GitHub/moto-9609/android_kernel_motorola_exynos9610.git
7 years agoscsi: pmcraid: switch to pci_alloc_irq_vectors
Christoph Hellwig [Fri, 18 Nov 2016 06:28:16 +0000 (07:28 +0100)]
scsi: pmcraid: switch to pci_alloc_irq_vectors

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: bfa: remove bfa_fcs_mod_s
Christoph Hellwig [Sun, 8 Jan 2017 09:41:15 +0000 (10:41 +0100)]
scsi: bfa: remove bfa_fcs_mod_s

Just call the functions directly instead of obsfucating the call chain.
This was in reply to a patch from Kees Cook to constify the function
pointer struct bfa_fcs_mod_s, but it turns out there is no reason to
have this indirection at all.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: qla2xxx: rename {vendor|hba}_indentifer to {vendor|hba}_identifer
Colin Ian King [Thu, 29 Dec 2016 22:20:38 +0000 (22:20 +0000)]
scsi: qla2xxx: rename {vendor|hba}_indentifer to {vendor|hba}_identifer

Rename the vendor_indentifer and hba_indentifer fields to correct
spelling.

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Acked-by: Himanshu Madhani <himanshu.madhani@cavium.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: qla2xxx: make msix_entries const
Nicolas Iooss [Mon, 26 Dec 2016 13:23:10 +0000 (14:23 +0100)]
scsi: qla2xxx: make msix_entries const

msix_entries and qla82xx_msix_entries arrays are never modified in
drivers/scsi/qla2xxx/qla_isr.c. Move their contents to read-only data.

Signed-off-by: Nicolas Iooss <nicolas.iooss_linux@m4x.org>
Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com>
Acked-by: Himanshu Madhani <himanshu.madhani@cavium.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: qla2xxx: silence -Wformat-security warning
Nicolas Iooss [Mon, 26 Dec 2016 13:23:09 +0000 (14:23 +0100)]
scsi: qla2xxx: silence -Wformat-security warning

qla24xx_enable_msix() calls scnprintf() with a non-literal format
string. This makes clang report -Wformat-security warnings when
compiling this function:

    drivers/scsi/qla2xxx/qla_isr.c:3083:7: error: format string is not a
    string literal (potentially insecure) [-Werror,-Wformat-security]
                        msix_entries[i].name);
                        ^~~~~~~~~~~~~~~~~~~~
    drivers/scsi/qla2xxx/qla_isr.c:3083:7: note: treat the string as an
    argument to avoid this
                        msix_entries[i].name);
                        ^
                        "%s",
    drivers/scsi/qla2xxx/qla_isr.c:3119:7: error: format string is not a
    string literal (potentially insecure) [-Werror,-Wformat-security]
                        msix_entries[QLA_ATIO_VECTOR].name);
                        ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
    drivers/scsi/qla2xxx/qla_isr.c:3119:7: note: treat the string as an
    argument to avoid this
                        msix_entries[QLA_ATIO_VECTOR].name);
                        ^
                        "%s",

Even though msix_entries[...].name are initialized as literal strings
with no % character and are never modified, introduce a "%s" format
parameter in order to silence this -Wformat-security warning and make
clang able to detect at compile time real bugs related to string
formatting.

[mkp: typo]

Signed-off-by: Nicolas Iooss <nicolas.iooss_linux@m4x.org>
Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com>
Acked-by: Himanshu Madhani <himanshu.madhani@cavium.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: lpfc: Reinstate lpfc_soft_wwn parameter
James Smart [Fri, 30 Dec 2016 14:57:47 +0000 (06:57 -0800)]
scsi: lpfc: Reinstate lpfc_soft_wwn parameter

The lpfc 11.2.0.4 patch set deprecated, by removing, the lpfc_soft_wwn
parameter support.

This patch reinstates support, but adds a warning in the enablement of
the feature that indicates Broadcom (Emulex) does not support the
feature.

Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: hisi_sas: lock sensitive region in hisi_sas_slot_abort()
John Garry [Tue, 3 Jan 2017 12:24:50 +0000 (20:24 +0800)]
scsi: hisi_sas: lock sensitive region in hisi_sas_slot_abort()

When we call hisi_sas_slot_task_free() we should grab the hisi_hba.lock,
as hisi_sas_slot_task_free() accesses common hisi_hba elements.
Function hisi_sas_slot_abort() is missing this, so add it.

Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Zhangfei Gao <zhangfei.gao@linaro.org>
Tested-by: Hanjun Guo <hanjun.guo@linaro.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: hisi_sas: lock sensitive regions when servicing CQ interrupt
John Garry [Tue, 3 Jan 2017 12:24:49 +0000 (20:24 +0800)]
scsi: hisi_sas: lock sensitive regions when servicing CQ interrupt

There is a bug in the current driver in that certain hisi_hba and port
structure elements which we access when servicing the CQ interrupt do
not use thread-safe accesses; these include hisi_sas_port linked-list of
active slots (hisi_sas_port.entry), bitmap of currently allocated IPTT
(in hisi_hba.slot_index_tags), and completion queue read pointer.

As a solution, lock these elements with the hisi_hba.lock.

Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Xiang Chen <chenxiang66@hisilicon.com>
Reviewed-by: Zhangfei Gao <zhangfei.gao@linaro.org>
Tested-by: Hanjun Guo <hanjun.guo@linaro.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: hisi_sas: service v2 hw CQ ISR with tasklet
John Garry [Tue, 3 Jan 2017 12:24:48 +0000 (20:24 +0800)]
scsi: hisi_sas: service v2 hw CQ ISR with tasklet

Currently the all the slot processing for the completion queue is done
in ISR context. It is judged that the slot processing can take a long
time, especially when a SATA NCQ completes (upto 32 slots).

So, as a solution, defer the bulk of the ISR processing to tasklet
context. Each CQ will have its down tasklet.

Signed-off-by: John Garry <john.garry@huawei.com>
Reviewed-by: Xiang Chen <chenxiang66@hisilicon.com>
Reviewed-by: Zhangfei Gao <zhangfei.gao@linaro.org>
Tested-by: Hanjun Guo <hanjun.guo@linaro.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: ufs-qcom: Fix module autoload
Javier Martinez Canillas [Mon, 2 Jan 2017 14:04:58 +0000 (11:04 -0300)]
scsi: ufs-qcom: Fix module autoload

If the driver is built as a module, autoload won't work because the module
alias information is not filled. So user-space can't match the registered
device with the corresponding module.

Export the module alias information using the MODULE_DEVICE_TABLE() macro.

Before this patch:

$ modinfo drivers/scsi/ufs/ufs-qcom.ko | grep alias
$

After this patch:

$ modinfo drivers/scsi/ufs/ufs-qcom.ko | grep alias
alias:          of:N*T*Cqcom,ufshcC*
alias:          of:N*T*Cqcom,ufshc

Signed-off-by: Javier Martinez Canillas <javier@osg.samsung.com>
Reviewed-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: ufs: Improve fatal error logs
Dolev Raviv [Fri, 23 Dec 2016 02:42:18 +0000 (18:42 -0800)]
scsi: ufs: Improve fatal error logs

Errors such as UIC error, illegal OCS values, and others may require
more information for debugging. Such information could be hibern8 events,
events sequences, recoverable errors, error history, and more.
This patch improves tracking of important errors and events in debug level
to be enabled when debugging a such issues. It includes:
* UIC error history
* Successful hibern8 events
* Successful command after hibern8 exit
* Clk-freq info
* Failed device command
* Infrastructure for dumping host controller debug information

Signed-off-by: Dolev Raviv <draviv@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: ufs: add trace event for ufs commands
Lee Susman [Fri, 23 Dec 2016 02:42:03 +0000 (18:42 -0800)]
scsi: ufs: add trace event for ufs commands

Use the ftrace infrastructure to conditionally trace ufs command events.
New trace event is created, which samples the following ufs command data:
- device name
- optional identification string
- task tag
- doorbell register
- number of transfer bytes
- interrupt status register
- request start LBA
- command opcode

Currently we only fully trace read(10) and write(10) commands.
All other commands which pass through ufshcd_send_command() will be
printed with "-1" in the lba and transfer_len fields.

