A lot of the REQ_* flags are only used on struct requests, and only of
use to the block layer and a few drivers that dig into struct request
internals.
This patch adds a new req_flags_t rq_flags field to struct request for
them, and thus dramatically shrinks the number of common requests. It
also removes the unfortunate situation where we have to fit the fields
from the same enum into 32 bits for struct bio and 64 bits for
struct request.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Shaun Tancheff <shaun.tancheff@seagate.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
Currently if we have PCI and UFSHCD configured in the kernel, both
SCSI_UFS_DWC_TC_PCI and SCSI_UFSHCD_PCI show up, which is not correct.
This patch changes the UFS Kconfig to assure hierarchy between the 2
options.
Signed-off-by: Joao Pinto <jpinto@synopsys.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Fix to return a negative error code from the error handling case instead
of 0, as done elsewhere in this function.
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Reviewed-by: Subhash Jadavani <subhashj@codeaurora.org>
Reviewed-by: Kiwoong Kim <kwmad.kim@samsung.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Some devices have problems handling a Query UPIU with Data Segment
set. Only set it for WRITE DESCRIPTOR commands.
[mkp: updated patch description]
Signed-off-by: Zang Leigang <zangleigang@hisilicon.com>
Reviewed-by: Subhash Jadavani <subhashj@codeaurora.org>
Reviewed-by: Kiwoong Kim <kwmad.kim@samsung.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
This patch enable no vccq quirk for SKHynix devices. SKHynix ufs device
don't need vccq vrail for device operation.
Signed-off-by: Kyuho Choi <kyuho.choi@sk.com>
Reviewed-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
When any UFS host controller receives a TM(Task Management) response
from a UFS device, UFS driver has been recognize like receiving a
message of "Task Management Function Complete"(00h) in all cases, so
far. That means there is no pending task for a tag of the TM request
sent before in the UFS device. That's because the byte offset 6 in TM
response which has been used to get a TM service response so far
represents just whether or not a TM transmission passes.
Regarding UFS spec, the correct byte offset to get TM service response
is 15, not 6.
I tested that UFS driver responds properly for the TM response from a
UFS device with an reference board with exynos8890, as follow: No
pending task -> Task Management Function Complete (00h) Pending task ->
Task Management Function Succeeded (08h)
[mkp: applied by hand]
Signed-off-by: Kiwoong Kim <kwmad.kim@samsung.com>
Signed-off-by: HeonGwang Chu <hg.chu@samsung.com>
Tested-by: : Kiwoong Kim <kwmad.kim@samsung.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
I modified a string as described in UFS spec as follow: umpcrs -> upmcrs.
[mkp: applied by hand]
Signed-off-by: Kiwoong Kim <kwmad.kim@samsung.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
We get 2 warnings when build kernel with W=1:
drivers/scsi/ufs/tc-dwc-g210.c:261:5: warning: no previous prototype for 'tc_dwc_g210_config_40_bit' [-Wmissing-prototypes]
drivers/scsi/ufs/tc-dwc-g210.c:293:5: warning: no previous prototype for 'tc_dwc_g210_config_20_bit' [-Wmissing-prototypes]
In fact, these functions are declared in ufs/tc-dwc-g210.h, so this
patch add missing header dependencies
Signed-off-by: Baoyou Xie <baoyou.xie@linaro.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
When buff_ascii kmalloc failed, there is no need to call kfree, it
should return -ENOMEM directly, this patch fixes it.
Signed-off-by: Tiezhu Yang <kernelpatch@126.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
The code was checking on PA_CONNECTEDRXLANES and PA_CONNECTEDTXLANES
attributes to program the Lane#1 attributes. The correct attributes are
PA_AVAILRXDATALANES and PA_AVAILTXDATALANES respectively.
Signed-off-by: Manjunath M B <manjumb@synopsys.com>
Tested-by: Joao Pinto <jpinto@synopsys.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
This patch adds license info to the tc-dwc-g210 and ufshcd-dwc files in
order for them to have access to some ufshcd symbols when all are built
as modules.
