Age | Commit message (Collapse) | Author |
|
Currently we calculate the first timeslice for every context
incorrectly - alloc_spu_context calls spu_set_timeslice before we set
ctx->prio so we always calculate the longest possible timeslice for the
lowest possible priority.
This patch makes sure to update the schedule-related fields before
calculating the timeslice and also makes sure we update the timeslice for
a non-running context when entering spu_run so a priority change affects
the context as soon as possible.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
|
|
From: Maynard Johnson <mpjohn@us.ibm.com>
This patch updates the existing arch/powerpc/oprofile/op_model_cell.c
to add in the SPU profiling capabilities. In addition, a 'cell' subdirectory
was added to arch/powerpc/oprofile to hold Cell-specific SPU profiling code.
Exports spu_set_profile_private_kref and spu_get_profile_private_kref which
are used by OProfile to store private profile information in spufs data
structures.
Also incorporated several fixes from other patches (rrn). Check pointer
returned from kzalloc. Eliminated unnecessary cast. Better error
handling and cleanup in the related area. 64-bit unsigned long parameter
was being demoted to 32-bit unsigned int and eventually promoted back to
unsigned long.
Signed-off-by: Carl Love <carll@us.ibm.com>
Signed-off-by: Maynard Johnson <mpjohn@us.ibm.com>
Signed-off-by: Bob Nelson <rrnelson@us.ibm.com>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
Acked-by: Paul Mackerras <paulus@samba.org>
|
|
This patch adds support for additional flags at spu_create, which relate
to the establishment of affinity between contexts and contexts to memory.
A fourth, optional, parameter is supported. This parameter represent
a affinity neighbor of the context being created, and is used when defining
SPU-SPU affinity.
Affinity is represented as a doubly linked list of spu_contexts.
Signed-off-by: Andre Detsch <adetsch@br.ibm.com>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
|
|
Currently a process is removed from the physical spu when spu_acquire_saved
is saved but never put back. This patch adds a new spu_release_saved
that is to be paired with spu_acquire_saved and put the process back if
it has been in RUNNABLE state before.
Niether Jeremy not be are entirely happy about this exact patch because
it adds another spu_activate call outside of the owner thread, but I
feel this is the best short-term fix we can come up with.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
|
|
This patch exports per-context statistics in spufs as long as spu
statistics in sysfs.
It was formed by merging:
"spufs: add spu stats in sysfs" From: Christoph Hellwig
"spufs: add stat file to spufs" From: Christoph Hellwig
"spufs: fix libassist accounting" From: Jeremy Kerr
"spusched: fix spu utilization statistics" From: Luke Browning
And some adjustments by myself, after suggestions on cbe-oss-dev.
Having separate patches was making the review process harder
than it should, as we end up integrating spus and ctx statistics
accounting much more than it was on the first implementation.
Signed-off-by: Andre Detsch <adetsch@br.ibm.com>
Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
|
|
Export per-context statistics in spufs.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
|
|
Provide load average information for spu context. The format
is identical to /proc/loadavg, which is also where a lot of code
and concepts is borrowed from.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
|
|
Add a cpus_allowed allowed filed to struct spu_context so that we always
use the cpu mask of the owning thread instead of the one happening to
call into the scheduler. Also use this information in
grab_runnable_context to avoid spurious wakeups.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
|
|
Update scheduling information on every spu_run to allow for setting
threads to realtime priority just before running them. This requires
some slightly ugly code in spufs_run_spu because we can just update
the information unlocked if the spu is not runnable, but we need to
acquire the active_mutex when it is runnable to protect against
find_victim. This locking scheme requires opencoding
spu_acquire_runnable in spufs_run_spu which actually is a nice cleanup
all by itself.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
|
|
Enable preemptive scheduling for non-RT contexts.
We use the same algorithms as the CPU scheduler to calculate the time
slice length, and for now we also use the same timeslice length as the
CPU scheduler. This might be not enough for good performance and can be
changed after some benchmarking.
