Roland Dreier [Mon, 27 Jun 2005 21:36:46 +0000 (14:36 -0700)]
[PATCH] IB: Fix race in sa_query
Use a copy of the id we'll return to the consumer so that we don't
dereference query->sa_query after calling send_mad(). A completion may
occur very quickly and end up freeing the query before we get to do
anything after send_mad().
Signed-off-by: Roland Dreier <rolandd@cisco.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Roland Dreier [Mon, 27 Jun 2005 21:36:45 +0000 (14:36 -0700)]
[PATCH] IB/mthca: Align FW command mailboxes to 4K
Future versions of Mellanox HCA firmware will require command mailboxes to be
aligned to 4K. Support this by using a pci_pool to allocate all mailboxes.
This has the added benefit of shrinking the source and text of mthca.
Signed-off-by: Roland Dreier <roland@topspin.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Roland Dreier [Mon, 27 Jun 2005 21:36:43 +0000 (14:36 -0700)]
[PATCH] IB/mthca: Move mthca_is_memfree checks
Make mthca_table_put() and mthca_table_put_range() NOPs if the device is not
mem-free, so that we don't have to have "if (mthca_is_memfree())" tests in the
callers of these functions. This makes our code more readable and
maintainable, and saves a couple dozen bytes of text in ib_mthca.ko as well.
Signed-off-by: Roland Dreier <roland@topspin.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Roland Dreier [Mon, 27 Jun 2005 21:36:42 +0000 (14:36 -0700)]
[PATCH] IB/mthca: Enable unreliable connected transport
Add support for unreliable connected (UC) transport to mthca driver:
- Add attributes for UC to modify QP table.
- Add support for posting UC work requests.
Signed-off-by: Roland Dreier <roland@topspin.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Roland Dreier [Mon, 27 Jun 2005 21:36:41 +0000 (14:36 -0700)]
[PATCH] IB/mthca: Set RDMA/atomic capabilities correctly
mthca apparently had the meanings of the max_rd_atomic and max_dest_rd_atomic
QP attributes backwards. max_rd_atomic limits the maximum number of
outstanding RDMA/atomic requests as an initiator (on a send queue), and
max_dest_rd_atomic specifies the resources allocated to handle RMDA/atomic
requests from the remote end of the connection. We were programming our QP
context with these values swapped.
Signed-off-by: Roland Dreier <roland@topspin.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Roland Dreier [Mon, 27 Jun 2005 21:36:40 +0000 (14:36 -0700)]
[PATCH] IB/mthca: Use dma_alloc_coherent instead of pci_alloc_consistent
Switch all allocations of coherent memory from pci_alloc_consistent() to
dma_alloc_coherent(), so that we can pass GFP_KERNEL. This should help when
the system is low on memory.
Signed-off-by: Roland Dreier <roland@topspin.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Andrea Arcangeli [Mon, 27 Jun 2005 21:36:36 +0000 (14:36 -0700)]
[PATCH] seccomp: tsc disable
I believe at least for seccomp it's worth to turn off the tsc, not just for
HT but for the L2 cache too. So it's up to you, either you turn it off
completely (which isn't very nice IMHO) or I recommend to apply this below
patch.
This has been tested successfully on x86-64 against current cogito
repository (i686 compiles so I didn't bother testing ;). People selling
the cpu through cpushare may appreciate this bit for a peace of mind.
There's no way to get any timing info anymore with this applied
(gettimeofday is forbidden of course). The seccomp environment is
completely deterministic so it can't be allowed to get timing info, it has
to be deterministic so in the future I can enable a computing mode that
does a parallel computing for each task with server side transparent
checkpointing and verification that the output is the same from all the 2/3
seller computers for each task, without the buyer even noticing (for now
the verification is left to the buyer client side and there's no
checkpointing, since that would require more kernel changes to track the
dirty bits but it'll be easy to extend once the basic mode is finished).
Eliminating a cold-cache read of the cr4 global variable will save one
cacheline during the tlb flush while making the code per-cpu-safe at the
same time. Thanks to Mikael Pettersson for noticing the tlb flush wasn't
per-cpu-safe.