Signed-off-by: Joao Pinto <jpinto@synopsys.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
This patch adds a glue pci driver for the Synopsys G210 Test Chip.
[mkp: Fixed Kconfig depends and module name]
Signed-off-by: Joao Pinto <jpinto@synopsys.com>
Reviewed-by: Hannes Reinicke <hare@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
This patch adds a glue platform driver for the Synopsys G210 Test Chip.
Signed-off-by: Joao Pinto <jpinto@synopsys.com>
Reviewed-by: Hannes Reinicke <hare@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
This patch adds support for Synopsys G210 Test Chip.
Signed-off-by: Joao Pinto <jpinto@synopsys.com>
Reviewed-by: Hannes Reinicke <hare@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
This patch has the goal to add support for DesignWare UFS Controller
specific operations.
Signed-off-by: Joao Pinto <jpinto@synopsys.com>
Reviewed-by: Hannes Reinicke <hare@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Add link status to ufshci.
Signed-off-by: Joao Pinto <jpinto@synopsys.com>
Acked-by: Tomas Winkler <tomas.winkler@intel.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Add unipro attributes.
Signed-off-by: Joao Pinto <jpinto@synopsys.com>
Acked-by: Tomas Winkler <tomas.winkler@intel.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Add UFS 2.0 support to the UFS core driver.
Signed-off-by: Joao Pinto <jpinto@synopsys.com>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Tomas Winkler <tomas.winkler@intel.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Fixed typo in ufshcd-pltfrm.
Signed-off-by: Joao Pinto <jpinto@synopsys.com>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Tomas Winkler <tomas.winkler@intel.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
A recent change to ufshcd introduced a call to utf16s_to_utf8s, a
function that is provided by the NLS module, so we get a link error when
that is not present:
drivers/scsi/built-in.o: In function `ufshcd_read_string_desc':
:(.text+0x124d0): undefined reference to `utf16s_to_utf8s'
This adds a Kconfig 'select' statement to avoid the build error.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Fixes: b573d484e4 ("scsi: ufs: add support to read device and string descriptors")
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
This change adds printouts of testbus and debug registers.
Reviewed-by: Gilad Broner <gbroner@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
This change enables the device ref clock before changing to HS mode
and disables it if entered to PWM mode.
Reviewed-by: Gilad Broner <gbroner@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Some UFS devices (and may be host) have issues if LCC is
enabled. So we are setting PA_Local_TX_LCC_Enable to 0
before link startup which will make sure that both host
and device TX LCC are disabled once link startup is
completed.
Reviewed-by: Gilad Broner <gbroner@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
We put the UFS device in sleep state & UFS link in hibern8 state during
runtime suspend. After this we put all the UFS rails in low power
modes immediately but it seems some devices may still draw more than
sleep current from UFS rails (especially from VCCQ rail) at-least for
500us.
To avoid this situation, this change adds 2ms delay before putting
these UFS rails in LPM mode.
Reviewed-by: Gilad Broner <gbroner@codeaurora.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Currently when we try to put the link in off/disabled state during
suspend, it seems link is not being kept in low power mode.
This patch fixes the issue by putting the link in hibern8 first
(so device also puts the link in low power mode) and then stop the
host controller.
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Optimal values of local UniPro parameters like PA_Hibern8Time &
PA_TActivate can help reduce the hibern8 exit latency. If both host and
device supports UniPro ver1.6 or later, these parameters will be
automatically tuned during link startup itself. But if either host or
device doesn't support UniPro ver 1.6 or later, we have to manually
tune them. But to keep manual tuning logic simple, we will only do
manual tuning if local unipro version doesn't support ver1.6 or later.
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
We are seeing that some devices are raising the urgent bkops exception
events even when BKOPS status doesn't indicate performace impacted or
critical. Handle these device by determining their urgent bkops status
at runtime.
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Query commands have 100ms timeout and it may timeout if they are
issued in parallel to ongoing read/write SCSI commands, this change
adds the retry (max: 10) in case command timeouts.
