David S. Miller [Mon, 29 Aug 2005 19:46:07 +0000 (12:46 -0700)]
[SPARC64]: Make debugging spinlocks usable again.
When the spinlock routines were moved out of line into
kernel/spinlock.c this made it so that the debugging
spinlocks record lock acquisition program counts in the
kernel/spinlock.c functions not in their callers.
This makes the debugging info kind of useless.
So record the correct caller's program counter and
now this feature is useful once more.
Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Mon, 29 Aug 2005 19:45:11 +0000 (12:45 -0700)]
[SPARC64]: Revamp Spitfire error trap handling.
Current uncorrectable error handling was poor enough
that the processor could just loop taking the same
trap over and over again. Fix things up so that we
at least get a log message and perhaps even some register
state.
In the process, much consolidation became possible,
particularly with the correctable error handler.
Prefix assembler and C function names with "spitfire"
to indicate that these are for Ultra-I/II/IIi/IIe only.
More work is needed to make these routines robust and
featureful to the level of the Ultra-III error handlers.
Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Mon, 29 Aug 2005 19:44:40 +0000 (12:44 -0700)]
[SPARC64]: Fix trap state reading for instruction_access_exception.
1) Read ASI_IMMU SFSR not ASI_DMMU.
2) IMMU has no SFAR, read TPC instead
3) Delete old and incorrect comment about the DTLB protection
trap having a dependency on the SFSR contents in order to
function correctly
Signed-off-by: David S. Miller <davem@davemloft.net>
Al Viro [Sun, 28 Aug 2005 02:19:14 +0000 (03:19 +0100)]
[PATCH] mod_devicetable.h fixes
* ieee1394_device_id has kernel_ulong_t field after an odd number of
__u32 ones. Since mod_devicetable.h is included both from kernel and
from host build helper, we may be in trouble if we are building on
32bit host for 64bit target - userland sees unsigned long long,
kernel sees unsigned long and while their sizes match, alignments
might not. Fixed by forcing alignment. Fortunately, almost nobody
else needs that - the rest of such fields is naturally aligned as it
is.
* of_device_id has void * in it. Host userland helpers need
kernel_ulong_t instead, since their void * might have nothing to do
with the kernel one. Fixed in the same way it's done for similar
problems in pcmcia_device_id (ifdef __KERNEL__).
* pcmcia_device_id has the same problem as ieee1394_device_id. Fixed
the same way.
Signed-off-by: Al Viro <viro@parcelfarce.linux.theplanet.co.uk> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Steven Rostedt [Mon, 29 Aug 2005 15:44:09 +0000 (11:44 -0400)]
[PATCH] convert signal handling of NODEFER to act like other Unix boxes.
It has been reported that the way Linux handles NODEFER for signals is
not consistent with the way other Unix boxes handle it. I've written a
program to test the behavior of how this flag affects signals and had
several reports from people who ran this on various Unix boxes,
confirming that Linux seems to be unique on the way this is handled.
The way NODEFER affects signals on other Unix boxes is as follows:
1) If NODEFER is set, other signals in sa_mask are still blocked.
2) If NODEFER is set and the signal is in sa_mask, then the signal is
still blocked. (Note: this is the behavior of all tested but Linux _and_
NetBSD 2.0 *).
The way NODEFER affects signals on Linux:
1) If NODEFER is set, other signals are _not_ blocked regardless of
sa_mask (Even NetBSD doesn't do this).
2) If NODEFER is set and the signal is in sa_mask, then the signal being
handled is not blocked.
The patch converts signal handling in all current Linux architectures to
the way most Unix boxes work.
Unix boxes that were tested: DU4, AIX 5.2, Irix 6.5, NetBSD 2.0, SFU
3.5 on WinXP, AIX 5.3, Mac OSX, and of course Linux 2.6.13-rcX.
* NetBSD was the only other Unix to behave like Linux on point #2. The
main concern was brought up by point #1 which even NetBSD isn't like
Linux. So with this patch, we leave NetBSD as the lonely one that
behaves differently here with #2.