Usage:
echo 1 > /sys/kernel/debug/tracing/events/ufs/enable
cat /sys/kernel/debug/tracing/trace_pipe

Signed-off-by: Lee Susman <lsusman@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: ufs: add time profiling support
subhashj@codeaurora.org [Fri, 23 Dec 2016 02:41:48 +0000 (18:41 -0800)]
scsi: ufs: add time profiling support

This patch adds the profiling support for some of the time critical
operations like hibern8 enter/exit, clock gating & clock scaling.

Reviewed-by: Venkat Gopalakrishnan <venkatg@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: ufs: fix setting init power mode
subhashj@codeaurora.org [Fri, 23 Dec 2016 02:41:33 +0000 (18:41 -0800)]
scsi: ufs: fix setting init power mode

Immediately after successful UFS link startup, UFS link power mode would
be in PWM-G1, 1-lane, SLOW-AUTO mode. But currently we are doing few
of the DME local/peer attributes access before setting the "hba->pwr_info"
to default power mode. If we are doing link startup as part of error
recovery then old power mode might be set to FAST mode and doing DME peer
access (after link startup but before updating "hba->pwr_info" to default
power mode) unintentionally tries to switch from FAST to FAST_AUTO mode (if
UFSHCD_QUIRK_DME_PEER_ACCESS_AUTO_MODE quirk is enabled).

Above issue is fixed by setting the default power mode immediately after
successful link startup.

Reviewed-by: Sahitya Tummala <stummala@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: ufs: add capability to keep auto bkops always enabled
subhashj@codeaurora.org [Fri, 23 Dec 2016 02:41:22 +0000 (18:41 -0800)]
scsi: ufs: add capability to keep auto bkops always enabled

UFS device requires to perform bkops (back ground operations) periodically
but host can control (via auto-bkops parameter of device) when device can
perform bkops based on its performance requirements. In general, host
would like to enable the device's auto-bkops only when it's not doing any
regular data transfer but sometimes device may not behave properly if host
keeps the auto-bkops disabled. This change adds the capability to let the
device auto-bkops always enabled except suspend.

Reviewed-by: Sahitya Tummala <stummala@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: ufs: set default UFS power management level
subhashj@codeaurora.org [Fri, 23 Dec 2016 02:41:11 +0000 (18:41 -0800)]
scsi: ufs: set default UFS power management level

UFS device and link can be put in multiple different low power modes hence
UFS driver supports multiple different low power modes.
This change sets the default UFS power management level which should put
the link hibernate state and device in sleep state. This default power
management level gives good  power savings with relatively less enter/exit
latencies.

Reviewed-by: Yaniv Gardi <ygardi@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: ufs: provide sysfs attribute to select the PM level
subhashj@codeaurora.org [Fri, 23 Dec 2016 02:41:00 +0000 (18:41 -0800)]
scsi: ufs: provide sysfs attribute to select the PM level

This patch provides the sysfs attribute to choose the power management
level for UFS runtime and system suspend.

Reviewed-by: Sujit Reddy Thumma <sthumma@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: ufs: Add sysfs node to dynamically control clock scaling
Sahitya Tummala [Fri, 23 Dec 2016 02:40:50 +0000 (18:40 -0800)]
scsi: ufs: Add sysfs node to dynamically control clock scaling

Provide an option to enable/disable clock scaling during runtime.
Write 1/0 to "clkscale_enable" sysfs node to enable/disable clock
scaling.

Signed-off-by: Sahitya Tummala <stummala@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: ufs: Add sysfs node to dynamically control clock gating
Sahitya Tummala [Fri, 23 Dec 2016 02:40:39 +0000 (18:40 -0800)]
scsi: ufs: Add sysfs node to dynamically control clock gating

Provide an option to enable/disable clock gating during runtime.
Write 1 or 0 to "clkgate_enable" sysfs node to enable/disable
clock gating.

Signed-off-by: Sahitya Tummala <stummala@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: ufs: fix multiple ufs spec violation
Dolev Raviv [Fri, 23 Dec 2016 02:40:07 +0000 (18:40 -0800)]
scsi: ufs: fix multiple ufs spec violation

When a command to a W-LU is timed out via scsi, error handling
will treat it as any other LU and send commands such as
START_STOP with wrong format or task abort. Those commands are
illegal for W-LU according to the UFS spec.
To solve it, when an error is recognized those steps are skipped
and the last step, reset and restore process, is initiated.

Signed-off-by: Dolev Raviv <draviv@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: ufs: add tracing support
subhashj@codeaurora.org [Fri, 23 Dec 2016 02:39:51 +0000 (18:39 -0800)]
scsi: ufs: add tracing support

This change adds the ftrace support for following:
1. UFS initialization time
2. Clock gating states
3. Clock scaling states
4. Power management APIs latency
5. BKOPs enable/disable

Usage:
echo 1 > /sys/kernel/debug/tracing/events/ufs/enable
cat /sys/kernel/debug/tracing/trace_pipe

Reviewed-by: Sahitya Tummala <stummala@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: ufs: dump debug info during failures
Dolev Raviv [Fri, 23 Dec 2016 02:39:42 +0000 (18:39 -0800)]
scsi: ufs: dump debug info during failures

Inserts driver dumps for UFS Host Controller registers, Transfer Requests
and Task Management Requests.
The dumps will occur on driver initialization failure, ufshcd_abort() and
on error handling path.

Signed-off-by: Dolev Raviv <draviv@codeaurora.org>
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: qla4xxx: comments correction
Cao jin [Mon, 19 Dec 2016 06:20:29 +0000 (14:20 +0800)]
scsi: qla4xxx: comments correction

Signed-off-by: Cao jin <caoj.fnst@cn.fujitsu.com>
Acked-by: Nilesh Javali <nilesh.javali@cavium.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: qedi: return via va_end to match corresponding va_start
Colin Ian King [Fri, 16 Dec 2016 14:10:43 +0000 (14:10 +0000)]
scsi: qedi: return via va_end to match corresponding va_start

Although on most systems va_end is a no-op, it is good practice to use
va_end on the function return path, especially since the va_start
documenation states:

  "Each invocation of va_start() must be matched by a corresponding
   invocation of va_end() in the same function."

Found with static analysis by CoverityScan, CIDs 1389477-1389479

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Acked-by: Manish Rangankar <manish.rangankar@cavium.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: be2iscsi: Update driver version
Jitendra Bhivare [Tue, 13 Dec 2016 10:26:06 +0000 (15:56 +0530)]
scsi: be2iscsi: Update driver version

Version 11.2.1.0

Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: be2iscsi: Add warning message for unsupported adapter
Ketan Mukadam [Tue, 13 Dec 2016 10:26:05 +0000 (15:56 +0530)]
scsi: be2iscsi: Add warning message for unsupported adapter

Add a warning message to indicate obsolete/unsupported
BE2 Adapter Family devices

Signed-off-by: Ketan Mukadam <ketan.mukadam@avagotech.com>
Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: be2iscsi: Reinit SGL handle, CID tables after TPE
Jitendra Bhivare [Tue, 13 Dec 2016 10:26:04 +0000 (15:56 +0530)]
scsi: be2iscsi: Reinit SGL handle, CID tables after TPE

After TPE recovery, CID table needs to be repopulated as per CIDs in
WRBQ creation responses.

SGL handles table needs to be recreated for posting and its indices need
to be resetted.

This is achieved by calling beiscsi_cleanup_port when disabling and
beiscsi_init_port in enabling port.

Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: be2iscsi: Add checks to validate CID alloc/free
Jitendra Bhivare [Tue, 13 Dec 2016 10:26:03 +0000 (15:56 +0530)]
scsi: be2iscsi: Add checks to validate CID alloc/free

Set CID slot to 0xffff to indicate empty.
Check if connection already exists in conn_table before binding.
Check if endpoint already NULL before putting back CID.
Break ep->conn link in free_ep to ignore completions after freeing.

Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: be2iscsi: Remove wq_name from beiscsi_hba
Jitendra Bhivare [Tue, 13 Dec 2016 10:26:02 +0000 (15:56 +0530)]
scsi: be2iscsi: Remove wq_name from beiscsi_hba

wq_name is used only to set WQ name when its being allocated.
Remove it from beiscsi_hba structure and define locally.

Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: be2iscsi: Remove unused struct members
Jitendra Bhivare [Tue, 13 Dec 2016 10:26:01 +0000 (15:56 +0530)]
scsi: be2iscsi: Remove unused struct members

Fix errors reported in static analysis.

Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: be2iscsi: Remove redundant receive buffers posting
Jitendra Bhivare [Tue, 13 Dec 2016 10:26:00 +0000 (15:56 +0530)]
scsi: be2iscsi: Remove redundant receive buffers posting

This duplicate code got added during manual merging.

Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: be2iscsi: Fix iSCSI cmd cleanup IOCTL
Jitendra Bhivare [Tue, 13 Dec 2016 10:25:59 +0000 (15:55 +0530)]
scsi: be2iscsi: Fix iSCSI cmd cleanup IOCTL

Prepare the IOCTL with appropriate sizes of buffers of V0 and V1.
Set missing chute number in V1 IOCTL.

Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: be2iscsi: Add checks to validate completions
Jitendra Bhivare [Tue, 13 Dec 2016 10:25:58 +0000 (15:55 +0530)]
scsi: be2iscsi: Add checks to validate completions

Added check in beiscsi_process_cq for pio_handle.
pio_handle is cleared in beiscsi_put_wrb_handle.
This catches any case where task gets cleaned up just before completion.

Use back_lock before accessing pio_handle.

Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: be2iscsi: Set WRB invalid bit for SkyHawk
Jitendra Bhivare [Tue, 13 Dec 2016 10:25:57 +0000 (15:55 +0530)]
scsi: be2iscsi: Set WRB invalid bit for SkyHawk

invalid bit in WRB indicates to FW that IO was invalidated before WRB
was fetched from host memory.

For SkyHawk, this invalid bit in WRB is at a different offset.
Use amap_iscsi_wrb_v2 to mark invalid bit for SkyHawk.

Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: be2iscsi: Take iscsi_task ref in abort handler
Jitendra Bhivare [Tue, 13 Dec 2016 10:25:56 +0000 (15:55 +0530)]
scsi: be2iscsi: Take iscsi_task ref in abort handler

Hold the reference of iscsi_task till invalidation completes.
This prevents use of ICD when invalidation of that ICD is being processed.

Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: be2iscsi: Fix for crash in beiscsi_eh_device_reset
Jitendra Bhivare [Tue, 13 Dec 2016 10:25:55 +0000 (15:55 +0530)]
scsi: be2iscsi: Fix for crash in beiscsi_eh_device_reset

System crashes when sg_reset is executed in a loop.
CPU: 13 PID: 7073 Comm: sg_reset Tainted: G            E   4.8.0-rc1+ #4
RIP: 0010:[<ffffffffa0825370>]  [<ffffffffa0825370>]
beiscsi_eh_device_reset+0x160/0x520 [be2iscsi]
Call Trace:
[<ffffffff814c7c77>] ? scsi_host_alloc_command+0x47/0xc0
[<ffffffff814caafa>] scsi_try_bus_device_reset+0x2a/0x50
[<ffffffff814cb46e>] scsi_ioctl_reset+0x13e/0x260
[<ffffffff814ca477>] scsi_ioctl+0x137/0x3d0
[<ffffffffa05e4ba2>] sg_ioctl+0x572/0xc20 [sg]
[<ffffffff8123f627>] do_vfs_ioctl+0xa7/0x5d0

The accesses to beiscsi_io_task is being protected in device reset handler
with frwd_lock but the freeing of task can happen under back_lock.

Hold the reference of iscsi_task till invalidation completes.
This prevents use of ICD when invalidation of that ICD is being processed.
Use frwd_lock for iscsi_tasks looping and back_lock to access
beiscsi_io_task structures.

Rewrite mgmt_invalidation_icds to handle allocation and freeing of IOCTL
buffer in one place.

Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: be2iscsi: Fix use of invalidate command table req
Jitendra Bhivare [Tue, 13 Dec 2016 10:25:54 +0000 (15:55 +0530)]
scsi: be2iscsi: Fix use of invalidate command table req

Remove shared structure inv_tbl in phba for all sessions to post
invalidation IOCTL.
Always allocate and then free the table after use in reset handler.
Abort handler needs just one instance so define it on stack.
Add checks for BE_INVLDT_CMD_TBL_SZ to not exceed invalidation
command table size in IOCTL.

Signed-off-by: Jitendra Bhivare <jitendra.bhivare@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: dpt_i2o: double free if adpt_i2o_online_hba() fails
Dan Carpenter [Fri, 16 Dec 2016 09:35:39 +0000 (12:35 +0300)]
scsi: dpt_i2o: double free if adpt_i2o_online_hba() fails

There are two places where adpt_i2o_online_hba() is called.  Both
callers call adpt_i2o_delete_hba(pHba) if adpt_i2o_online_hba() fails
and since we also free it here that causes a double free bug.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: mptlan: Remove linux/miscdevice.h from mptlan.h
Corentin Labbe [Thu, 15 Dec 2016 14:12:16 +0000 (15:12 +0100)]
scsi: mptlan: Remove linux/miscdevice.h from mptlan.h

This patch remove linux/miscdevice.h from mptlan.h since mptlan.h does
not contain any miscdevice.  The only user of it is mptctl.c which
already include linux/miscdevice.h So no need to include it twice.

Signed-off-by: Corentin Labbe <clabbe.montjoie@gmail.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: cciss: use designated initializers
Kees Cook [Sat, 17 Dec 2016 01:05:30 +0000 (17:05 -0800)]
scsi: cciss: use designated initializers

Prepare to mark sensitive kernel structures for randomization by making
sure they're using designated initializers. These were identified during
allyesconfig builds of x86, arm, and arm64, with most initializer fixes
extracted from grsecurity.

Signed-off-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: hpsa: use designated initializers
Kees Cook [Sat, 17 Dec 2016 01:04:49 +0000 (17:04 -0800)]
scsi: hpsa: use designated initializers

Prepare to mark sensitive kernel structures for randomization by making
sure they're using designated initializers. These were identified during
allyesconfig builds of x86, arm, and arm64, with most initializer fixes
extracted from grsecurity.

Signed-off-by: Kees Cook <keescook@chromium.org>
Acked-by: Don Brace <don.brace@microsemi.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: lpfc: lpfc version change to 11.2.0.4
James Smart [Mon, 19 Dec 2016 23:07:31 +0000 (15:07 -0800)]
scsi: lpfc: lpfc version change to 11.2.0.4

lpfc version change to 11.2.0.4

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: lpfc: Add missing memory barrier
James Smart [Mon, 19 Dec 2016 23:07:30 +0000 (15:07 -0800)]
scsi: lpfc: Add missing memory barrier

On loosely ordered memory systems (PPC for example), the WQE elements
were being updated in memory, but not necessarily flushed before the
separate doorbell was written to hw which would cause hw to dma the
WQE element. Thus, the hardware occasionally received partially
updated WQE data.

Add the memory barrier after updating the WQE memory.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: lpfc: Correct oops on vport port resets
James Smart [Mon, 19 Dec 2016 23:07:29 +0000 (15:07 -0800)]
scsi: lpfc: Correct oops on vport port resets

Correct oops on vport port resets. Incorrect WQE type, thus the clearing
code actually overstepped the WQE.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: lpfc: Deprecate lpfc_prot_sg_seg_cnt parameter
James Smart [Mon, 19 Dec 2016 23:07:27 +0000 (15:07 -0800)]
scsi: lpfc: Deprecate lpfc_prot_sg_seg_cnt parameter

Deprecate lpfc_prot_sg_seg_cnt parameter. Eliminates driver from
unnecessarily limiting DIF s/g list length.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: lpfc: Fix Xlane dynamic LUN set for LUN priority.
James Smart [Mon, 19 Dec 2016 23:07:26 +0000 (15:07 -0800)]
scsi: lpfc: Fix Xlane dynamic LUN set for LUN priority.

