Paul Mundt [Wed, 27 Sep 2006 09:17:31 +0000 (18:17 +0900)]
sh: machvec rework.
Some more machvec overhauling and setup code cleanup. Kill off
get_system_type() and platform_setup(), we can do these both
through the machvec. While we're add it, kill off more useless
mach.c's and drop some legacy cruft from setup.c.
Paul Mundt [Wed, 27 Sep 2006 08:53:59 +0000 (17:53 +0900)]
watchdog: Add a simple mmap() stub for shwdt.
In some applications people have expressed a need for an mmap() method,
so we implement a simple stub for this that maps back a page with the
counter in it.
Paul Mundt [Wed, 27 Sep 2006 08:07:07 +0000 (17:07 +0900)]
sh: Fixup __strnlen_user() behaviour.
Drop TIF_USERSPACE and add addr_limit to the thread_info struct.
Subsequently, use that for address checking in strnlen_user() to
ward off bogus -EFAULTs.
Make __strnlen_user() return 0 on exception, rather than -EFAULT.
Paul Mundt [Wed, 27 Sep 2006 07:45:22 +0000 (16:45 +0900)]
sh: Inhibit mapping PCI apertures through page tables.
Inhibit mapping through page tables in __ioremap() for PCI memory
apertures on SH7751 and SH7780-style PCI controllers, translation is
not possible for these areas. For other users that map a small window
in P1/P2 space, ioremap() traps that already, and should never make
it to __ioremap().
Paul Mundt [Wed, 27 Sep 2006 07:43:28 +0000 (16:43 +0900)]
sh: Consolidated SH7751/SH7780 PCI support.
This cleans up quite a lot of the PCI mess that we
currently have, and attempts to consolidate the
duplication in the SH7780 and SH7751 PCI controllers.
Paul Mundt [Wed, 27 Sep 2006 07:32:13 +0000 (16:32 +0900)]
serial: Rework sh-sci for driver model.
sh-sci was turning in to an unmaintainable mess, especially with
regards to the port list. This cleans it up quite a bit, and
switches over to a platform device model where subtypes will
register their port list individually in their setup code.
Paul Mundt [Wed, 27 Sep 2006 07:24:55 +0000 (16:24 +0900)]
sh: kgdb stub cleanups.
Some kgdb cleanup. Move hexchars/highhex/lowhex to the header, so it can
be reused by sh-sci. Also drop silly ctrl_inl/outl() overloading being
done by the kgdb stub.
Paul Mundt [Wed, 27 Sep 2006 07:17:17 +0000 (16:17 +0900)]
sh: Free up some and document PTEL flags.
Drop _PAGE_SHARED/_PAGE_U0_SHARED and document Linux PTE encodings in
the PTEL value. Preserve the swap cache entry encoding semantics for
now, though it will need rework to free up _PAGE_WT from _PAGE_FILE.
Paul Mundt [Wed, 27 Sep 2006 06:49:57 +0000 (15:49 +0900)]
sh: Store Queue API rework.
Rewrite the store queue API for a per-cpu interface in the driver
model. The old miscdevice is dropped, due to TASK_SIZE limitations,
and no one was using it anyways.
Carve up and allocate store queue space with a bitmap, back sq
mapping objects with a slab cache, and let userspace worry about
its own prefetching.
Paul Mundt [Wed, 27 Sep 2006 06:16:42 +0000 (15:16 +0900)]
sh: ioremap() overhaul.
ioremap() overhaul. Add support for transparent PMB mapping, get rid of
p3_ioremap(), etc. Also drop ioremap() and iounmap() routines from the
machvec, as everyone can use the generic ioremap() API instead. For PCI
memory apertures and other special cases, use the pci_iomap() API, as
boards are already required to get the mapping right there.
Paul Mundt [Wed, 27 Sep 2006 06:13:36 +0000 (15:13 +0900)]
sh: page table alloc cleanups and page fault optimizations.
Cleanup of page table allocators, using generic folded PMD and PUD
helpers. TLB flushing operations are moved to a more sensible spot.
The page fault handler is also optimized slightly, we no longer waste
cycles on IRQ disabling for flushing of the page from the ITLB, since
we're already under CLI protection by the initial exception handler.
Paul Mundt [Wed, 27 Sep 2006 05:57:44 +0000 (14:57 +0900)]
sh: Add control register barriers.
Currently when making changes to control registers, we
typically need some time for changes to take effect (8
nops, generally). However, for sh4a we simply need to
do an icbi..