David Gibson [Thu, 11 Aug 2005 06:55:21 +0000 (16:55 +1000)]
[PATCH] Dynamic hugepage addresses for ppc64
Paulus, I think this is now a reasonable candidate for the post-2.6.13
queue.
Relax address restrictions for hugepages on ppc64
Presently, 64-bit applications on ppc64 may only use hugepages in the
address region from 1-1.5T. Furthermore, if hugepages are enabled in
the kernel config, they may only use hugepages and never normal pages
in this area. This patch relaxes this restriction, allowing any
address to be used with hugepages, but with a 1TB granularity. That
is if you map a hugepage anywhere in the region 1TB-2TB, that entire
area will be reserved exclusively for hugepages for the remainder of
the process's lifetime. This works analagously to hugepages in 32-bit
applications, where hugepages can be mapped anywhere, but with 256MB
(mmu segment) granularity.
This patch applies on top of the four level pagetable patch
(http://patchwork.ozlabs.org/linuxppc64/patch?id=1936).
Signed-off-by: David Gibson <dwg@au1.ibm.com> Signed-off-by: Paul Mackerras <paulus@samba.org>
[PATCH] ppc64: unflatten_device_tree() should check if lmb_alloc() fails
unflatten_device_tree() doesn't check if lmb_alloc() succeeds or not, it
should. All it can do is panic, but at least there's an error message
(assuming you have some sort of console at that point).
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
arch/ppc64/kernel/prom.c | 9 +++++++--
1 files changed, 7 insertions(+), 2 deletions(-) Signed-off-by: Paul Mackerras <paulus@samba.org>
Olaf Hering [Thu, 4 Aug 2005 17:26:42 +0000 (19:26 +0200)]
[PATCH] ppc64: allow xmon=off
If both CONFIG_XMON and CONFIG_XMON_DEFAULT is enabled in the .config,
there is no way to disable xmon again. setup_system calls first xmon_init,
later parse_early_param. So a new 'xmon=off' cmdline option will do the right
thing.
Signed-off-by: Olaf Hering <olh@suse.de> Signed-off-by: Paul Mackerras <paulus@samba.org>
We can now remove CONFIG_MSCHUNKS as it doesn't do anything interesting
anymore.
The only macro in abs_addr.h which is called by non-iSeries code is
phys_to_abs(), so remove the other dummy implementations, and we add a
firmware feature check to phys_to_abs().
Signed-off-by: Michael Ellerman <michael@ellerman.id.au> Signed-off-by: Paul Mackerras <paulus@samba.org>
lmb_phys_mem_size() can always return lmb.memory.size, as long as it's called
after lmb_analyze(), which it is. There's no need to recalculate the size on
every call.
lmb_analyze() was calculating a few things we then threw away, so just don't
calculate them to start with.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au> Signed-off-by: Paul Mackerras <paulus@samba.org>
abs_to_phys() is a macro that turns out to do nothing, and also has the
unfortunate property that it's not the inverse of phys_to_abs() on iSeries.
The following is for my benefit as much as everyone else.
With CONFIG_MSCHUNKS enabled, the lmb code is changed such that it keeps
a physbase variable for each lmb region. This is used to take the possibly
discontiguous lmb regions and present them as a contiguous address space
beginning from zero.
In this context each lmb region's base address is its "absolute" base
address, and its physbase is it's "physical" address (from Linux's point of
view). The abs_to_phys() macro does the mapping from "absolute" to "physical".
Note: This is not related to the iSeries mapping of physical to absolute
(ie. Hypervisor) addresses which is maintained with the msChunks structure.
And the msChunks structure is not controlled via CONFIG_MSCHUNKS.
Once upon a time you could compile for non-iSeries with CONFIG_MSCHUNKS
enabled. But these days CONFIG_MSCHUNKS depends on CONFIG_PPC_ISERIES, so
for non-iSeries code abs_to_phys() is a no-op.