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Some vendor's UFS device sends back to back NACs for the DL data frames
causing the host controller to raise the DFES error status. Sometimes
such UFS devices send back to back NAC without waiting for new
retransmitted DL frame from the host and in such cases it might be
possible the Host UniPro goes into bad state without raising the DFES
error interrupt. If this happens then all the pending commands would
timeout only after respective SW command (which is generally too
large).
This change workarounds such device behaviour like this:
- As soon as SW sees the DL NAC error, it would schedule the error
handler
- Error handler would sleep for 50ms to see if there any fatal errors
raised by UFS controller.
- If there are fatal errors then SW does normal error recovery.
- If there are no fatal errors then SW sends the NOP command to
device to check if link is alive.
- If NOP command times out, SW does normal error recovery
- If NOP command succeed, skip the error handling.
If DL NAC error is seen multiple times with some vendor's UFS devices
then enable this quirk to initiate quick error recovery and also
silence related error logs to reduce spamming of kernel logs.
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
UFS driver's error handler forcefully tries to clear all the pending
requests. For each pending request in the queue, it waits 1 sec for it
to get cleared. If we have multiple requests in the queue then it's
possible that we might end up waiting for those many seconds before
resetting the host. But note that resetting host would any way clear
all the pending requests from the hardware. Hence this change skips
the forceful clear of the pending requests if we are anyway going to
reset the host (for fatal errors).
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Some UFS devices don't require VCCQ rail for device operations hence
this change adds support to recognize such devices and remove vote for
the unused VCCQ rail.
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Currently we use the host quirks mechanism in order to
handle both device and host controller quirks.
In order to support various of UFS devices we should separate
handling the device quirks from the host controller's.
Reviewed-by: Gilad Broner <gbroner@codeaurora.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Raviv Shvili <rshvili@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
This change adds support to read device descriptor and string descriptor
from a UFS device
Reviewed-by: Gilad Broner <gbroner@codeaurora.org>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Raviv Shvili <rshvili@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Sometimes due to hw issues it takes some time to the
host controller register to update. In order to verify the register
has updated, a polling is done until its value is set.
In addition the functions ufshcd_hba_stop() and
ufshcd_wait_for_register() was updated with an additional input
parameter, indicating the timeout between reads will
be done by sleeping or spinning the cpu.
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Raviv Shvili <rshvili@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
A race condition exists between request requeueing and scsi layer
error handling:
When UFS driver queuecommand returns a busy status for a request,
it will be requeued and its tag will be freed and set to -1.
At the same time it is possible that the request will timeout and
scsi layer will start error handling for it. The scsi layer reuses
the request and its tag to send error related commands to the device,
however its tag is no longer valid.
As this request was never really sent to the device, there is no
point to start error handling with the device.
Implement the scsi error handling timeout callback and bypass SCSI
error handling for request that were not actually sent to the device.
For such requests simply reset the block layer timer. Otherwise, let
SCSI layer perform the usual error handling.
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Gilad Broner <gbroner@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
When control reaches to Linux UFS driver during UFS boot mode, UFS host
controller interrupt status/enable registers may have left over
settings.
In order to avoid any spurious interrupts due to these left overs,
it's important to clear these interrupt status/enable registers before
enabling UFS interrupt handling.
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Different platform may have different number of lanes
for the UFS link.
Add parameter to device tree specifying how many lanes
should be configured for the UFS link.
Reviewed-by: Hannes Reinecke <hare@suse.de>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Gilad Broner <gbroner@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
QUERY_DESC_GEOMETRY_MAZ_SIZE
QUERY_DESC_GEOMETRY_MAX_SIZE
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Sometimes queries from the device might return a failure so it is
recommended to retry sending the query, before giving up. This change
adds a wrapper to retry sending a query attribute, in cases where we
need to wait longer, before we continue, or before reporting a failure.
Reviewed-by: Gilad Broner <gbroner@codeaurora.org>
Reviewed-by: Dolev Raviv <draviv@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Add a write memory barrier to make sure descriptors prepared are
actually written to memory before ringing the doorbell. We have also
added the write memory barrier after ringing the doorbell register so
that controller sees the new request immediately.
