Now that all drivers implementing new EH are converted to new probing
mechanism, ops->probe_reset doesn't have any user. Kill it.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Convert to new probing mechanism and add hotplug support by enabling
PORT_IRQ_PHYRDY_CHG, marking ehi for hotplug and scheduling EH on
PORT_IRQ_PHYRDY_CHG or PORT_IRQ_DEV_XCHG.
Sil3124/32 family of controllers don't have any mechanism to wait for
the first D2H FIS after hotplug, so ATA_FLAG_SKIP_D2H_BSY is used.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Convert to new probing mechanism and add hotplug support by enabling
PORT_IRQ_PHYRDY, marking ehi for hotplug and scheduling EH on
CONNECT/PHYRDY interrupts.
Unfortunately, ahci cannot reliably wait for the first D2H FIS after
hotplug. It sometimes succeeds but times out more often than not, so
ATA_FLAG_SKIP_D2H_BSY is used.
This patch also fixes ahci_hardreset() such that D2H Register FIS RX
area is cleared before issuing COMRESET. Without this,
ata_busy_sleep() after COMRESET might prematually finish if the
previous TF contains DRDY && !BSY.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Convert to new probing mechanism and add hotplug support by enabling
SATA IRQ for SError.N, marking ehi for hotplug and scheduling EH on
SATA IRQs.
Sil3112/3512/3114 family of controllers use COMRESET as TF clearing
point and can reliably wait for D2H FIS after COMRESET whether the FIS
is the first D2H FIS after POR or in response to the COMRESET. Thus,
setting ATA_FLAG_HRST_TO_RESUME is enough for device detection after
hotplug.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Convert ata_piix to new probing mechanism. Automatic hotplug is not
supported due to hardware limitation (no PHY event interrupt), but
warm plugging works.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Update unload unplug - driver unloading / PCI removal. This is done
by ata_port_detach() which short-circuits EH, disables all devices and
freezes the port. With this patch, EH and unloading/unplugging are
properly synchronized.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Implement bootplug - boot probing via hotplug path. While loading,
ata_host_add() simply schedules probing and invokes EH. After EH
completes, ata_host_add() scans and assicates them with SCSI devices.
EH path is slightly modified to handle this (e.g. no autopsy during
bootplug). The SCSI part is left in ata_host_add() because it's
shared with legacy path and to keep probing order as before (ATA scan
all ports in host_set then attach all).
Signed-off-by: Tejun Heo <htejun@gmail.com>
Implement warmplug. User-initiated unplug can be detected by
hostt->slave_destroy() and plug by transportt->user_scan(). This
patch only implements the two callbacks. The next function will hook
them.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Implement SCSI part of hotplug.
This must be done in a separate context as SCSI makes use of EH during
probing. SCSI scan fails silently if EH is in progress. In such
cases, libata pauses briefly and retries until every device is
attached.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Implement ATA part of hotplug. To avoid probing broken devices over
and over again, disabled devices are not automatically detached. They
are detached only if probing is requested for the device or the
associated port is offline. Also, to avoid infinite probing loop,
Each device is probed only once per EH run.
As SATA PHY status is fragile, devices are detached only after it has
used up its recovery chances unless explicitly requested by LLDD or
user (LLDD may request direct detach if, for example, it supports cold
presence detection).
Signed-off-by: Tejun Heo <htejun@gmail.com>
Implement ata_eh_detach_dev(). This function is responsible for
detaching an ATA device and offlining the associated SCSI device
atomically so that the detached device is not accessed after ATA
detach is complete.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Use phy debouncing instead of unconditional wait after DEV_RST and
make sil24_hardreset() to request followup SRST as that's the only way
to wait for !BSY. Note that the original implementation never worked
- if the cached status was !BSY, ata_busy_sleep() finished
immediately; otherwise, it timed out regardless of the actual device
status.
Signed-off-by: Tejun Heo <htejun@gmail.com>
PORT_PRB is a misnomer as the area also contains other stuff. Rename
it to PORT_LRAM and add PORT_LRAM_SLOT_SZ.
Signed-off-by: Tejun Heo <htejun@gmail.com>
The DMA complete bit of these controllers reflects ATA IRQ status
while no DMA command is in progress. So, we can tell whether the
controller is raising an interrupt or not in deterministic manner.
This patch gives sata_sil its own interrupt handler which behaves much
better than the original one in terms of error detection and handling.
This change is also necessary for later hotplug support.
Further improvements are possible, in both 2 and 4 ports versions, we
can get all status with only one readl and using custom bmdma
operations can further cut down register accesses.
Signed-off-by: Tejun Heo <htejun@gmail.com>
ata_hsm_move() will be used by LLDDs which depend on standard PIO HSM
but implement their own interrupt handlers.