On iSeries we always have one lmb region which spans from 0 to
systemcfg->physicalMemorySize (arch/ppc64/kernel/iSeries_setup.c line 383).
This region has a base (ie. absolute) address of 0, and a physbase address
of 0 (as calculated in lmb_analyze() (arch/ppc64/kernel/lmb.c line 144)).
On iSeries, abs_to_phys(aa) is defined as lmb_abs_to_phys(aa), which finds
the lmb region containing aa (and there's only one, ie. 0), and then does:
[PATCH] ppc64: Remove redundant use of pointers in lmb code
The lmb code is all written to use a pointer to an lmb struct. But it's always
the same lmb struct, called "lmb". So we take the address of lmb, call it
_lmb and then start using _lmb->foo everywhere, which is silly.
This patch removes the _lmb pointers and replaces them with direct references
to the one "lmb" struct. We do the same for some _mem and _rsv pointers which
point to lmb.memory and lmb.reserved respectively.
This patch looks quite busy, but it's basically just:
s/_lmb->/lmb./g
s/_mem->/lmb.memory./g
s/_rsv->/lmb.reserved./g
s/_rsv/&lmb.reserved/g
s/mem->/lmb.memory./g
Signed-off-by: Michael Ellerman <michael@ellerman.id.au> Signed-off-by: Paul Mackerras <paulus@samba.org>
Chunks are 256KB, so use constants for the size/shift/mask, rather than
getting them from the msChunks struct. The iSeries debugger (??) might still
need access to the values in the msChunks struct, so we keep them around
for now, but set them from the constant values.
Replace msChunks_entry typedef with regular u32.
Simplify msChunks_alloc() to manipulate klimit directly, rather than via
a parameter.
Move msChunks_alloc() and msChunks into iSeries_setup.c, as that's where
they're used.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au> Signed-off-by: Paul Mackerras <paulus@samba.org>
[PATCH] ppc64: Remove PTRRELOC() from msChunks code
The msChunks code was written to work on pSeries, but now it's only used on
iSeries. This means there's no need to do PTRRELOC anymore, so remove it all.
A few places were getting "extern reloc_offset()" from abs_addr.h, move it
into system.h instead.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au> Signed-off-by: Paul Mackerras <paulus@samba.org>
[PATCH] ppc64: remove firmware features from cpu_spec
The firmware_features field of struct cpu_spec should really be a separate
variable as the firmware features do not depend on the chip and the
bitmask is constructed independently. By removing it, we save 112 bytes
from the cpu_specs array and we access the bitmask directly instead of via
the cur_cpu_spec pointer.
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au> Signed-off-by: Paul Mackerras <paulus@samba.org>
David Gibson [Fri, 19 Aug 2005 04:52:32 +0000 (14:52 +1000)]
[PATCH] Move variables in ppc64 head.S from .data to .bss
The ppc64 head.S defines several zero-initialized structures, such as
the empty_zero_page and the kernel top-level pagetable. Currently
they are defined to be in the data section. However, they're not used
until after the bss is cleared, so this patch moves them to the bss,
saving two and a half pages from the vmlinux.
Signed-off-by: David Gibson <dwg@au1.ibm.com> Signed-off-by: Paul Mackerras <paulus@samba.org>
David Gibson [Fri, 19 Aug 2005 04:52:32 +0000 (14:52 +1000)]
[PATCH] Remove unneeded #defines in head.S
arch/ppc64/kernel/head.S #defines SECONDARY_PROCESSORS then has some
#ifdefs based on it. Whatever purpose this had is long lost, this
patch removes it.
Likewise, head.S defines H_SET_ASR, which is now defined, along with
other hypervisor call numbers in hvcall.h. This patch deletes it, as
well, from head.S.
Signed-off-by: David Gibson <dwg@au1.ibm.com> Signed-off-by: Paul Mackerras <paulus@samba.org>
David Gibson [Fri, 19 Aug 2005 04:52:32 +0000 (14:52 +1000)]
[PATCH] Fix apparent code overlap in ppc64 head.S
An #if/#else construct near the top of ppc64's head.S appears to
create overlapping sections of code for iSeries and pSeries (i.e. one
thing on iSeries and something different in the same place on
pSeries). In fact, checking the various absolute offsets, it doesn't.