Fix Xlane dynamic LUN set for LUN priority. Dynamic changing of the
priority was not getting reflected on the LUN.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: lpfc: FCoE VPort enable-disable does not bring up the VPort
James Smart [Mon, 19 Dec 2016 23:07:25 +0000 (15:07 -0800)]
scsi: lpfc: FCoE VPort enable-disable does not bring up the VPort

FCoE VPort enable-disable does not bring up the VPort.
VPI structure needed to be initialized before being re-registered.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: lpfc: Correct host name in symbolic_name field
James Smart [Mon, 19 Dec 2016 23:07:24 +0000 (15:07 -0800)]
scsi: lpfc: Correct host name in symbolic_name field

Correct host name in symbolic_name field of nameserver registrations

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: lpfc: Correct issue leading to oops during link reset
James Smart [Mon, 19 Dec 2016 23:07:23 +0000 (15:07 -0800)]
scsi: lpfc: Correct issue leading to oops during link reset

Correct issue leading to oops during link reset. Missing vport pointer.

[mkp: fixed typo]

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: lpfc: Deprecate lpfc_soft_wwn parameter
James Smart [Mon, 19 Dec 2016 23:07:22 +0000 (15:07 -0800)]
scsi: lpfc: Deprecate lpfc_soft_wwn parameter

Deprecate lpfc_soft_wwn parameter.
No longer allow override of hw-assigned wwns

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: lpfc: Correct error in setting OS Driver Version with FW
James Smart [Mon, 19 Dec 2016 23:07:21 +0000 (15:07 -0800)]
scsi: lpfc: Correct error in setting OS Driver Version with FW

Correct error in setting OS Driver Version with FW.  Prior length was
too short.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@Suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: lpfc: Clear the VendorVersion in the PLOGI/PLOGI ACC payload
James Smart [Mon, 19 Dec 2016 23:07:20 +0000 (15:07 -0800)]
scsi: lpfc: Clear the VendorVersion in the PLOGI/PLOGI ACC payload

Clear the VendorVersion in the PLOGI/PLOGI ACC payload

Vendor version info may have been set on fabric login. Before sending
PLOGI payloads, ensure that it's cleared.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: storvsc: properly set residual data length on errors
Long Li [Thu, 15 Dec 2016 02:46:03 +0000 (18:46 -0800)]
scsi: storvsc: properly set residual data length on errors

On I/O errors, the Windows driver doesn't set data_transfer_length
on error conditions other than SRB_STATUS_DATA_OVERRUN.
In these cases we need to set data_transfer_length to 0,
indicating there is no data transferred. On SRB_STATUS_DATA_OVERRUN,
data_transfer_length is set by the Windows driver to the actual data transferred.

Reported-by: Shiva Krishna <Shiva.Krishna@nimblestorage.com>
Signed-off-by: Long Li <longli@microsoft.com>
Reviewed-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: storvsc: properly handle SRB_ERROR when sense message is present
Long Li [Thu, 15 Dec 2016 02:46:02 +0000 (18:46 -0800)]
scsi: storvsc: properly handle SRB_ERROR when sense message is present

When sense message is present on error, we should pass along to the upper
layer to decide how to deal with the error.
This patch fixes connectivity issues with Fiber Channel devices.

Signed-off-by: Long Li <longli@microsoft.com>
Reviewed-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: storvsc: use tagged SRB requests if supported by the device
Long Li [Thu, 15 Dec 2016 02:46:01 +0000 (18:46 -0800)]
scsi: storvsc: use tagged SRB requests if supported by the device

Properly set SRB flags when hosting device supports tagged queuing.
This patch improves the performance on Fiber Channel disks.

Signed-off-by: Long Li <longli@microsoft.com>
Reviewed-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: storvsc: Enable multi-queue support
K. Y. Srinivasan [Thu, 15 Dec 2016 02:46:00 +0000 (18:46 -0800)]
scsi: storvsc: Enable multi-queue support

Enable multi-q support. We will allocate the outgoing channel using
the following policy:

        1. We will make every effort to pick a channel that is in the
           same NUMA node that is initiating the I/O
        2. The mapping between the guest CPU and the outgoing channel
           is persistent.

Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Reviewed-by: Long Li <longli@microsoft.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: storvsc: Remove the restriction on max segment size
K. Y. Srinivasan [Thu, 15 Dec 2016 02:45:59 +0000 (18:45 -0800)]
scsi: storvsc: Remove the restriction on max segment size

Remove the artificially imposed restriction on max segment size.

Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Reviewed-by: Long Li <longli@microsoft.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoscsi: storvsc: Enable tracking of queue depth
K. Y. Srinivasan [Thu, 15 Dec 2016 02:45:58 +0000 (18:45 -0800)]
scsi: storvsc: Enable tracking of queue depth

Enable tracking of queue depth.

Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Reviewed-by: Long Li <longli@microsoft.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
7 years agoLinux 4.10-rc2
Linus Torvalds [Sun, 1 Jan 2017 22:31:53 +0000 (14:31 -0800)]
Linux 4.10-rc2

7 years agoMerge branch 'libnvdimm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdim...
Linus Torvalds [Sun, 1 Jan 2017 20:27:05 +0000 (12:27 -0800)]
Merge branch 'libnvdimm-fixes' of git://git./linux/kernel/git/nvdimm/nvdimm

Pull DAX updates from Dan Williams:
 "The completion of Jan's DAX work for 4.10.

  As I mentioned in the libnvdimm-for-4.10 pull request, these are some
  final fixes for the DAX dirty-cacheline-tracking invalidation work
  that was merged through the -mm, ext4, and xfs trees in -rc1. These
  patches were prepared prior to the merge window, but we waited for
  4.10-rc1 to have a stable merge base after all the prerequisites were
  merged.

  Quoting Jan on the overall changes in these patches:

     "So I'd like all these 6 patches to go for rc2. The first three
      patches fix invalidation of exceptional DAX entries (a bug which
      is there for a long time) - without these patches data loss can
      occur on power failure even though user called fsync(2). The other
      three patches change locking of DAX faults so that ->iomap_begin()
      is called in a more relaxed locking context and we are safe to
      start a transaction there for ext4"

  These have received a build success notification from the kbuild
  robot, and pass the latest libnvdimm unit tests. There have not been
  any -next releases since -rc1, so they have not appeared there"

* 'libnvdimm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm:
  ext4: Simplify DAX fault path
  dax: Call ->iomap_begin without entry lock during dax fault
  dax: Finish fault completely when loading holes
  dax: Avoid page invalidation races and unnecessary radix tree traversals
  mm: Invalidate DAX radix tree entries only if appropriate
  ext2: Return BH_New buffers for zeroed blocks

7 years agoMerge tag 'docs-4.10-rc1-fix' of git://git.lwn.net/linux
Linus Torvalds [Fri, 30 Dec 2016 17:32:26 +0000 (09:32 -0800)]
Merge tag 'docs-4.10-rc1-fix' of git://git.lwn.net/linux

Pull documentation fixes from Jonathan Corbet:
 "Two small fixes:

   - A merge error on my part broke the DocBook build. I've
     requisitioned one of tglx's frozen sharks for appropriate
     disciplinary action and resolved to be more careful about testing
     the DocBook stuff as long as it's still around.

