- 25 Jun, 2005 40 commits
-
-
Eric W. Biederman authored
For one kernel to report a crash another kernel has created we need to have 2 kernels loaded simultaneously in memory. To accomplish this the two kernels need to built to run at different physical addresses. This patch adds the CONFIG_PHYSICAL_START option to the x86_64 kernel so we can do just that. You need to know what you are doing and the ramifications are before changing this value, and most users won't care so I have made it depend on CONFIG_EMBEDDED bzImage kernels will work and run at a different address when compiled with this option but they will still load at 1MB. If you need a kernel loaded at a different address as well you need to boot a vmlinux. Signed-off-by:
Eric Biederman <ebiederm@xmission.com> Signed-off-by:
Andrew Morton <akpm@osdl.org> Signed-off-by:
Linus Torvalds <torvalds@osdl.org>
-
Vivek Goyal authored
This patch fixes a problem with reserving memory during boot up of a kernel built for non-default location. Currently boot memory allocator reserves the memory required by kernel image, boot allocaotor bitmap etc. It assumes that kernel is loaded at 1MB (HIGH_MEMORY hard coded to 1024*1024). But kernel can be built for non-default locatoin, hence existing hardcoding will lead to reserving unnecessary memory. This patch fixes it. Signed-off-by:
Vivek Goyal <vgoyal@in.ibm.com> Signed-off-by:
-
Eric W. Biederman authored
For one kernel to report a crash another kernel has created we need to have 2 kernels loaded simultaneously in memory. To accomplish this the two kernels need to built to run at different physical addresses. This patch adds the CONFIG_PHYSICAL_START option to the x86 kernel so we can do just that. You need to know what you are doing and the ramifications are before changing this value, and most users won't care so I have made it depend on CONFIG_EMBEDDED bzImage kernels will work and run at a different address when compiled with this option but they will still load at 1MB. If you need a kernel loaded at a different address as well you need to boot a vmlinux. Signed-off-by:
-
Eric W. Biederman authored
The vmlinux on x86_64 does not report the correct physical address of the kernel. Instead in the physical address field it currently reports the virtual address of the kernel. This is patch is a bug fix that corrects vmlinux to report the proper physical addresses. This is potentially a help for crash dump analysis tools. This definitiely allows bootloaders that load vmlinux as a standard ELF executable. Bootloaders directly loading vmlinux become of practical importance when we consider the kexec on panic case. Signed-off-by:
-
Eric W. Biederman authored
The vmlinux on i386 does not report the correct physical address of the kernel. Instead in the physical address field it currently reports the virtual address of the kernel. This is patch is a bug fix that corrects vmlinux to report the proper physical addresses. This is potentially a help for crash dump analysis tools. This definitiely allows bootloaders that load vmlinux as a standard ELF executable. Bootloaders directly loading vmlinux become of practical importance when we consider the kexec on panic case. Signed-off-by:
-
Eric W. Biederman authored
In vmlinux.lds.h the code is carefull to define every section so vmlinux properly reports the correct physical load address of code, as well as it's virtual address. The new SECURITY_INIT definition fails to follow that convention and and causes incorrect physical address to appear in the vmlinux if there are any security initcalls. This patch updates the SECURITY_INIT to follow the convention in the rest of the file. Signed-off-by:
-
Eric W. Biederman authored
When coming out of apic mode attempt to set the appropriate apic back into virtual wire mode. This improves on previous versions of this patch by by never setting bot the local apic and the ioapic into veritual wire mode. This code looks at data from the mptable to see if an ioapic has an ExtInt input to make this decision. A future improvement is to figure out which apic or ioapic was in virtual wire mode at boot time and to remember it. That is potentially a more accurate method, of selecting which apic to place in virutal wire mode. Signed-off-by:
-
Eric W. Biederman authored
When coming out of apic mode attempt to set the appropriate apic back into virtual wire mode. This improves on previous versions of this patch by by never setting bot the local apic and the ioapic into veritual wire mode. This code looks at data from the mptable to see if an ioapic has an ExtInt input to make this decision. A future improvement is to figure out which apic or ioapic was in virtual wire mode at boot time and to remember it. That is potentially a more accurate method, of selecting which apic to place in virutal wire mode. Signed-off-by:
-
Eric W. Biederman authored
From: Eric W. Biederman <ebiederm@xmission.com The following patch simply adds a shutdown method to the x86_64 i8259 code. Signed-off-by:
-
Eric W. Biederman authored