This patch unravels the #ifdefs to make it more obvious what's going
on. This accomplishes another microstep towards a single kernel image
which can boot both iSeries and pSeries.
Signed-off-by: David Gibson <dwg@au1.ibm.com> Signed-off-by: Paul Mackerras <paulus@samba.org>
David Gibson [Fri, 19 Aug 2005 04:52:31 +0000 (14:52 +1000)]
[PATCH] Remove general use functions from head.S
As well as the interrupt vectors and initialization code, head.S
contains several asm functions which are used during runtime. This
patch moves these to misc.S, a more sensible location for random asm
support code. A couple The functions moved are:
disable_kernel_fp
giveup_fpu
disable_kernel_altivec
giveup_altivec
__setup_cpu_power3 (empty function)
Signed-off-by: David Gibson <dwg@au1.ibm.com> Signed-off-by: Paul Mackerras <paulus@samba.org>
David Gibson [Fri, 19 Aug 2005 04:52:31 +0000 (14:52 +1000)]
[PATCH] Change address of ppc64 initial segment table
On ppc64 machines with segment tables, CPU0's segment table is at a
fixed address, currently 0x9000. This patch moves it to the free
space at 0x6000, just below the fwnmi data area. This saves 8k of
space in vmlinux and the runtime kernel image.
Signed-off-by: David Gibson <dwg@au1.ibm.com> Signed-off-by: Paul Mackerras <paulus@samba.org>
David Gibson [Fri, 19 Aug 2005 04:52:31 +0000 (14:52 +1000)]
[PATCH] Move iSeries and common vectors into unused space in head.S
In the ppc64 kernel head.S there is currently quite a lot of unused
space between the naca (at fixed address 0x4000) and the fwnmi data
area (at fixed address 0x7000). This patch moves various exception
vectors and support code into this region to use the wasted space.
The functions load_up_fpu and load_up_altivec are moved down as well,
since they are essentially continuations of the fp_unavailable_common
and altivec_unavailable_common vectors, respectively.
Likewise, the fwnmi vectors themselves are moved down into this area,
because while the location of the fwnmi data area is fixed by the RPA,
the vectors themselves can be anywhere sufficiently low.
Signed-off-by: David Gibson <dwg@au1.ibm.com> Signed-off-by: Paul Mackerras <paulus@samba.org>
David Gibson [Fri, 19 Aug 2005 04:52:31 +0000 (14:52 +1000)]
[PATCH] Remove NACA fixed address constraint
Comments in head.S suggest that the iSeries naca has a fixed address,
because tools expect to find it there. The only tool which appears to
access the naca is addRamDisk, but both the in-kernel version and the
version used in RHEL and SuSE in fact locate the NACA the same way as
the hypervisor does, by following the pointer in the hvReleaseData
structure.
Since the requirement for a fixed address seems to be obsolete, this
patch removes the naca from head.S and replaces it with a normal C
initializer.
For good measure, it removes an old version of addRamDisk.c which was
sitting, unused, in the ppc32 tree.
Signed-off-by: David Gibson <dwg@au1.ibm.com> Signed-off-by: Paul Mackerras <paulus@samba.org>
Frank Rowand [Tue, 28 Jun 2005 23:48:04 +0000 (16:48 -0700)]
[PATCH] ppc64: change duplicate Kconfig menu "General setup" to "Bus Options"
arch/ppc64/Kconfig defines a "General setup" menu, but also sources
init/Kconfig which also defines a "General setup" menu. Both of these
menus appear at the top level of make menuconfig. Having two menus with
the same name is confusing. This patch renames the ppc64/Kconfig menu to
be "Bus Options" and moves options in this menu which are not bus related
to the end of the "Platform support" menu.