The global tlb flush can run from irq (IPI calling do_flush_tlb_all) but
it'll be transparent to the switch_to code since the IPI won't make any
change to the cr4 contents from the point of view of the interrupted code
and since it's now all per-cpu stuff, it will not race. So no need to
disable irqs in switch_to slow path.
Signed-off-by: Andrea Arcangeli <andrea@cpushare.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This patch removes CONFIG_PMAC_PBOOK (PowerBook support). This is now
split into CONFIG_PMAC_MEDIABAY for the actual hotswap bay that some
powerbooks have, CONFIG_PM for power management related code, and just left
out of any CONFIG_* option for some generally useful stuff that can be used
on non-laptops as well.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The macserial driver has been obsoleted by the new pmac_zilog driver for a
while now and probably doesn't even work anymore on recent kernels. This
patch removes it.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The Power Management Unit on PowerMacs is very sensitive to timeouts during
async message exchanges. It uses rather crude protocol based on a shift
register with an interrupt and is almost continuously exchanging messages with
the host CPU on laptops.
This patch adds a routine to the open_pic driver to be able to select a PMU
driver so that it bumps it's interrupt priority to above the normal level.
This will allow PMU interrupts to occur while another interrupt is pending,
and thus reduce the risk of machine beeing abruptly shutdown by the PMU due to
a timeout in PMU communication caused by excessive interrupt latency. The
problem is very rare, and usually just doesn't happen, but it is still useful
to make things even more robust.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Marcelo Tosatti [Mon, 27 Jun 2005 16:09:00 +0000 (13:09 -0300)]
[PATCH] 8xx: avoid "dcbst" misbehaviour with unpopulated TLB
The proposed _tlbie call at update_mmu_cache() is safe because:
Addresses for which update_mmu_cache() gets invocated are never inside the
static kernel virtual mapping, meaning that there is no risk for the
_tlbie() here to be thrashing the pinned entry, as Dan suspected.
The intermediate TLB state in which this bug can be triggered is not
visible by userspace or any other contexts, except the page fault handling
path. So there is no need to worry about userspace dcbxxx users.
The other solution to this is to avoid dcbst misbehaviour in the first
place, which involves changing in-kernel "dcbst" callers to use 8xx
specific SPR's.
Summary:
On 8xx, cache control instructions (particularly "dcbst" from
flush_dcache_icache) fault as write operation if there is an unpopulated
TLB entry for the address in question. To workaround that, we invalidate
the TLB here, thus avoiding dcbst misbehaviour.
Bob Picco [Mon, 27 Jun 2005 21:36:28 +0000 (14:36 -0700)]
[PATCH] fix WANT_PAGE_VIRTUAL in memmap_init
I spotted this issue while in memmap_init last week. I can't say the
change has any test coverage by me. start_pfn was formerly used in main
"for" loop. The fix is replace start_pfn with pfn.
Signed-off-by: Bob Picco <bob.picco@hp.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Jens Axboe [Mon, 27 Jun 2005 12:49:39 +0000 (14:49 +0200)]
[PATCH] CFQ io scheduler: scheduler switch oops
If cfq is managing a queue and a new scheduler is later selected, it is
possible for the cfqd unplug_work work to be queued after the kblockd
work struct has been flushed. The problem is the ordering of
cfq_shutdown_timer_wq() and blk_put_queue() in cfq_put_cfqd(). The
latter may rearm the work, leaving cfq_kick_queue() with dead data.
Jens Axboe [Mon, 27 Jun 2005 08:56:24 +0000 (10:56 +0200)]
[PATCH] CFQ io scheduler updates
- Adjust slice values
- Instead of one async queue, one is defined per priority level. This
prevents kernel threads (such as reiserfs/x and others) that run at
higher io priority from conflicting with others. Previously, it was a
coin toss what io prio the async queue got, it was defined by who
first set up the queue.
- Let a time slice only begin, when the previous slice is completely
done. Previously we could be somewhat unfair to a new sync slice, if
the previous slice was async and had several ios queued. This might
need a little tweaking if throughput suffers a little due to this,
allowing perhaps an overlap of a single request or so.