Signed-off-by: Tejun Heo <htejun@gmail.com>
In some cases, hardreset must be followed by SRST.
* some controllers can't classify with hardreset
* some controllers can't wait for !BSY after hardreset (LLDD should
explicitly request followup softreset by returning -EAGAIN)
* (later) PM needs SRST w/ PMP==15 to operate after hardreset
To handle above cases, this patch implements follow-up softreset.
After a hardreset, ata_eh_reset() checks whether any of above
conditions are met and do a follow-up softreset if necessary.
Signed-off-by: Tejun Heo <htejun@gmail.com>
With hotplug, every reset might be a probing reset and thus something
similar to probe_init() is needed. prereset() method is called before
a series of resets to a port and is the counterpart of postreset().
prereset() can tell EH to use different type of reset or skip reset by
modifying ehc->i.action.
This patch also implements ata_std_prereset(). Most controllers
should be able to use this function directly or with some wrapping.
After hotplug, different controllers need different actions to resume
the PHY and detect the newly attached device. Controllers can be
categorized as follows.
* Controllers which can wait for the first D2H FIS after hotplug.
Note that if the waiting is implemented by polling TF status, there
needs to be a way to set BSY on PHY status change. It can be
implemented by hardware or with the help of the driver.
* Controllers which can wait for the first D2H FIS after sending
COMRESET. These controllers need to issue COMRESET to wait for the
first FIS. Note that the received D2H FIS could be the first D2H
FIS after POR (power-on-reset) or D2H FIS in response to the
COMRESET. Some controllers use COMRESET as TF status
synchronization point and clear TF automatically (sata_sil).
* Controllers which cannot wait for the first D2H FIS reliably.
Blindly issuing SRST to spinning-up device often results in command
issue failure or timeout, causing extended delay. For these
controllers, ata_std_prereset() explicitly waits ATA_SPINUP_WAIT
(currently 8s) to give newly attached device time to spin up, then
issues reset. Note that failing to getting ready in ATA_SPINUP_WAIT
is not critical. libata will retry. So, the timeout needs to be
long enough to spin up most devices.
LLDDs can tell ata_std_prereset() which of above action is needed with
ATA_FLAG_HRST_TO_RESUME and ATA_FLAG_SKIP_D2H_BSY flags. These flags
are PHY-specific property and will be moved to ata_link later.
While at it, this patch unifies function typedef's such that they all
have named arguments.
Signed-off-by: Tejun Heo <htejun@gmail.com>
With hotplug, PHY always needs to be debounced before a reset as any
reset might find new devices. Extract PHY waiting code from
sata_phy_resume() and extend it to include SStatus debouncing. Note
that sata_phy_debounce() is superset of what used to be done inside
sata_phy_resume().
Three default debounce timing parameters are defined to be used by
hot/boot plug. As resume failure during probing will be properly
handled as errors, timeout doesn't have to be long as before.
probeinit() uses the same timeout to retain the original behavior.
Signed-off-by: Tejun Heo <htejun@gmail.com>
It's best to run ATA hotplug from EH but attaching SCSI devices needs
working EH. ata_aux_wq is used to give SCSI hotplug operations a
separate context.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Add device persistent field dev->sdev and store the attached SCSI
device. With hotplug, libata needs to know the attached SCSI device
to offline and detach it, but scsi_device_lookup() cannot be used
because libata will reuse SCSI ID numbers - dead but not gone devices
(due to zombie opens, etc...) interfere with the lookup.
dev->sdev doesn't hold reference to the SCSI device. It's cleared
when the SCSI device goes away.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Add ap->hw_sata_spd_limit and initialize it once during the boot
initialization (or driver load initialization). ap->sata_spd_limit is
reset to ap->hw_sata_spd_limit on hotplug. This prevents spd limits
introduced by earlier devices from affecting new devices.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Separate out ata_find_dev() and __ata_scsi_find_dev() from
ata_scsi_find_dev(). These will be used by later hotplug
implementation.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Lifetimes of some fields span over device plugging/unplugging. This
patch moves such persistent fields to the top of ata_device and
separate them with ATA_DEVICE_CLEAR_OFFSET. Fields above the offset
are initialized once during host initializatino while all other fields
are cleared before hotplugging. Currently ->ap, devno and part of
flags are persistent.
Note that flags is partially cleared while holding host_set lock.