Reviewed-by: Dolev Raviv <draviv@codeaurora.org>
Signed-off-by: Gilad Broner <gbroner@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Performing several writes to UFS host controller registers has no
guarantee of ordering, so we must make sure register writes to setup
request list base address etc. are performed before the run/stop
register is enabled. In addition, when setting up a task request, we
must make sure the updating of descriptors takes places before ringing
the doorbell, similarly to setting up a transfer request.
Reviewed-by: Dolev Raviv <draviv@codeaurora.org>
Signed-off-by: Gilad Broner <gbroner@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
DME commands such as Hibern8 enter/exit and gear switch generate 2
completion interrupts, one for confirmation that command is received by
local UniPro and 2nd one is the final confirmation after communication
with remote UniPro. Currently both of these completions are registered
as interrupt events which is not quite necessary and instead we can just
wait for the interrupt of 2nd completion, this should reduce the number
of interrupts and could reduce the unnecessary CPU wakeups to handle
extra interrupts.
Reviewed-by: Gilad Broner <gbroner@codeaurora.org>
Reviewed-by: Dolev Raviv <draviv@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
UFS flag query requests may fail sometimes due to timeouts etc. Add a
wrapper function to retry up to 10 times in case of such failure,
similar to retries being made for attribute queries.
Reviewed-by: Dolev Raviv <draviv@codeaurora.org>
Signed-off-by: Gilad Broner <gbroner@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Hibern8 exit can be called from 3 different contexts:
- ufshcd_hibern8_exit_work
- ufshcd_ungate_work
- runtime/system resume
If hibern8 exit fails for some reason then we try to bring the link to
active state by link startup but this recovery mechanism results into
deadlock or errors from first 2 context listed above. This change fixes
the recovery by adding proper error handling mechanism.
Reviewed-by: Gilad Broner <gbroner@codeaurora.org>
Reviewed-by: Dolev Raviv <draviv@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
If hibern8 enter command fails then UFS link state may be unknown which
may result into timeout of all the commands issued after failure.
This change does 2 things (for pre-defined number of retry counts) after
hibern8 enter failure:
1. Recovers the UFS link to active state
2. If link is recovered to active state, tries to put the UFS link in
hibern8 enter again until retry count expires.
Reviewed-by: Gilad Broner <gbroner@codeaurora.org>
Reviewed-by: Dolev Raviv <draviv@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
The dme_peer get/set attribute commands are prone to errors, therefore
we add three retries for the UIC command sending. Error code returned
from ufshcd_send_uic_cmd() is checked, and unless it was successful or
the retries have finished, another command will be sent.
Reviewed-by: Gilad Broner <gbroner@codeaurora.org>
Reviewed-by: Dolev Raviv <draviv@codeaurora.org>
Signed-off-by: Lee Susman <lsusman@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
If device raises the exception event in the response to the commands
sent during the runtime/system PM callbacks, exception event handler
might run in parallel with PM callbacks and may see unclocked register
accesses. This change fixes this issue by not scheduling the exception
event handler while PM callbacks are running.
Reviewed-by: Gilad Broner <gbroner@codeaurora.org>
Reviewed-by: Dolev Raviv <draviv@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
fDeviceInit query response time for some devices is too long that
default query request timeout of 100ms may not be enough. Experiments
show that fDeviceInit response sometimes takes 500ms so to be on safer
side this change sets the timeout to 600ms. Without this change, we
might unnecessarily have to retry fDeviceInit query requests multiple
times and each query request timeout prints one error message.
Reviewed-by: Gilad Broner <gbroner@codeaurora.org>
Reviewed-by: Dolev Raviv <draviv@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
When sending a query to the device returns with a timeout error, we
clear the corresponding bit in the DOORBELL register but we don't clear
the outstanding_request field as we should. This patch fixes this bug.
Reviewed-by: Subhash Jadavani <subhashj@codeaurora.org>
Reviewed-by: Gilad Broner <gbroner@codeaurora.org>
Reviewed-by: Dolev Raviv <draviv@codeaurora.org>
Signed-off-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Christoph Hellwig