Note that currently we do not boost the priority for contexts waiting
on the runqueue for a long time, so contexts with a higher nice value
could starve ones with less priority. This could easily be fixed once
the rework of the spu lists that Luke and I discussed is done.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
|
|
Get rid of the scheduler workqueues that complicated things a lot to
a dedicated spu scheduler thread that gets woken by a traditional
scheduler tick. By default this scheduler tick runs a HZ * 10, aka
one spu scheduler tick for every 10 cpu ticks.
Currently the tick is not disabled when we have less context than
available spus, but I will implement this later.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
|
|
Make sure the mapping_lock also protects access to the various address_space
pointers used for tearing down the ptes on a spu context switch.
Because unmap_mapping_range can sleep we need to turn mapping_lock from
a spinlock into a sleeping mutex.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
|
|
This adds an option to spufs when the kernel is configured for
4K page to give it the ability to use 64K pages for SPE local store
mappings.
Currently, we are optimistic and try order 4 allocations when creating
contexts. If that fails, the code will fallback to 4K automatically.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
|
|
There is no reason for run_sema to be a struct semaphore. Changing
it to a mutex and rename it accordingly.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
|
|
For quite a while now spu state is protected by a simple mutex instead
of the old rw_semaphore, and this means we can simplify the locking
around spu_setup_isolated a lot.
Instead of doing an spu_release before entering spu_setup_isolated and
then calling the complicated spu_acquire_exclusive we can now simply
enter the function locked an in guaranteed runnable state, so that the
only bit of spu_acquire_exclusive that's left is the call to
spu_unmap_mappings.
Similarly there's no more need to unlock and reacquire the state_mutex
when spu_setup_isolated is done, but we can always return with the
lock held and only drop it in spu_run_init in the failure case.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
|
|
A single context should only be woken once, and we should not have
more wakeups for a given priority than the number of contexts on
that runqueue position.
Also add some asserts to trap future problems in this area more
easily.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
|
|
Make sure the pointers to various mappings are cleared once the last
user stopped using them. This avoids accessing freed memory when
tearing down the gang directory aswell as optimizing away
pte invalidations if no one uses these.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
|
|
For SCHED_RR tasks we can do some really trivial timeslicing. Basically
we fire up a time for every scheduler tick that searches for a higher
or same priority thread that is on the runqueue and if there is one
context switches to it. Because we can't lock spus from timer context
we actually run this from a delayed runqueue instead of a timer.
A nice optimization would be to skip the actual priority bitmap search
when there are less contexts than physical spus available. To implement
this I need a so far unpublished patch from Andre, and it will be added
after we have that patch in.
Note that right now we only do the time slicing for SCHED_RR tasks.
The code would work for SCHED_OTHER tasks aswell, but their prio
value is defered from the one the PPU thread has at time of spu_run,
and using this for spu scheduling decisions would make the code very
unfair. SCHED_OTHER support will be enabled once we the spu scheduler
knows how to calculcate cpu_context.prio (very soon)
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
|
|
If we start a spu context with realtime priority we want it to run
immediately and not wait until some other lower priority thread has
finished. Try to find a suitable victim and use it's spu in this
case.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
|
|
There is no need to directly wake up contexts in spu_activate when
called from spu_run, so add a flag to surpress this wakeup.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
|
|
It doesn't make any sense to have a priority field in the physical spu
structure. Move it into the spu context instead.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
|
|
Various cleanups in code surrounding the state semaphore:
- inline spu_acquire/spu_release
- cleanup spu_acquire_* and add kerneldoc comments to these functions
- remove spu_release_exclusive and replace it with spu_release
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
|
|
The r/w semaphore to lock the spus was overkill and can be replaced
with a mutex to make it faster, simpler and easier to debug. It also
helps to allow making most spufs interruptible in future patches.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
|
|
Only bind_context/unbind_context change the spu context state. Thus
we can move all assignents of SPU_STATE_RUNNABLE into bind_context,
which parallels the unbind side aswell.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
|
|
unbind_context already sets the context state to SPU_STATE_SAVED, thus
the spu_deactivate callers don't need to do it again.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
|
|
It looks like we've had some serious bitrot there mostly due to tracking
of address_space's of mmap'ed files getting out of sync with the actual
mmap code. The mfc, mss and psmap were not tracked properly and thus
not invalidated on context switches (oops !)