There are many variations among architectures on the exact naming of the
"Bus Options" menu. I chose to use the simplest one, which is also used
in arch/ppc/Kconfig.
Signed-off-by: Frank Rowand <frowand@mvista.com> Signed-off-by: Paul Mackerras <paulus@samba.org>
Jake Moilanen [Thu, 9 Jun 2005 14:31:12 +0000 (09:31 -0500)]
[PATCH] ppc64: PCI device-node failure detection
OpenFirmware marks devices as failed in the device-tree when a hardware
problem is detected. The kernel needs to fail config reads/writes to
prevent a kernel crash when incorrect data is read.
This patch validates that the device-node is not marked "fail" when
config space reads/writes are attempted.
Signed-off-by: Jake Moilanen <moilanen@austin.ibm.com> Signed-off-by: Paul Mackerras <paulus@samba.org>
This patch updates the format of the flattened device-tree passed
between the boot trampoline and the kernel to support a more compact
representation, for use by embedded systems mostly.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
David Gibson [Fri, 5 Aug 2005 09:39:06 +0000 (19:39 +1000)]
[PATCH] Four level pagetables for ppc64
Implement 4-level pagetables for ppc64
This patch implements full four-level page tables for ppc64, thereby
extending the usable user address range to 44 bits (16T).
The patch uses a full page for the tables at the bottom and top level,
and a quarter page for the intermediate levels. It uses full 64-bit
pointers at every level, thus also increasing the addressable range of
physical memory. This patch also tweaks the VSID allocation to allow
matching range for user addresses (this halves the number of available
contexts) and adds some #if and BUILD_BUG sanity checks.
Signed-off-by: David Gibson <dwg@au1.ibm.com> Signed-off-by: Paul Mackerras <paulus@samba.org>
Olaf Hering [Mon, 8 Aug 2005 03:24:38 +0000 (13:24 +1000)]
[PATCH] ppc64: make arch/ppc64/boot standalone
Make the bootheader for ppc64 independent from kernel and libc headers.
* add -nostdinc -isystem $gccincludes to not include libc headers
* declare all functions in header files, also the stuff from string.S
* declare some functions static
* use stddef.h to get size_t (hopefully ok)
* remove ppc32-types.h, only elf.h used the __NN types
With further modifications by Paul Mackerras and Stephen Rothwell.
Signed-off-by: Olaf Hering <olh@suse.de> Signed-off-by: Paul Mackerras <paulus@samba.org>
Jeff Garzik [Mon, 29 Aug 2005 00:18:39 +0000 (20:18 -0400)]
[libata] license change, other bits
- changes license of all code from OSL+GPL to plain ole GPL
- except for NVIDIA, who hasn't yet responded about sata_nv
- copyright holders were already contacted privately
- adds info in each driver about where hardware/protocol docs may be
obtained
- where I have made major contributions, updated copyright dates
Paul Mackerras [Sat, 27 Aug 2005 23:40:01 +0000 (09:40 +1000)]
[PATCH] Remove race between con_open and con_close
[ Same race and same patch also by Steven Rostedt <rostedt@goodmis.org> ]
I have a laptop (G3 powerbook) which will pretty reliably hit a race
between con_open and con_close late in the boot process and oops in
vt_ioctl due to tty->driver_data being NULL.
What happens is this: process A opens /dev/tty6; it comes into
con_open() (drivers/char/vt.c) and assign a non-NULL value to
tty->driver_data. Then process A closes that and concurrently process
B opens /dev/tty6. Process A gets through con_close() and clears
tty->driver_data, since tty->count == 1. However, before process A
can decrement tty->count, we switch to process B (e.g. at the
down(&tty_sem) call at drivers/char/tty_io.c line 1626).
So process B gets to run and comes into con_open with tty->count == 2,
as tty->count is incremented (in init_dev) before con_open is called.
Because tty->count != 1, we don't set tty->driver_data. Then when the
process tries to do anything with that fd, it oopses.