From: Eric W. Biederman <ebiederm@xmission.com> This patch disables interrupt generation from the legacy pic on reboot. Now that there is a sys_device class it should not be called while drivers are still using interrupts. There is a report about this breaking ACPI power off on some systems. http://bugme.osdl.org/show_bug.cgi?id=4041 However the final comment seems to exonerate this code. So until I get more information I believe that was a false positive. Signed-off-by:
-
Eric W. Biederman authored
From: Eric W. Biederman <ebiederm@xmission.com> It is ok to reserve resources > 4G on x86_64 struct resource is 64bit now :) Signed-off-by:
-
Eric W. Biederman authored
From: "Maciej W. Rozycki" <macro@linux-mips.org> Fix a kexec problem whcih causes local APIC detection failure. The problem is detect_init_APIC() is called early, before the command line have been processed. Therefore "lapic" (and "nolapic") have not been seen, yet. Signed-off-by:
Maciej W. Rozycki <macro@linux-mips.org> Signed-off-by:
-
Eric W. Biederman authored
From: "Maciej W. Rozycki" <macro@linux-mips.org> Rename APIC_MODE_EXINT to APIC_MODE_EXTINT - I think it should be named after what the mode is called in documentation. From: "Eric W. Biederman" <ebiederm@lnxi.com> I have reduced this patch to just the name change in the header. And integrated the changes into the patches that add those lines. Otherwise I ran into some ugly dependencies. Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org Signed-off-by:
-
Ingo Molnar authored
This patch adds a new preemption model: 'Voluntary Kernel Preemption'. The 3 models can be selected from a new menu: (X) No Forced Preemption (Server) ( ) Voluntary Kernel Preemption (Desktop) ( ) Preemptible Kernel (Low-Latency Desktop) we still default to the stock (Server) preemption model. Voluntary preemption works by adding a cond_resched() (reschedule-if-needed) call to every might_sleep() check. It is lighter than CONFIG_PREEMPT - at the cost of not having as tight latencies. It represents a different latency/complexity/overhead tradeoff. It has no runtime impact at all if disabled. Here are size stats that show how the various preemption models impact the kernel's size: text data bss dec hex filename 3618774 547184 179896 4345854 424ffe vmlinux.stock 3626406 547184 179896 4353486 426dce vmlinux.voluntary +0.2% 3748414 548640 179896 4476950 445016 vmlinux.preempt +3.5% voluntary-preempt is +0.2% of .text, preempt is +3.5%. This feature has been tested for many months by lots of people (and it's also included in the RHEL4 distribution and earlier variants were in Fedora as well), and it's intended for users and distributions who dont want to use full-blown CONFIG_PREEMPT for one reason or another. Signed-off-by:Ingo Molnar <mingo@elte.hu> Signed-off-by:
-
Ingo Molnar authored
The only sane way to clean up the current 3 lock_kernel() variants seems to be to remove the spinlock-based BKL implementations altogether, and to keep the semaphore-based one only. If we dont want to do that for whatever reason then i'm afraid we have to live with the current complexity. (but i'm open for other cleanup suggestions as well.) To explore this possibility we'll (at a minimum) have to know whether the semaphore-based BKL works fine on plain SMP too. The patch below enables this. The patch may make sense in isolation as well, as it might bring performance benefits: code that would formerly spin on the BKL spinlock will now schedule away and give up the CPU. It might introduce performance regressions as well, if any performance-critical code uses the BKL heavily and gets overscheduled due to the semaphore. I very much hope there is no such performance-critical codepath left though. Signed-off-by:
-
Ingo Molnar authored
This patch consolidates the CONFIG_PREEMPT and CONFIG_PREEMPT_BKL preemption options into kernel/Kconfig.preempt. This, besides reducing source-code, also enables more centralized tweaking of preemption related options. Signed-off-by:
-
Dinakar Guniguntala authored
ia64 changes similar to kernel/sched.c. Signed-off-by:
Dinakar Guniguntala <dino@in.ibm.com> Acked-by:
Paul Jackson <pj@sgi.com> Acked-by:
Nick Piggin <nickpiggin@yahoo.com.au> Signed-off-by:
-
Dinakar Guniguntala authored
Adds the core update_cpu_domains code and updated cpusets documentation Signed-off-by:
-
Dinakar Guniguntala authored
The following patches add dynamic sched domains functionality that was extensively discussed on lkml and lse-tech. I would like to see this added to -mm o The main advantage with this feature is that it ensures that the scheduler load balacing code only balances against the cpus that are in the sched domain as defined by an exclusive cpuset and not all of the cpus in the system. This removes any overhead due to load balancing code trying to pull tasks outside of the cpu exclusive cpuset only to be prevented by the tasks' cpus_allowed mask. o cpu exclusive cpusets are useful for servers running orthogonal workloads such as RT applications requiring low latency and HPC applications that are throughput sensitive o It provides a new API partition_sched_domains in sched.c that makes dynamic sched domains possible. o cpu_exclusive cpusets sets are now associated with a sched domain. Which means that the users can dynamically modify the sched domains through the cpuset file system interface o ia64 sched domain code has been updated to support this feature as well o Currently, this does not support hotplug. (However some of my tests indicate hotplug+preempt is currently broken) o I have tested it extensively on x86. o This should have very minimal impact on performance as none of the fast paths are affected Signed-off-by:
Matthew Dobson <colpatch@us.ibm.com> Signed-off-by:
-
Olivier Croquette authored
Presently, a process without the capability CAP_SYS_NICE can not change its own policy, which is OK. But it can also not decrease its RT priority (if scheduled with policy SCHED_RR or SCHED_FIFO), which is what this patch changes. The rationale is the same as for the nice value: a process should be able to require less priority for itself. Increasing the priority is still not allowed. This is for example useful if you give a multithreaded user process a RT priority, and the process would like to organize its internal threads using priorities also. Then you can give the process the highest priority needed N, and the process starts its threads with lower priorities: N-1, N-2... The POSIX norm says that the permissions are implementation specific, so I think we can do that. In a sense, it makes the permissions consistent whatever the policy is: with this patch, process scheduled by SCHED_FIFO, SCHED_RR and SCHED_OTHER can all decrease their priority. From: Ingo Molnar <mingo@elte.hu> cleaned up and merged to -mm. Signed-off-by:
-
Chen Shang authored
micro-optimize task requeueing in schedule() & clean up recalc_task_prio(). Signed-off-by:
-
Nick Piggin authored
The maximum rebalance interval allowed by the multiprocessor balancing backoff is often not large enough to handle corner cases where there are lots of tasks pinned on a CPU. Suresh reported: I see system livelock's if for example I have 7000 processes pinned onto one cpu (this is on the fastest 8-way system I have access to). After this patch, the machine is reported to go well above this number. Signed-off-by:
-
Nick Piggin authored
Consolidate balance-on-exec with balance-on-fork. This is made easy by the sched-domains RCU patches. As well as the general goodness of code reduction, this allows the runqueues to be unlocked during balance-on-fork. schedstats is a problem. Maybe just have balance-on-event instead of distinguishing fork and exec? Signed-off-by:
-
Nick Piggin authored
One of the problems with the multilevel balance-on-fork/exec is that it needs to jump through hoops to satisfy sched-domain's locking semantics (that is, you may traverse your own domain when not preemptable, and you may traverse others' domains when holding their runqueue lock). balance-on-exec had to potentially migrate between more than one CPU before finding a final CPU to migrate to, and balance-on-fork needed to potentially take multiple runqueue locks. So bite the bullet and make sched-domains go completely RCU. This actually simplifies the code quite a bit. From: Ingo Molnar <mingo@elte.hu> schedstats RCU fix, and a nice comment on for_each_domain, from Ingo. Signed-off-by:
-
Nick Piggin authored
The fundamental problem that Suresh has with balance on exec and fork is that it only tries to balance the top level domain with the flag set. This was worked around by removing degenerate domains, but is still a problem if people want to start using more complex sched-domains, especially multilevel NUMA that ia64 is already using. This patch makes balance on fork and exec try balancing over not just the top most domain with the flag set, but all the way down the domain tree. Signed-off-by:
-
Suresh Siddha authored
Remove degenerate scheduler domains during the sched-domain init. For example on x86_64, we always have NUMA configured in. On Intel EM64T systems, top most sched domain will be of NUMA and with only one sched_group in it. With fork/exec balances(recent Nick's fixes in -mm tree), we always endup taking wrong decisions because of this topmost domain (as it contains only one group and find_idlest_group always returns NULL). We will endup loading HT package completely first, letting active load balance kickin and correct it. In general, this patch also makes sense with out recent Nick's fixes in -mm. From: Nick Piggin <nickpiggin@yahoo.com.au> Modified to account for more than just sched_groups when scanning for degenerate domains by Nick Piggin. And allow a runqueue's sd to go NULL rather than keep a single degenerate domain around (this happens when you run with maxcpus=1). Signed-off-by:
Suresh Siddha <suresh.b.siddha@intel.com> Signed-off-by:
-
Nick Piggin authored
Fix the last 2 places that directly access a runqueue's sched-domain and assume it cannot be NULL. That allows the use of NULL for domain, instead of a dummy domain, to signify no balancing is to happen. No functional changes. Signed-off-by:
-
Nick Piggin authored