   - Fix an error in unaligned-memory-access.txt"

* tag 'docs-4.10-rc1-fix' of git://git.lwn.net/linux:
  Documentation/unaligned-memory-access.txt: fix incorrect comparison operator
  docs: Fix build failure

7 years agoMerge branch 'linus' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6
Linus Torvalds [Fri, 30 Dec 2016 17:29:50 +0000 (09:29 -0800)]
Merge branch 'linus' of git://git./linux/kernel/git/herbert/crypto-2.6

Pull crypto fix from Herbert Xu:
 "This fixes a boot failure on some platforms when crypto self test is
  enabled along with the new acomp interface"

* 'linus' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6:
  crypto: testmgr - Use heap buffer for acomp test input

8 years agomm/filemap: fix parameters to test_bit()
Olof Johansson [Thu, 29 Dec 2016 22:16:07 +0000 (14:16 -0800)]
mm/filemap: fix parameters to test_bit()

 mm/filemap.c: In function 'clear_bit_unlock_is_negative_byte':
  mm/filemap.c:933:9: error: too few arguments to function 'test_bit'
    return test_bit(PG_waiters);
         ^~~~~~~~

Fixes: b91e1302ad9b ('mm: optimize PageWaiters bit use for unlock_page()')
Signed-off-by: Olof Johansson <olof@lixom.net>
Brown-paper-bag-by: Linus Torvalds <dummy@duh.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
8 years agomm: optimize PageWaiters bit use for unlock_page()
Linus Torvalds [Tue, 27 Dec 2016 19:40:38 +0000 (11:40 -0800)]
mm: optimize PageWaiters bit use for unlock_page()

In commit 62906027091f ("mm: add PageWaiters indicating tasks are
waiting for a page bit") Nick Piggin made our page locking no longer
unconditionally touch the hashed page waitqueue, which not only helps
performance in general, but is particularly helpful on NUMA machines
where the hashed wait queues can bounce around a lot.

However, the "clear lock bit atomically and then test the waiters bit"
sequence turns out to be much more expensive than it needs to be,
because you get a nasty stall when trying to access the same word that
just got updated atomically.

On architectures where locking is done with LL/SC, this would be trivial
to fix with a new primitive that clears one bit and tests another
atomically, but that ends up not working on x86, where the only atomic
operations that return the result end up being cmpxchg and xadd.  The
atomic bit operations return the old value of the same bit we changed,
not the value of an unrelated bit.

On x86, we could put the lock bit in the high bit of the byte, and use
"xadd" with that bit (where the overflow ends up not touching other
bits), and look at the other bits of the result.  However, an even
simpler model is to just use a regular atomic "and" to clear the lock
bit, and then the sign bit in eflags will indicate the resulting state
of the unrelated bit #7.

So by moving the PageWaiters bit up to bit #7, we can atomically clear
the lock bit and test the waiters bit on x86 too.  And architectures
with LL/SC (which is all the usual RISC suspects), the particular bit
doesn't matter, so they are fine with this approach too.

This avoids the extra access to the same atomic word, and thus avoids
the costly stall at page unlock time.

The only downside is that the interface ends up being a bit odd and
specialized: clear a bit in a byte, and test the sign bit.  Nick doesn't
love the resulting name of the new primitive, but I'd rather make the
name be descriptive and very clear about the limitation imposed by
trying to work across all relevant architectures than make it be some
generic thing that doesn't make the odd semantics explicit.

So this introduces the new architecture primitive

    clear_bit_unlock_is_negative_byte();

and adds the trivial implementation for x86.  We have a generic
non-optimized fallback (that just does a "clear_bit()"+"test_bit(7)"
combination) which can be overridden by any architecture that can do
better.  According to Nick, Power has the same hickup x86 has, for
example, but some other architectures may not even care.

All these optimizations mean that my page locking stress-test (which is
just executing a lot of small short-lived shell scripts: "make test" in
the git source tree) no longer makes our page locking look horribly bad.
Before all these optimizations, just the unlock_page() costs were just
over 3% of all CPU overhead on "make test".  After this, it's down to
0.66%, so just a quarter of the cost it used to be.

(The difference on NUMA is bigger, but there this micro-optimization is
likely less noticeable, since the big issue on NUMA was not the accesses
to 'struct page', but the waitqueue accesses that were already removed
by Nick's earlier commit).

Acked-by: Nick Piggin <npiggin@gmail.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Bob Peterson <rpeterso@redhat.com>
Cc: Steven Whitehouse <swhiteho@redhat.com>
Cc: Andrew Lutomirski <luto@kernel.org>
Cc: Andreas Gruenbacher <agruenba@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
8 years agoMerge branch 'linus' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6
Linus Torvalds [Wed, 28 Dec 2016 01:51:36 +0000 (17:51 -0800)]
Merge branch 'linus' of git://git./linux/kernel/git/herbert/crypto-2.6

Pull crypto fix from Herbert Xu:
 "This fixes a hash corruption bug in the marvell driver"

* 'linus' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6:
  crypto: marvell - Copy IVDIG before launching partial DMA ahash requests

8 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Wed, 28 Dec 2016 00:04:37 +0000 (16:04 -0800)]
Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:

 1) Various ipvlan fixes from Eric Dumazet and Mahesh Bandewar.

    The most important is to not assume the packet is RX just because
    the destination address matches that of the device. Such an
    assumption causes problems when an interface is put into loopback
    mode.

 2) If we retry when creating a new tc entry (because we dropped the
    RTNL mutex in order to load a module, for example) we end up with
    -EAGAIN and then loop trying to replay the request. But we didn't
    reset some state when looping back to the top like this, and if
    another thread meanwhile inserted the same tc entry we were trying
    to, we re-link it creating an enless loop in the tc chain. Fix from
    Daniel Borkmann.

 3) There are two different WRITE bits in the MDIO address register for
    the stmmac chip, depending upon the chip variant. Due to a bug we
    could set them both, fix from Hock Leong Kweh.

 4) Fix mlx4 bug in XDP_TX handling, from Tariq Toukan.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net:
  net: stmmac: fix incorrect bit set in gmac4 mdio addr register
  r8169: add support for RTL8168 series add-on card.
  net: xdp: remove unused bfp_warn_invalid_xdp_buffer()
  openvswitch: upcall: Fix vlan handling.
  ipv4: Namespaceify tcp_tw_reuse knob
  net: korina: Fix NAPI versus resources freeing
  net, sched: fix soft lockup in tc_classify
  net/mlx4_en: Fix user prio field in XDP forward
  tipc: don't send FIN message from connectionless socket
  ipvlan: fix multicast processing
  ipvlan: fix various issues in ipvlan_process_multicast()

8 years agoDocumentation/unaligned-memory-access.txt: fix incorrect comparison operator
Cihangir Akturk [Sat, 17 Dec 2016 17:42:17 +0000 (19:42 +0200)]
Documentation/unaligned-memory-access.txt: fix incorrect comparison operator

In the actual implementation ether_addr_equal function tests for equality to 0
when returning. It seems in commit 0d74c4 it is somehow overlooked to change
this operator to reflect the actual function.

Signed-off-by: Cihangir Akturk <cakturk@gmail.com>
Signed-off-by: Jonathan Corbet <corbet@lwn.net>
8 years agodocs: Fix build failure
John Brooks [Fri, 23 Dec 2016 00:53:10 +0000 (00:53 +0000)]
docs: Fix build failure

The 80211.tmpl DocBook file was removed in commit 819bf593767c ("docs-rst:
sphinxify 802.11 documentation"), but the 80211.xml target was re-added to
the Makefile by commit 7ddedebb03b7 ("ALSA: doc: ReSTize
writing-an-alsa-driver document"), leading to a failure when building the
documentation:

*** No rule to make target 'Documentation/DocBook/80211.xml', needed by
'Documentation/DocBook/80211.aux.xml'.

cc: stable@vger.kernel.org
Signed-off-by: John Brooks <john@fastquake.com>
Mea-culpa-by: Jonathan Corbet <corbet@lwn.net>
Signed-off-by: Jonathan Corbet <corbet@lwn.net>
8 years agoMerge tag 'v4.10-rc1' into docs-next
Jonathan Corbet [Tue, 27 Dec 2016 19:53:44 +0000 (12:53 -0700)]
Merge tag 'v4.10-rc1' into docs-next

Linux 4.10-rc1

8 years agonet: stmmac: fix incorrect bit set in gmac4 mdio addr register
Kweh, Hock Leong [Tue, 27 Dec 2016 20:07:41 +0000 (04:07 +0800)]
net: stmmac: fix incorrect bit set in gmac4 mdio addr register

Fixing the gmac4 mdio write access to use MII_GMAC4_WRITE only instead of
OR together with MII_WRITE.