- Optimize the calling of kblockd_schedule_work() by doing it only when
it is strictly necessary (no requests in driver and work left to do).
- Correct sync vs async logic. A 'normal' process can be purely async as
well, and a flusher can be purely sync as well. Sync or async is now a
property of the class defined and requests pending. Previously writers
could be considered sync, when they were really async.
- Get rid of the bit fields in cfqq and crq, use flags instead.
Jens Axboe [Mon, 27 Jun 2005 08:55:49 +0000 (10:55 +0200)]
[PATCH] Fix cfq_find_next_crq()
In cfq_find_next_crq(), cfq tries to find the next request by choosing
one of two requests before and after the current one. Currently, when
choosing the next request, if there's no next request, the next
candidate is NULL, resulting in selection of the previous request. This
results in weird scheduling. Once we reach the end, we always seek
backward.
The correct behavior is using the first request as the next candidate.
cfq_choose_req() already has logics for handling wrapped requests.
Jens Axboe [Mon, 27 Jun 2005 08:55:12 +0000 (10:55 +0200)]
[PATCH] Update cfq io scheduler to time sliced design
This updates the CFQ io scheduler to the new time sliced design (cfq
v3). It provides full process fairness, while giving excellent
aggregate system throughput even for many competing processes. It
supports io priorities, either inherited from the cpu nice value or set
directly with the ioprio_get/set syscalls. The latter closely mimic
set/getpriority.
It doesn't make sense to have the PGD kernel pointers initialisation
separate from the PGD user pointers, especially when we clean the
data cache over the whole range.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Russell King [Mon, 27 Jun 2005 10:12:54 +0000 (11:12 +0100)]
[PATCH] Serial: Split 8250 port table
Add separate files for the different 8250 ISA-based serial boards.
Looking across all the various architectures, it seems reasonable that
we can key the availability of the configuration options for these
beasts to the bus-related symbols (iow, CONFIG_ISA). We also standardise
the base baud/uart clock rate for these boards - I'm sure that isn't
architecture specific, but is solely dependent on the crystal fitted
on the board (which should be the same no matter what type of machine
its fitted into.)
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Russell King [Mon, 27 Jun 2005 09:29:26 +0000 (10:29 +0100)]
[PATCH] Serial: Remove {un,}register_serial from 68328serial.c
68328serial.c does not make use of register_serial/unregister_serial,
which is traditionally used to register 8250-compatible UARTs with
the 8250-compatible serial driver.
Acked-by: David McCullough Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Tobias Klauser [Sat, 28 May 2005 23:17:29 +0000 (01:17 +0200)]
[PATCH] sound/oss/via82cxxx_audio: Use the DMA_32BIT_MASK constant
Use the DMA_32BIT_MASK constant from dma-mapping.h
when calling pci_set_dma_mask() or pci_set_consistent_dma_mask()
See http://marc.theaimsgroup.com/?t=108001993000001&r=1&w=2 for details
Signed-off-by: Tobias Klauser <tklauser@nuerscht.ch> Signed-off-by: Domen Puncer <domen@coderock.org>
Check the link state during b44_open. This closes a 1 HZ window
that existed after b44_open ran but before the b44_timer handler ran,
during which ethtool would report "Link detected: yes" no matter what
the link state actually was.
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Olaf Hering [Wed, 25 May 2005 23:16:51 +0000 (01:16 +0200)]
[PATCH] uninitialized variable in prism54 isl38xx_trigger_device
drivers/net/wireless/prism54/isl_38xx.c:131: warning: 'current_time.tv_sec' is used uninitialized in this function
drivers/net/wireless/prism54/isl_38xx.c:131: warning: 'current_time.tv_usec' is used uninitialized in this function
Nicolas Pitre [Mon, 20 Jun 2005 03:56:21 +0000 (23:56 -0400)]
[PATCH] smc91x: plug race between TX tasklet and driver reset
The race causes a kernel oops when smc_hardware_send_pkt() tries to
dereference pending_tx_skb which would have been freed from one of the
driver reset paths just after the tx_task tasklet has been scheduled.
This race is possible on SMP but was uncovered by the kernel RT work.
