the copying of the bitmap intelligently. (patch attached) The patched kernel gives: # ./ioperm-latency default no ioperm: scheduling latency: 2423 cycles turning on port 80 ioperm: scheduling latency: 2503 cycles turning on port 65535 ioperm: scheduling latency: 10607 cycles this is much more acceptable - the full overhead only occurs in the very unlikely event of a task using the high ioport range. X doesnt suffer any significant overhead. (tracking the maximum allowed port # also allows a simplification of io_bitmap handling: e.g. we dont do the invalid-offset trick anymore - the IO bitmap in the TSS is always valid and secure.) I tested the patch on x86 SMP and UP, it works fine for me. I tested boundary conditions as well, it all seems secure. Ingo #include <errno.h> #include <stdio.h> #include <sched.h> #include <signal.h> #include <sys/io.h> #include <stdlib.h> #include <unistd.h> #include <linux/unistd.h> #define CYCLES(x) asm volatile ("rdtsc" :"=a" (x)::"edx") #define __NR_sched_set_affinity 241 _syscall3 (int, sched_set_affinity, pid_t, pid, unsigned int, mask_len, unsigned long *, mask) /* * Use a pair of RT processes bound to the same CPU to measure * context-switch overhead: */ static void measure(void) { unsigned long i, min = ~0UL, pid, mask = 1, t1, t2; sched_set_affinity(0, sizeof(mask), &mask); pid = fork(); if (!pid) for (;;) { asm volatile ("sti; nop; cli"); sched_yield(); } sched_yield(); for (i = 0; i < 100; i++) { asm volatile ("sti; nop; cli"); CYCLES(t1); sched_yield(); CYCLES(t2); if (i > 10) { if (t2 - t1 < min) min = t2 - t1; } } asm volatile ("sti"); kill(pid, 9); printf("scheduling latency: %ld cycles\n", min); sched_yield(); } int main(void) { struct sched_param p = { sched_priority: 2 }; unsigned long mask = 1; if (iopl(3)) { printf("need to run as root!\n"); exit(-1); } sched_setscheduler(0, SCHED_FIFO, &p); sched_set_affinity(0, sizeof(mask), &mask); printf("default no ioperm: "); measure(); printf("turning on port 80 ioperm: "); ioperm(0x80,1,1); measure(); printf("turning on port 65535 ioperm: "); if (ioperm(0xffff,1,1)) printf("FAILED - older kernel.\n"); else measure(); return 0; } Signed-off-by:
Ingo Molnar <mingo@elte.hu> Signed-off-by:
-
- 22 Aug, 2004 6 commits
-
-
Linus Torvalds authored
Signed single-bit bitfields really are a pretty strange thing to have. They work, but it wasn't really intentional.
-
http://linux-watchdog.bkbits.net/linux-2.6-watchdogLinus Torvalds authored
into ppc970.osdl.org:/home/torvalds/v2.6/linux
-
http://xfs.org:8090/xfs-linux-2.6Linus Torvalds authored
into ppc970.osdl.org:/home/torvalds/v2.6/linux
-
Cal Peake authored
net/ipv4/proc.c was updated to use a new mechanism for outputting /proc/net/snmp and /proc/net/netstat. However, a superfluous '\n' snuck in, breaking `netstat -s`
-
bk://linux-dj.bkbits.net/cpufreqLinus Torvalds authored
into ppc970.osdl.org:/home/torvalds/v2.6/linux
-
Linus Torvalds authored
This one from Dave Jones, who read the Intel docs even more.
-