Signed-off-by: Kweh, Hock Leong <hock.leong.kweh@intel.com>
Acked-By: Joao Pinto <jpinto@synopsys.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agor8169: add support for RTL8168 series add-on card.
Chun-Hao Lin [Tue, 27 Dec 2016 08:29:43 +0000 (16:29 +0800)]
r8169: add support for RTL8168 series add-on card.

This chip is the same as RTL8168, but its device id is 0x8161.

Signed-off-by: Chun-Hao Lin <hau@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: xdp: remove unused bfp_warn_invalid_xdp_buffer()
Jason Wang [Tue, 27 Dec 2016 02:49:54 +0000 (10:49 +0800)]
net: xdp: remove unused bfp_warn_invalid_xdp_buffer()

After commit 73b62bd085f4737679ea9afc7867fa5f99ba7d1b ("virtio-net:
remove the warning before XDP linearizing"), there's no users for
bpf_warn_invalid_xdp_buffer(), so remove it. This is a revert for
commit f23bc46c30ca5ef58b8549434899fcbac41b2cfc.

Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoopenvswitch: upcall: Fix vlan handling.
pravin shelar [Mon, 26 Dec 2016 16:31:27 +0000 (08:31 -0800)]
openvswitch: upcall: Fix vlan handling.

Networking stack accelerate vlan tag handling by
keeping topmost vlan header in skb. This works as
long as packet remains in OVS datapath. But during
OVS upcall vlan header is pushed on to the packet.
When such packet is sent back to OVS datapath, core
networking stack might not handle it correctly. Following
patch avoids this issue by accelerating the vlan tag
during flow key extract. This simplifies datapath by
bringing uniform packet processing for packets from
all code paths.

Fixes: 5108bbaddc ("openvswitch: add processing of L3 packets").
CC: Jarno Rajahalme <jarno@ovn.org>
CC: Jiri Benc <jbenc@redhat.com>
Signed-off-by: Pravin B Shelar <pshelar@ovn.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoipv4: Namespaceify tcp_tw_reuse knob
Haishuang Yan [Sun, 25 Dec 2016 06:33:16 +0000 (14:33 +0800)]
ipv4: Namespaceify tcp_tw_reuse knob

Different namespaces might have different requirements to reuse
TIME-WAIT sockets for new connections. This might be required in
cases where different namespace applications are in place which
require TIME_WAIT socket connections to be reduced independently
of the host.

Signed-off-by: Haishuang Yan <yanhaishuang@cmss.chinamobile.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agocrypto: testmgr - Use heap buffer for acomp test input
Laura Abbott [Wed, 21 Dec 2016 20:32:54 +0000 (12:32 -0800)]
crypto: testmgr - Use heap buffer for acomp test input

Christopher Covington reported a crash on aarch64 on recent Fedora
kernels:

kernel BUG at ./include/linux/scatterlist.h:140!
Internal error: Oops - BUG: 0 [#1] PREEMPT SMP
Modules linked in:
CPU: 2 PID: 752 Comm: cryptomgr_test Not tainted 4.9.0-11815-ge93b1cc #162
Hardware name: linux,dummy-virt (DT)
task: ffff80007c650080 task.stack: ffff800008910000
PC is at sg_init_one+0xa0/0xb8
LR is at sg_init_one+0x24/0xb8
...
[<ffff000008398db8>] sg_init_one+0xa0/0xb8
[<ffff000008350a44>] test_acomp+0x10c/0x438
[<ffff000008350e20>] alg_test_comp+0xb0/0x118
[<ffff00000834f28c>] alg_test+0x17c/0x2f0
[<ffff00000834c6a4>] cryptomgr_test+0x44/0x50
[<ffff0000080dac70>] kthread+0xf8/0x128
[<ffff000008082ec0>] ret_from_fork+0x10/0x50

The test vectors used for input are part of the kernel image. These
inputs are passed as a buffer to sg_init_one which eventually blows up
with BUG_ON(!virt_addr_valid(buf)). On arm64, virt_addr_valid returns
false for the kernel image since virt_to_page will not return the
correct page. Fix this by copying the input vectors to heap buffer
before setting up the scatterlist.

Reported-by: Christopher Covington <cov@codeaurora.org>
Fixes: d7db7a882deb ("crypto: acomp - update testmgr with support for acomp")
Signed-off-by: Laura Abbott <labbott@redhat.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
8 years agoext4: Simplify DAX fault path
Jan Kara [Fri, 21 Oct 2016 09:33:49 +0000 (11:33 +0200)]
ext4: Simplify DAX fault path

Now that dax_iomap_fault() calls ->iomap_begin() without entry lock, we
can use transaction starting in ext4_iomap_begin() and thus simplify
ext4_dax_fault(). It also provides us proper retries in case of ENOSPC.

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
8 years agodax: Call ->iomap_begin without entry lock during dax fault
Jan Kara [Wed, 19 Oct 2016 12:34:31 +0000 (14:34 +0200)]
dax: Call ->iomap_begin without entry lock during dax fault

Currently ->iomap_begin() handler is called with entry lock held. If the
filesystem held any locks between ->iomap_begin() and ->iomap_end()
(such as ext4 which will want to hold transaction open), this would cause
lock inversion with the iomap_apply() from standard IO path which first
calls ->iomap_begin() and only then calls ->actor() callback which grabs
entry locks for DAX (if it faults when copying from/to user provided
buffers).

Fix the problem by nesting grabbing of entry lock inside ->iomap_begin()
- ->iomap_end() pair.

Reviewed-by: Ross Zwisler <ross.zwisler@linux.intel.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
8 years agodax: Finish fault completely when loading holes
Jan Kara [Wed, 19 Oct 2016 12:48:38 +0000 (14:48 +0200)]
dax: Finish fault completely when loading holes

The only case when we do not finish the page fault completely is when we
are loading hole pages into a radix tree. Avoid this special case and
finish the fault in that case as well inside the DAX fault handler. It
will allow us for easier iomap handling.

Reviewed-by: Ross Zwisler <ross.zwisler@linux.intel.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
8 years agodax: Avoid page invalidation races and unnecessary radix tree traversals
Jan Kara [Wed, 10 Aug 2016 15:10:28 +0000 (17:10 +0200)]
dax: Avoid page invalidation races and unnecessary radix tree traversals

Currently dax_iomap_rw() takes care of invalidating page tables and
evicting hole pages from the radix tree when write(2) to the file
happens. This invalidation is only necessary when there is some block
allocation resulting from write(2). Furthermore in current place the
invalidation is racy wrt page fault instantiating a hole page just after
we have invalidated it.

So perform the page invalidation inside dax_iomap_actor() where we can
do it only when really necessary and after blocks have been allocated so
nobody will be instantiating new hole pages anymore.

Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Ross Zwisler <ross.zwisler@linux.intel.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
8 years agomm: Invalidate DAX radix tree entries only if appropriate
Jan Kara [Wed, 10 Aug 2016 15:22:44 +0000 (17:22 +0200)]
mm: Invalidate DAX radix tree entries only if appropriate

Currently invalidate_inode_pages2_range() and invalidate_mapping_pages()
just delete all exceptional radix tree entries they find. For DAX this
is not desirable as we track cache dirtiness in these entries and when
they are evicted, we may not flush caches although it is necessary. This
can for example manifest when we write to the same block both via mmap
and via write(2) (to different offsets) and fsync(2) then does not
properly flush CPU caches when modification via write(2) was the last
one.