I also removed the various file->f_mapping = inode->i_mapping;
assignments that were done in the other open() routines since that
is already done for us by __dentry_open.
One improvement we might want to do later is to assign the various
ctx-> fields at mmap time instead of file open/close time so that we
don't call unmap_mapping_range() on thing that have not been mmap'ed
Finally, I added some smp_wmb's after assigning the ctx-> fields to make
sure they are visible to other CPUs. I don't think this is really
necessary as I suspect locking in the fs layer will make that happen
anyway but better safe than sorry.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
|
|
When the user changes the runcontrol register, an SPU might be
running without a process being attached to it and waiting for
events. In order to prevent this, make sure we always disable
the priv1 master control when we're not inside of spu_run.
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
|
|
When in isolated mode, SPEs have access to an area of persistent
storage, which is per-SPE. In order for isolated-mode apps to
communicate arbitrary data through this storage, we need to ensure that
isolated physical SPEs can be reused for subsequent applications.
Add a file ("recycle") in a spethread dir to enable isolated-mode
recycling. By writing to this file, the kernel will reload the
isolated-mode loader kernel, allowing a new app to be run on the same
physical SPE.
This requires the spu_acquire_exclusive function to enforce exclusive
access to the SPE while the loader is initialised.
Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
|
|
Add the concept of a gang to spufs as a new type of object.
So far, this has no impact whatsover on scheduling, but makes
it possible to add that later.
A new type of object in spufs is now a spu_gang. It is created
with the spu_create system call with the flags argument set
to SPU_CREATE_GANG (0x2). Inside of a spu_gang, it
is then possible to create spu_context objects, which until
now was only possible at the root of spufs.
There is a new member in struct spu_context pointing to
the spu_gang it belongs to, if any. The spu_gang maintains
a list of spu_context structures that are its children.
This information can then be used in the scheduler in the
future.
There is still a bug that needs to be resolved in this
basic infrastructure regarding the order in which objects
are removed. When the spu_gang file descriptor is closed
before the spu_context descriptors, we leak the dentry
and inode for the gang. Any ideas how to cleanly solve
this are appreciated.
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
|
|
Use kzalloc when allocating a new spu context, rather than kmalloc +
zeroing.
Booted & tested on cell.
Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
|
|
the mfc member of a new context was not initialized to zero,
which potentially leads to wild memory accesses.
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
|
|
This patch is layered on top of CONFIG_SPARSEMEM
and is patterned after direct mapping of LS.
This patch allows mmap() of the following regions:
"mfc", which represents the area from [0x3000 - 0x3fff];
"cntl", which represents the area from [0x4000 - 0x4fff];
"signal1" which begins at offset 0x14000; "signal2" which
begins at offset 0x1c000.
The signal1 & signal2 files may be mmap()'d by regular user
processes. The cntl and mfc file, on the other hand, may
only be accessed if the owning process has CAP_SYS_RAWIO,
because they have the potential to confuse the kernel
with regard to parallel access to the same files with
regular file operations: the kernel always holds a spinlock
when accessing registers in these areas to serialize them,
which can not be guaranteed with user mmaps,
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
|
|
This patch adds a new file called 'mfc' to each spufs directory.
The file accepts DMA commands that are a subset of what would
be legal DMA commands for problem state register access. Upon
reading the file, a bitmask is returned with the completed
tag groups set.
The file is meant to be used from an abstraction in libspe
that is added by a different patch.