This is a simple patch for implementing a general purpose
ctrl_barrier() which functions as a control register write
barrier. There's some additional documentation in the patch
itself, but it's pretty self explanatory.
There were also some places where we were not doing the
barrier, which didn't seem to have any adverse effects on
legacy parts, but certainly did on sh4a. It's safer to have
the barrier in place for legacy parts as well in these cases,
though this does make flush_tlb_all() more expensive (by an
order of 8 nops). We can ifdef around the flush_tlb_all()
case for now if it's clear that all legacy parts won't have
a problem with this.
sh: Fix TCP payload csum bug in csum_partial_copy_generic().
There's a bug in the Hitachi SuperH csum_partial_copy_generic()
implementation. If the supplied length is 1 (and several alignment
conditions are met), the function immediately branches to label 4.
However, the assembly at label 4 expects the length to be stored in
register r2. Since this has not occurred, subsequent behavior is
undefined.
This can cause bad payload checksums in TCP connections.
I've fixed the problem by initializing register r2 prior to the branch
instruction.
Signed-off-by: Ollie Wild <aaw@rincewind.tv> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
Paul Mundt [Wed, 27 Sep 2006 05:38:02 +0000 (14:38 +0900)]
sh: Fix fatal oops in copy_user_page() on sh4a (SH7780).
We had a pretty interesting oops happening, where copy_user_page()
was down()'ing p3map_sem[] with a bogus offset (particularly, an
offset that hadn't been initialized with sema_init(), due to the
mismatch between cpu_data->dcache.n_aliases and what was assumed
based off of the old CACHE_ALIAS value).
Luckily, spinlock debugging caught this for us, and so we drop
the old hardcoded CACHE_ALIAS for sh4 completely and rely on the
run-time probed cpu_data->dcache.alias_mask. This in turn gets
the p3map_sem[] index right, and everything works again.
While we're at it, also convert to 4-level page tables..
Paul Mundt [Wed, 27 Sep 2006 05:26:53 +0000 (14:26 +0900)]
sh: earlyprintk= support and cleanups.
Allow multiple early printk consoles via earlyprintk=.
With this change earlyprintk is no longer enabled by default,
it must be specified on the kernel command line. Optionally
with ,keep to prevent unreg by tty_io.
Richard Curnow [Wed, 27 Sep 2006 05:09:26 +0000 (14:09 +0900)]
sh: Optimized cache handling for SH-4/SH-4A caches.
This reworks some of the SH-4 cache handling code to more easily
accomodate newer-style caches (particularly for the > direct-mapped
case), as well as optimizing some of the old code.
Signed-off-by: Richard Curnow <richard.curnow@st.com> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
Paul Mundt [Wed, 27 Sep 2006 04:11:57 +0000 (13:11 +0900)]
sh: hugetlb updates.
For some of the larger sizes we permitted spanning pages
across several PTEs, but this turned out to not be generally
useful. This reverts the sh hugetlbpage interface to something
more sensible using huge pages at single PTE granularity.
Paul Mundt [Wed, 27 Sep 2006 03:31:01 +0000 (12:31 +0900)]
sh: Various cosmetic cleanups.
We had quite a bit of whitespace damage, clean most of it up..
Signed-off-by: Stuart Menefy <stuart.menefy@st.com> Signed-off-by: Arthur Othieno <a.othieno@bluewin.ch> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
Paul Mundt [Wed, 27 Sep 2006 02:40:05 +0000 (11:40 +0900)]
sh: Kill off the .stack section.
We had a special .stack section in the ld script that
was being used to position r15 initially. This is
nonsensical, as we can just use a THREAD_SIZE offset
from the init_thread_union instead (as every other arch
does).
Paul Mundt [Wed, 27 Sep 2006 02:36:10 +0000 (11:36 +0900)]
sh: Move syscall table in to syscall.S.
Move the syscall table in to its own file, as per sh64. The entry.S
bits will end up being considerably different in the sh2/sh2a cases,
so this lets us keep things in sync somewhat..
Paul Mundt [Wed, 27 Sep 2006 02:29:55 +0000 (11:29 +0900)]
sh: flush_cache_range() cleanup and optimizations.
flush_cache_range() wasn't page aligning the end of the range,
we can't assume that it will always be page aligned, and we
ended up getting unaligned faults in some rare call paths.
Additionally, we add a small optimization to just purge the
dcache entirely if the range is large enough that the page
table walking will take longer. We use an arbitrary value of
64 pages for the large range size, as per sh64.
projects / linux-beck.git / log