Create appropriate DAX functions to handle invalidation of DAX entries
for invalidate_inode_pages2_range() and invalidate_mapping_pages() and
wire them up into the corresponding mm functions.

Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: Ross Zwisler <ross.zwisler@linux.intel.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
8 years agoext2: Return BH_New buffers for zeroed blocks
Jan Kara [Wed, 10 Aug 2016 14:42:53 +0000 (16:42 +0200)]
ext2: Return BH_New buffers for zeroed blocks

So far we did not return BH_New buffers from ext2_get_blocks() when we
allocated and zeroed-out a block for DAX inode to avoid racy zeroing in
DAX code. This zeroing is gone these days so we can remove the
workaround.

Reviewed-by: Ross Zwisler <ross.zwisler@linux.intel.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
8 years agox86/mce/AMD: Make the init code more robust
Thomas Gleixner [Mon, 26 Dec 2016 21:58:20 +0000 (22:58 +0100)]
x86/mce/AMD: Make the init code more robust

If mce_device_init() fails then the mce device pointer is NULL and the
AMD mce code happily dereferences it.

Add a sanity check.

Reported-by: Markus Trippelsdorf <markus@trippelsdorf.de>
Reported-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
8 years agosmp/hotplug: Undo tglxs brainfart
Thomas Gleixner [Mon, 26 Dec 2016 21:58:19 +0000 (22:58 +0100)]
smp/hotplug: Undo tglxs brainfart

The attempt to prevent overwriting an active state resulted in a
disaster which effectively disables all dynamically allocated hotplug
states.

Cleanup the mess.

Fixes: dc280d936239 ("cpu/hotplug: Prevent overwriting of callbacks")
Reported-by: Markus Trippelsdorf <markus@trippelsdorf.de>
Reported-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
8 years agoarm64: don't pull uaccess.h into *.S
Al Viro [Mon, 26 Dec 2016 09:10:19 +0000 (04:10 -0500)]
arm64: don't pull uaccess.h into *.S

Split asm-only parts of arm64 uaccess.h into a new header and use that
from *.S.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
8 years agonet: korina: Fix NAPI versus resources freeing
Florian Fainelli [Sat, 24 Dec 2016 03:56:56 +0000 (19:56 -0800)]
net: korina: Fix NAPI versus resources freeing

Commit beb0babfb77e ("korina: disable napi on close and restart")
introduced calls to napi_disable() that were missing before,
unfortunately this leaves a small window during which NAPI has a chance
to run, yet we just freed resources since korina_free_ring() has been
called:

Fix this by disabling NAPI first then freeing resource, and make sure
that we also cancel the restart task before doing the resource freeing.

Fixes: beb0babfb77e ("korina: disable napi on close and restart")
Reported-by: Alexandros C. Couloumbis <alex@ozo.com>
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet, sched: fix soft lockup in tc_classify
Daniel Borkmann [Wed, 21 Dec 2016 17:04:11 +0000 (18:04 +0100)]
net, sched: fix soft lockup in tc_classify

Shahar reported a soft lockup in tc_classify(), where we run into an
endless loop when walking the classifier chain due to tp->next == tp
which is a state we should never run into. The issue only seems to
trigger under load in the tc control path.

What happens is that in tc_ctl_tfilter(), thread A allocates a new
tp, initializes it, sets tp_created to 1, and calls into tp->ops->change()
with it. In that classifier callback we had to unlock/lock the rtnl
mutex and returned with -EAGAIN. One reason why we need to drop there
is, for example, that we need to request an action module to be loaded.

This happens via tcf_exts_validate() -> tcf_action_init/_1() meaning
after we loaded and found the requested action, we need to redo the
whole request so we don't race against others. While we had to unlock
rtnl in that time, thread B's request was processed next on that CPU.
Thread B added a new tp instance successfully to the classifier chain.
When thread A returned grabbing the rtnl mutex again, propagating -EAGAIN
and destroying its tp instance which never got linked, we goto replay
and redo A's request.

This time when walking the classifier chain in tc_ctl_tfilter() for
checking for existing tp instances we had a priority match and found
the tp instance that was created and linked by thread B. Now calling
again into tp->ops->change() with that tp was successful and returned
without error.

tp_created was never cleared in the second round, thus kernel thinks
that we need to link it into the classifier chain (once again). tp and
*back point to the same object due to the match we had earlier on. Thus
for thread B's already public tp, we reset tp->next to tp itself and
link it into the chain, which eventually causes the mentioned endless
loop in tc_classify() once a packet hits the data path.

Fix is to clear tp_created at the beginning of each request, also when
we replay it. On the paths that can cause -EAGAIN we already destroy
the original tp instance we had and on replay we really need to start
from scratch. It seems that this issue was first introduced in commit
12186be7d2e1 ("net_cls: fix unconfigured struct tcf_proto keeps chaining
and avoid kernel panic when we use cls_cgroup").

Fixes: 12186be7d2e1 ("net_cls: fix unconfigured struct tcf_proto keeps chaining and avoid kernel panic when we use cls_cgroup")
Reported-by: Shahar Klein <shahark@mellanox.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Tested-by: Shahar Klein <shahark@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoLinux 4.10-rc1
Linus Torvalds [Mon, 26 Dec 2016 00:13:08 +0000 (16:13 -0800)]
Linux 4.10-rc1

8 years agopowerpc: Fix build warning on 32-bit PPC
Larry Finger [Fri, 23 Dec 2016 03:06:53 +0000 (21:06 -0600)]
powerpc: Fix build warning on 32-bit PPC

I am getting the following warning when I build kernel 4.9-git on my
PowerBook G4 with a 32-bit PPC processor:

    AS      arch/powerpc/kernel/misc_32.o
  arch/powerpc/kernel/misc_32.S:299:7: warning: "CONFIG_FSL_BOOKE" is not defined [-Wundef]

This problem is evident after commit 989cea5c14be ("kbuild: prevent
lib-ksyms.o rebuilds"); however, this change in kbuild only exposes an
error that has been in the code since 2005 when this source file was
created.  That was with commit 9994a33865f4 ("powerpc: Introduce
entry_{32,64}.S, misc_{32,64}.S, systbl.S").

The offending line does not make a lot of sense.  This error does not
seem to cause any errors in the executable, thus I am not recommending
that it be applied to any stable versions.

Thanks to Nicholas Piggin for suggesting this solution.

Fixes: 9994a33865f4 ("powerpc: Introduce entry_{32,64}.S, misc_{32,64}.S, systbl.S")
Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Cc: Nicholas Piggin <npiggin@gmail.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: linuxppc-dev@lists.ozlabs.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
8 years agoavoid spurious "may be used uninitialized" warning
Linus Torvalds [Sun, 25 Dec 2016 22:56:58 +0000 (14:56 -0800)]
avoid spurious "may be used uninitialized" warning

The timer type simplifications caused a new gcc warning:

  drivers/base/power/domain.c: In function ‘genpd_runtime_suspend’:
  drivers/base/power/domain.c:562:14: warning: ‘time_start’ may be used uninitialized in this function [-Wmaybe-uninitialized]
     elapsed_ns = ktime_to_ns(ktime_sub(ktime_get(), time_start));

despite the actual use of "time_start" not having changed in any way.
It appears that simply changing the type of ktime_t from a union to a
plain scalar type made gcc check the use.

The variable wasn't actually used uninitialized, but gcc apparently
failed to notice that the conditional around the use was exactly the
same as the conditional around the initialization of that variable.

Add an unnecessary initialization just to shut up the compiler.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
8 years agoMerge branch 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 25 Dec 2016 22:30:04 +0000 (14:30 -0800)]
Merge branch 'timers-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull timer type cleanups from Thomas Gleixner:
 "This series does a tree wide cleanup of types related to
  timers/timekeeping.

   - Get rid of cycles_t and use a plain u64. The type is not really
     helpful and caused more confusion than clarity

   - Get rid of the ktime union. The union has become useless as we use
     the scalar nanoseconds storage unconditionally now. The 32bit
     timespec alike storage got removed due to the Y2038 limitations
     some time ago.