The simple and effective fix for this is to test tty->driver_data
rather than tty->count in con_open. The testing and setting of
tty->driver_data is serialized with respect to the clearing of
tty->driver_data in con_close by the console_sem. We can't get a
situation where con_open sees tty->driver_data != NULL and then
con_close on a different fd clears tty->driver_data, because
tty->count is incremented before con_open is called. Thus this patch
eliminates the race, and in fact with this patch my laptop doesn't
oops.
Signed-off-by: Paul Mackerras <paulus@samba.org>
[ Same patch Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
in http://marc.theaimsgroup.com/?l=linux-kernel&m=112450820432121&w=2 ] Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Jan Blunck [Sat, 27 Aug 2005 18:07:52 +0000 (11:07 -0700)]
[PATCH] sg.c: fix a memory leak in devices seq_file implementation
I know that scsi procfs is legacy code but this is a fix for a memory leak.
While reading through sg.c I realized that the implementation of
/proc/scsi/sg/devices with seq_file is leaking memory due to freeing the
pointer returned by the next() iterator method. Since next() might return
NULL or an error this is wrong. This patch fixes it through using the
seq_files private field for holding the reference to the iterator object.
Here is a small bash script to trigger the leak. Use slabtop to watch
the size-32 usage grow and grow.
#!/bin/sh
while true; do
cat /proc/scsi/sg/devices > /dev/null
done
Signed-off-by: Jan Blunck <j.blunck@tu-harburg.de> Acked-by: James Bottomley <James.Bottomley@steeleye.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
[PATCH] fix for race problem in DVB USB drivers (dibusb)
Fixed race between submitting streaming URBs in the driver and starting
the actual transfer in hardware (demodulator and USB controller) which
sometimes lead to garbled data transfers. URBs are now submitted first,
then the transfer is enabled. Dibusb devices and clones are now fully
functional again.
Signed-off-by: Patrick Boettcher <pb@linuxtv.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Francois Romieu [Tue, 23 Aug 2005 23:14:23 +0000 (01:14 +0200)]
[PATCH] r8169: avoid conflict between revisions 2 and 3 of the Linksys EG1032
Both revisions share the same PCI device ID and vendor ID but revision 2
of the device uses SysKonnect's chipset whereas revision 3 of the device
uses Realtek's 8169 chipset.
Credit goes to Christiaan Lutzer <mythtv.lutzer@gmail.com> for reporting
the issue and giving the actual value for the different revisions.
Signed-off-by: Francois Romieu <romieu@fr.zoreil.com> Signed-off-by: Jeff Garzik <jgarzik@pobox.com>
Ralf Baechle [Wed, 24 Aug 2005 17:01:33 +0000 (18:01 +0100)]
[PATCH] Fix 6pack setting of MAC address
Don't check type of sax25_family; dev_set_mac_address has already done
that before and anyway, the type to check against would have been
ARPHRD_AX25. We only got away because AF_AX25 and ARPHRD_AX25 both happen
to be defined to the same value.
Don't check sax25_ndigis either; it's value is insignificant for the
purpose of setting the MAC address and the check has shown to break
some application software for no good reason.
Signed-off-by: Ralf Baechle DL5RB <ralf@linux-mips.org> Signed-off-by: Jeff Garzik <jgarzik@pobox.com>
Roland Dreier [Thu, 25 Aug 2005 20:40:04 +0000 (13:40 -0700)]
[PATCH] IB: move include files to include/rdma
Move the InfiniBand headers from drivers/infiniband/include to include/rdma.
This allows InfiniBand-using code to live elsewhere, and lets us remove the
ugly EXTRA_CFLAGS include path from the InfiniBand Makefiles.
Currently we may have work scheduled in default kernel workqueue when
the device is going down. The device could get freed before this
workqueue gets serviced. I am actually seeing this causing system
hangs.
The following patch fixes this by using ipoib_workqueue which gets
flushed when the device is going down.
Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
Sean Hefty [Fri, 19 Aug 2005 20:46:34 +0000 (13:46 -0700)]
[PATCH] IB: fix userspace CM deadlock
Fix deadlock condition resulting from trying to destroy a cm_id
from the context of a CM thread. The synchronization around the
ucm context structure is simplified as a result, and some simple
code cleanup is included.