From the kernel perspective, this means a process can now
offload a memory copy from or into an SPE local store
without having to run code on the SPE itself.
The transfer will only be performed while the SPE is owned
by one thread that is waiting in the spu_run system call
and the data will be transferred into that thread's
address space, independent of which thread started the
transfer.
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
|
|
When spu_activate fails in spu_acquire_runnable, the
state must still be SPU_STATE_SAVED, we were
incorrectly setting it to SPU_STATE_RUNNABLE.
Signed-off-by: Arnd Bergmann <arndb@de.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
|
|
During an earlier cleanup, we lost the serialization
of multiple spu_run calls performed on the same
spu_context. In order to get this back, introduce a
mutex in the spu_context that is held inside of spu_run.
Noticed by Al Viro.
Signed-off-by: Arnd Bergmann <arndb@de.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
|
|
We need to check for validity of owner under down_write,
down_read is not enough.
Noticed by Al Viro.
Signed-off-by: Arnd Bergmann <arndb@de.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
|
|
This patch reduces lock complexity of SPU scheduler, particularly
for involuntary preemptive switches. As a result the new code
does a better job of mapping the highest priority tasks to SPUs.
Lock complexity is reduced by using the system default workqueue
to perform involuntary saves. In this way we avoid nasty lock
ordering problems that the previous code had. A "minimum timeslice"
for SPU contexts is also introduced. The intent here is to avoid
thrashing.
While the new scheduler does a better job at prioritization it
still does nothing for fairness.
From: Mark Nutter <mnutter@us.ibm.com>
Signed-off-by: Arnd Bergmann <arndb@de.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
|
|
This patch makes it easier to preempt an SPU context by
having the scheduler hold ctx->state_sema for much shorter
periods of time.
As part of this restructuring, the control logic for the "run"
operation is moved from arch/ppc64/kernel/spu_base.c to
fs/spufs/file.c. Of course the base retains "bottom half"
handlers for class{0,1} irqs. The new run loop will re-acquire
an SPU if preempted.
From: Mark Nutter <mnutter@us.ibm.com>
Signed-off-by: Arnd Bergmann <arndb@de.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
|
|
This adds a scheduler for SPUs to make it possible to use
more logical SPUs than physical ones are present in the
system.
Currently, there is no support for preempting a running
SPU thread, they have to leave the SPU by either triggering
an event on the SPU that causes it to return to the
owning thread or by sending a signal to it.
This patch also adds operations that enable accessing an SPU
in either runnable or saved state. We use an RW semaphore
to protect the state of the SPU from changing underneath
us, while we are holding it readable. In order to change
the state, it is acquired writeable and a context save
or restore is executed before downgrading the semaphore
to read-only.
From: Mark Nutter <mnutter@us.ibm.com>,
Uli Weigand <Ulrich.Weigand@de.ibm.com>
Signed-off-by: Arnd Bergmann <arndb@de.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
|
|
Add some infrastructure for saving and restoring the context of an
SPE. This patch creates a new structure that can hold the whole
state of a physical SPE in memory. It also contains code that
avoids races during the context switch and the binary code that
is loaded to the SPU in order to access its registers.
The actual PPE- and SPE-side context switch code are two separate
patches.
Signed-off-by: Arnd Bergmann <arndb@de.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
|
|
This is the current version of the spu file system, used
for driving SPEs on the Cell Broadband Engine.
This release is almost identical to the version for the
2.6.14 kernel posted earlier, which is available as part
of the Cell BE Linux distribution from
http://www.bsc.es/projects/deepcomputing/linuxoncell/.
The first patch provides all the interfaces for running
spu application, but does not have any support for
debugging SPU tasks or for scheduling. Both these
functionalities are added in the subsequent patches.
See Documentation/filesystems/spufs.txt on how to use
spufs.
Signed-off-by: Arnd Bergmann <arndb@de.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
|