Instead of requiring architecture code to interact with the scheduler's locking implementation, provide a couple of defines that can be used by the architecture to request runqueue unlocked context switches, and ask for interrupts to be enabled over the context switch. Also replaces the "switch_lock" used by these architectures with an oncpu flag (note, not a potentially slow bitflag). This eliminates one bus locked memory operation when context switching, and simplifies the task_running function. Signed-off-by:
-
Ingo Molnar authored
"Chen, Kenneth W" <kenneth.w.chen@intel.com> uninline task_timeslice() - reduces code footprint noticeably, and it's slowpath code. Signed-off-by: -
Nick Piggin authored
Do some basic initial tuning. Signed-off-by:
-
Nick Piggin authored
Add SCHEDSTAT statistics for sched-balance-fork. Signed-off-by:
-
Nick Piggin authored
Reimplement the balance on exec balancing to be sched-domains aware. Use this to also do balance on fork balancing. Make x86_64 do balance on fork over the NUMA domain. The problem that the non sched domains aware blancing became apparent on dual core, multi socket opterons. What we want is for the new tasks to be sent to a different socket, but more often than not, we would first load up our sibling core, or fill two cores of a single remote socket before selecting a new one. This gives large improvements to STREAM on such systems. Signed-off-by:
-
Nick Piggin authored
Remove the very aggressive idle stuff that has recently gone into 2.6 - it is going against the direction we are trying to go. Hopefully we can regain performance through other methods. Signed-off-by:
-
Nick Piggin authored
Do less affine wakeups. We're trying to reduce dbt2-pgsql idle time regressions here... make sure we don't don't move tasks the wrong way in an imbalance condition. Also, remove the cache coldness requirement from the calculation - this seems to induce sharp cutoff points where behaviour will suddenly change on some workloads if the load creeps slightly over or under some point. It is good for periodic balancing because in that case have otherwise have no other context to determine what task to move. But also make a minor tweak to "wake balancing" - the imbalance tolerance is now set at half the domain's imbalance, so we get the opportunity to do wake balancing before the more random periodic rebalancing gets preformed. Signed-off-by:
-
Nick Piggin authored
Do CPU load averaging over a number of different intervals. Allow each interval to be chosen by sending a parameter to source_load and target_load. 0 is instantaneous, idx > 0 returns a decaying average with the most recent sample weighted at 2^(idx-1). To a maximum of 3 (could be easily increased). So generally a higher number will result in more conservative balancing. Signed-off-by:
-
Nick Piggin authored
Remove the special casing for idle CPU balancing. Things like this are hurting for example on SMT, where are single sibling being idle doesn't really warrant a really aggressive pull over the NUMA domain, for example. Signed-off-by:
-
Nick Piggin authored
These conditions should now be impossible, and we need to fix them if they happen. Signed-off-by:
-
Nick Piggin authored
SMT balancing has a couple of problems. Firstly, active_load_balance is too complex - basically it should be a dumb helper for when the periodic balancer has determined there is an imbalance, but gets stuck because the task is running. So rip out all its "smarts", and just make it move one task to the target CPU. Second, the busy CPU's sched-domain tree was being used for active balancing. This means that it may not see that nr_balance_failed has reached a critical level. So use the target CPU's sched-domain tree for this. We can do this because we hold its runqueue lock. Lastly, reset nr_balance_failed to a point where we allow cache hot migration. This will help ensure active load balancing is successful. Thanks to Suresh Siddha for pointing out these issues. Signed-off-by:
-
Nick Piggin authored
Fix up active load balancing a bit so it doesn't get called when it shouldn't. Reset the nr_balance_failed counter at more points where we have found conditions to be balanced. This reduces too aggressive active balancing seen on some workloads. Signed-off-by:
-
Nick Piggin authored
John Hawkes explained the problem best: A large number of processes that are pinned to a single CPU results in every other CPU's load_balance() seeing this overloaded CPU as "busiest", yet move_tasks() never finds a task to pull-migrate. This condition occurs during module unload, but can also occur as a denial-of-service using sys_sched_setaffinity(). Several hundred CPUs performing this fruitless load_balance() will livelock on the busiest CPU's runqueue lock. A smaller number of CPUs will livelock if the pinned task count gets high. Expanding slightly on John's patch, this one attempts to work out whether the balancing failure has been due to too many tasks pinned on the runqueue. This allows it to be basically invisible to the regular blancing paths (ie. when there are no pinned tasks). We can use this extra knowledge to shut down the balancing faster, and ensure the migration threads don't start running which is another problem observed in the wild. Signed-off-by:
-