Signed-off-by: Sean Hefty <sean.hefty@intel.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
Roland Dreier [Fri, 19 Aug 2005 19:03:17 +0000 (12:03 -0700)]
[PATCH] IPoIB: Set full membership bit in P_Keys
Always make sure that the full membership bit is set in the P_Keys
that IPoIB uses. This makes sure that all hosts join the correct
multicast groups so that hosts that are partial partition members
can talk to the rest of the network.
Roland Dreier [Fri, 19 Aug 2005 17:36:11 +0000 (10:36 -0700)]
[PATCH] IB/mthca: Handle context tables smaller than our chunk size
When creating a table in context memory where the table is smaller
than our chunk size, we don't want to allocate and map a full chunk.
Instead, allocate just enough memory to cover the table.
This can be pretty simple because all tables are a power-of-2 size, so
either the table is a multiple of the chunk size, or it's smaller than
one chunk.
Roland Dreier [Fri, 19 Aug 2005 17:33:35 +0000 (10:33 -0700)]
[PATCH] IB/mthca: Move WQE structures into their own header
Move the definitions of the WQE structures from mthca_qp.c into
mthca_wqe.h, so that we'll be able to share them when we add the
SRQ code in mthca_srq.c.
Roland Dreier [Fri, 19 Aug 2005 16:19:05 +0000 (09:19 -0700)]
[PATCH] IB/mthca: Simplify handling of completions with error
Mem-free HCAs never generate error CQEs that complete multiple WQEs,
so just skip the call to mthca_free_err_wqe() for them rather than
having logic to handle the mem-free case in mthca_free_err_wqe().
Roland Dreier [Thu, 18 Aug 2005 20:39:31 +0000 (13:39 -0700)]
[PATCH] IB/mthca: Factor out common queue alloc code
Clean up the allocation of memory for queues by factoring out the
common code into mthca_buf_alloc() and mthca_buf_free(). Now CQs and
QPs share the same queue allocation code, which we'll also use for SRQs.
Roland Dreier [Thu, 18 Aug 2005 19:24:13 +0000 (12:24 -0700)]
[PATCH] IB: userspace SRQ support
Add SRQ support to userspace verbs module. This adds several commands
and associated structures, but it's OK to do this without bumping the
ABI version because the commands are added at the end of the list so
they don't change the existing numbering. There are two cases to
worry about:
1. New kernel, old userspace. This is OK because old userspace simply
won't try to use the new SRQ commands. None of the old commands are
changed.
2. Old kernel, new userspace. This works perfectly as long as
userspace doesn't try to use SRQ commands. If userspace tries to
use SRQ commands, it will get EINVAL, which is perfectly
reasonable: the kernel doesn't support SRQs, so we couldn't do any
better.
Roland Dreier [Thu, 18 Aug 2005 19:14:11 +0000 (12:14 -0700)]
[PATCH] IB/mthca: Report correct max_msg_sz
Set the max_msg_sz port property correctly in mthca's port_query
function. Also zero out the attr struct so that we don't leave
any other members uninitialized.
Roland Dreier [Wed, 17 Aug 2005 14:39:10 +0000 (07:39 -0700)]
[PATCH] IB/mthca: Use correct port width capability value
When we call the INIT_IB firmware command to bring up a port, use
the actual port width capability returned by the QUERY_DEV_LIM
command instead of always trying to enable both 1X and 4X. This
fixes breakage seen when the firmware is build to allow 4X only.
Roland Dreier [Mon, 15 Aug 2005 14:35:16 +0000 (07:35 -0700)]
[PATCH] IB: unmap FMRs when destroying FMR pool
Make sure that all FMRs are unmapped before we deallocate them so that
we don't leak references to our protection domain when destroying an
FMR pool. (Bug reported by Guy German <guyg@voltaire.com>)
projects / linux-2.6 / log