This is to synchronize with later warm plug implementation which will
record hotplug request in dev->flags.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Move initialization of struct ata_device into ata_dev_init() in
preparation for hotplug. This patch calls ata_dev_init() from
ata_host_init() and thus makes no functional difference.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Implement ata_eh_wait(). On return from this function, it's
guaranteed that the EH which was pending or in progress when the
function was called is complete - including the tailing part of SCSI
EH. This will be used by hotplug and others to synchronize with EH.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Set ata_device->pio_mode to XFER_PIO_0 after a successful reset. This
is to keep EH resets consistent with probe resets as updated by the
commit b6079ca409. Note that, with
soon-to-follow hotplug update, EH resets will include probe resets.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Uses of smp_processor_id() get pushed earlier and earlier in
the start_kernel() sequence. So just get it working before
we call start_kernel() to avoid all possible problems.
Signed-off-by: David S. Miller <davem@davemloft.net>
We don't enable the BTB on the ixp2350 as that can cause weird
crashes (erratum #42.) However, some bootloaders enable the BTB,
which means that we have to disable the BTB explicitly.
Found thanks to Tom Rini.
Signed-off-by: Lennert Buytenhek <buytenh@wantstofly.org>
Signed-off-by: Deepak Saxena <dsaxena@plexity.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This reverts commit 5491d0f3e2.
As per Andi:
"After some discussion with people who have the affected system it
seems best to revert for 2.6.17. It broke a common BIOS workaround
and PCI-X still doesn't work. Alternative is for people to change
the BIOS which seems to be better right now."
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
int_ret_from_syscall already does syscall exit tracing, so
no need to do it again in the caller.
This caused problems for UML and some other special programs doing
syscall interception.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
From: Robert Hentosh <robert_hentosh@dell.com>
Actually, we just stumbled on a different bug found in find_e820_area() in
e820.c. The following code does not handle the edge condition correctly:
while (bad_addr(&addr, size) && addr+size < ei->addr + ei->size)
;
last = addr + size;
if ( last > ei->addr + ei->size )
continue;
The second statement in the while loop needs to be a <= b so that it is the
logical negavite of the if (a > b) outside it. It needs to read:
while (bad_addr(&addr, size) && addr+size <= ei->addr + ei->size)
;
In the case that failed bad_addr was returning an address that is exactly size
bellow the end of the e820 range.
AK: Again together with the earlier avoid edma fix this fixes
boot on a Dell PE6850/16GB
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
From: Daniel Yeisley <dan.yeisley@unisys.com>
It is possible to boot a Unisys ES7000 with CPUs from multiple cells, and not
also include the memory from those cells. This can create a scenario where
node 0 has cpus, but no associated memory. The system will boot fine in a
configuration where node 0 has memory, but nodes 2 and 3 do not.
[AK: I rechecked the code and generic code seems to indeed handle that already.
Dan's original patch had a change for mm/slab.c that seems to be already in now.]
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
From: "Jan Beulich" <jbeulich@novell.com>
The PM timer code updates vxtime.last_tsc, but this update was done
incorrectly in two ways:
- offset_delay being in microseconds requires multiplying with cpu_mhz
rather than cpu_khz
- the multiplication of offset_delay and cpu_khz (both being 32-bit
values) on most current CPUs would overflow (observed value of the
delay was approximately 4000us, yielding an overflow for frequencies
starting a little above 1GHz)
Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
From: "Jan Beulich" <jbeulich@novell.com>
When using apic= on the kernel command line, this had no effect for machines
matched by either the ACPI MADT or the MPS OEM table scan. However, when such
option is specified, it should also take effect for this set of systems.
Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Complaining about the IOMMU not compiled in doesn't make sense
here because it is clearly compiled in.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
ia32_setup_arg_pages would ignore the passed in random stack top
and use its own static value.
Now it uses the 8bit of randomness native i386 would use too.
This indirectly fixes mmap randomization for 32bit processes too,
which depends on the stack randomization.
Should also give slightly better virtual cache colouring and
possibly better performance with HyperThreading.
Signed-off-by: Andi Kleen <ak@suse.de>
Acked-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
- Drop cic from the list when seen as dead.
- Fixup the locking, just use a simple spinlock.
Signed-off-by: Jens Axboe <axboe@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
* master.kernel.org:/pub/scm/linux/kernel/git/davej/agpgart:
[AGPGART] VIA PT880 Ultra support.
[AGPGART] Fix Nforce3 suspend on amd64.
[AGPGART] Enable SIS AGP driver on x86-64 for EM64T systems
* git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6:
[[CIFS] Pass truncate open flag through on file open in case setattr fails
[CIFS] Fix typos in previous fix
[CIFS] endian fix for new POSIX byte range lock support
[CIFS] fix memory leak in cifs session info struct on reconnect
[CIFS] ACPI suspend oops
[CIFS] Do not limit the length of share names (was 100 for whole UNC name)
[CIFS] Fix new POSIX Locking for setting lock_type correctly on unlock