Malli Chilakala [Sat, 18 Jun 2005 00:27:41 +0000 (17:27 -0700)]
[PATCH] e100: fixed e100 MDI/MDI-X issues
Added patch from Eran Mann to fix following e100 MDI/MDI-X issues
* MDI/MDI-X autodetection should never be enabled for 82551ER/QM chips
* enabling this feature based on eeprom settings
Signed-off-by: Mallikarjuna R Chilakala <mallikarjuna.chilakala@intel.com> Signed-off-by: Ganesh Venkatesan <ganesh.venkatesan@intel.com> Signed-off-by: John Ronciak <john.ronciak@intel.com>
[PATCH] orinoco: always use 802.11 header for rx processing
If the frame has ToDS flag set, mark it by setting skb->pkt_type to
PACKET_OTHERHOST, so that applications unaware of promiscous mode won't get
uplink (STA->AP) packets for STA->STA transmissions relayed by the AP.
Thanks to John Denker and David Gibson for finding the problem and the
solution.
Tobias Klauser [Mon, 20 Jun 2005 21:49:08 +0000 (23:49 +0200)]
[PATCH] drivers/block/sx8.c: Use the DMA_{64, 32}BIT_MASK constants
Use the DMA_{64,32}BIT_MASK constants from dma-mapping.h when calling
pci_set_dma_mask() or pci_set_consistent_dma_mask()
These patches include dma-mapping.h explicitly because it caused errors
on some architectures otherwise.
See http://marc.theaimsgroup.com/?t=108001993000001&r=1&w=2 for details
Signed-off-by: Tobias Klauser <tklauser@nuerscht.ch> Signed-off-by: Domen Puncer <domen@coderock.org>
[PATCH] net/farsync: add set_current_state() before schedule_timeout()
Insert set_current_state() before schedule_timeout() so the
function delays as expected. Without the addition, schedule_timeout()
will return immediately.
Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com> Signed-off-by: Maximilian Attems <janitor@sternwelten.at> Signed-off-by: Domen Puncer <domen@coderock.org>
[PATCH] net/sb1000: replace nicedelay() with ssleep()
Use ssleep() instead of nicedelay()
to guarantee the task delays as expected. Remove the prototype and
definition of nicedelay(). This is a very weird function, because it is
called to sleep in terms of usecs, but always sleeps for 1 second,
completely ignoring the parameter. I have gone ahead and followed suit,
just sleeping for a second in all cases, but maybe someone with the
hardware could tell me if perhaps the paramter *should* matter. Additionally,
nicedelay() is called in TASK_INTERRUPTIBLE state, but doesn't deal with signals
in case these longer delays do not complete, so I believe ssleep() is more
appropriate.
Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com> Signed-off-by: Domen Puncer <domen@coderock.org>
more outputs from Alexey Dobriyan:
2.95.3:
text data bss dec hex filename
before 51118 2156 160 53434 d0ba drivers/net/wireless/airo.o
after 51118 2156 160 53434 d0ba drivers/net/wireless/airo.o
3.3.5-20050130:
before 46999 2156 160 49315 c0a3 drivers/net/wireless/airo.o
after 46994 2156 160 49310 c09e drivers/net/wireless/airo.o
4.1.0-20050522:
before 45555 2220 160 47935 bb3f drivers/net/wireless/airo.o
after 45550 2220 160 47930 bb3a drivers/net/wireless/airo.o
pageexec [Sun, 26 Jun 2005 23:00:19 +0000 (16:00 -0700)]
[IPVS]: Fix for overflows
From: <pageexec@freemail.hu>
$subject was fixed in 2.4 already, 2.6 needs it as well.
The impact of the bugs is a kernel stack overflow and privilege escalation
from CAP_NET_ADMIN via the IP_VS_SO_SET_STARTDAEMON/IP_VS_SO_GET_DAEMON
ioctls. People running with 'root=all caps' (i.e., most users) are not
really affected (there's nothing to escalate), but SELinux and similar
users should take it seriously if they grant CAP_NET_ADMIN to other users.
Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: David S. Miller <davem@davemloft.net>