     That leaves the odd union access around for no reason. Clean it up.

  Both changes have been done with coccinelle and a small amount of
  manual mopping up"

* 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  ktime: Get rid of ktime_equal()
  ktime: Cleanup ktime_set() usage
  ktime: Get rid of the union
  clocksource: Use a plain u64 instead of cycle_t

8 years agoMerge branch 'smp-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 25 Dec 2016 22:05:56 +0000 (14:05 -0800)]
Merge branch 'smp-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull SMP hotplug notifier removal from Thomas Gleixner:
 "This is the final cleanup of the hotplug notifier infrastructure. The
  series has been reintgrated in the last two days because there came a
  new driver using the old infrastructure via the SCSI tree.

  Summary:

   - convert the last leftover drivers utilizing notifiers

   - fixup for a completely broken hotplug user

   - prevent setup of already used states

   - removal of the notifiers

   - treewide cleanup of hotplug state names

   - consolidation of state space

  There is a sphinx based documentation pending, but that needs review
  from the documentation folks"

* 'smp-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  irqchip/armada-xp: Consolidate hotplug state space
  irqchip/gic: Consolidate hotplug state space
  coresight/etm3/4x: Consolidate hotplug state space
  cpu/hotplug: Cleanup state names
  cpu/hotplug: Remove obsolete cpu hotplug register/unregister functions
  staging/lustre/libcfs: Convert to hotplug state machine
  scsi/bnx2i: Convert to hotplug state machine
  scsi/bnx2fc: Convert to hotplug state machine
  cpu/hotplug: Prevent overwriting of callbacks
  x86/msr: Remove bogus cleanup from the error path
  bus: arm-ccn: Prevent hotplug callback leak
  perf/x86/intel/cstate: Prevent hotplug callback leak
  ARM/imx/mmcd: Fix broken cpu hotplug handling
  scsi: qedi: Convert to hotplug state machine

8 years agoMerge branch 'turbostat' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux
Linus Torvalds [Sun, 25 Dec 2016 22:01:28 +0000 (14:01 -0800)]
Merge branch 'turbostat' of git://git./linux/kernel/git/lenb/linux

Pull turbostat updates from Len Brown.

* 'turbostat' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux:
  tools/power turbostat: remove obsolete -M, -m, -C, -c options
  tools/power turbostat: Make extensible via the --add parameter
  tools/power turbostat: Denverton uses a 25 MHz crystal, not 19.2 MHz
  tools/power turbostat: line up headers when -M is used
  tools/power turbostat: fix SKX PKG_CSTATE_LIMIT decoding
  tools/power turbostat: Support Knights Mill (KNM)
  tools/power turbostat: Display HWP OOB status
  tools/power turbostat: fix Denverton BCLK
  tools/power turbostat: use intel-family.h model strings
  tools/power/turbostat: Add Denverton RAPL support
  tools/power/turbostat: Add Denverton support
  tools/power/turbostat: split core MSR support into status + limit
  tools/power turbostat: fix error case overflow read of slm_freq_table[]
  tools/power turbostat: Allocate correct amount of fd and irq entries
  tools/power turbostat: switch to tab delimited output
  tools/power turbostat: Gracefully handle ACPI S3
  tools/power turbostat: tidy up output on Joule counter overflow

8 years agomm: add PageWaiters indicating tasks are waiting for a page bit
Nicholas Piggin [Sun, 25 Dec 2016 03:00:30 +0000 (13:00 +1000)]
mm: add PageWaiters indicating tasks are waiting for a page bit

Add a new page flag, PageWaiters, to indicate the page waitqueue has
tasks waiting. This can be tested rather than testing waitqueue_active
which requires another cacheline load.

This bit is always set when the page has tasks on page_waitqueue(page),
and is set and cleared under the waitqueue lock. It may be set when
there are no tasks on the waitqueue, which will cause a harmless extra
wakeup check that will clears the bit.

The generic bit-waitqueue infrastructure is no longer used for pages.
Instead, waitqueues are used directly with a custom key type. The
generic code was not flexible enough to have PageWaiters manipulation
under the waitqueue lock (which simplifies concurrency).

This improves the performance of page lock intensive microbenchmarks by
2-3%.

Putting two bits in the same word opens the opportunity to remove the
memory barrier between clearing the lock bit and testing the waiters
bit, after some work on the arch primitives (e.g., ensuring memory
operand widths match and cover both bits).

Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Bob Peterson <rpeterso@redhat.com>
Cc: Steven Whitehouse <swhiteho@redhat.com>
Cc: Andrew Lutomirski <luto@kernel.org>
Cc: Andreas Gruenbacher <agruenba@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
8 years agomm: Use owner_priv bit for PageSwapCache, valid when PageSwapBacked
Nicholas Piggin [Sun, 25 Dec 2016 03:00:29 +0000 (13:00 +1000)]
mm: Use owner_priv bit for PageSwapCache, valid when PageSwapBacked

A page is not added to the swap cache without being swap backed,
so PageSwapBacked mappings can use PG_owner_priv_1 for PageSwapCache.

Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Acked-by: Hugh Dickins <hughd@google.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Bob Peterson <rpeterso@redhat.com>
Cc: Steven Whitehouse <swhiteho@redhat.com>
Cc: Andrew Lutomirski <luto@kernel.org>
Cc: Andreas Gruenbacher <agruenba@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
8 years agoktime: Get rid of ktime_equal()
Thomas Gleixner [Sun, 25 Dec 2016 11:43:07 +0000 (12:43 +0100)]
ktime: Get rid of ktime_equal()

No point in going through loops and hoops instead of just comparing the
values.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
8 years agoktime: Cleanup ktime_set() usage
Thomas Gleixner [Sun, 25 Dec 2016 11:30:41 +0000 (12:30 +0100)]
ktime: Cleanup ktime_set() usage

ktime_set(S,N) was required for the timespec storage type and is still
useful for situations where a Seconds and Nanoseconds part of a time value
needs to be converted. For anything where the Seconds argument is 0, this
is pointless and can be replaced with a simple assignment.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
8 years agoktime: Get rid of the union
Thomas Gleixner [Sun, 25 Dec 2016 10:38:40 +0000 (11:38 +0100)]
ktime: Get rid of the union

ktime is a union because the initial implementation stored the time in
scalar nanoseconds on 64 bit machine and in a endianess optimized timespec
variant for 32bit machines. The Y2038 cleanup removed the timespec variant
and switched everything to scalar nanoseconds. The union remained, but
become completely pointless.

Get rid of the union and just keep ktime_t as simple typedef of type s64.

The conversion was done with coccinelle and some manual mopping up.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
8 years agoclocksource: Use a plain u64 instead of cycle_t
Thomas Gleixner [Wed, 21 Dec 2016 19:32:01 +0000 (20:32 +0100)]
clocksource: Use a plain u64 instead of cycle_t

There is no point in having an extra type for extra confusion. u64 is
unambiguous.

Conversion was done with the following coccinelle script:

@rem@
@@
-typedef u64 cycle_t;

@fix@
typedef cycle_t;
@@
-cycle_t
+u64

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: John Stultz <john.stultz@linaro.org>
8 years agoirqchip/armada-xp: Consolidate hotplug state space
Thomas Gleixner [Wed, 21 Dec 2016 19:19:57 +0000 (20:19 +0100)]
irqchip/armada-xp: Consolidate hotplug state space

The mpic is either the main interrupt controller or is cascaded behind a
GIC. The mpic is single instance and the modes are mutually exclusive, so
there is no reason to have seperate cpu hotplug states.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Sebastian Siewior <bigeasy@linutronix.de>
Cc: Marc Zyngier <marc.zyngier@arm.com>
Link: http://lkml.kernel.org/r/20161221192112.333161745@linutronix.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
8 years agoirqchip/gic: Consolidate hotplug state space
Thomas Gleixner [Wed, 21 Dec 2016 19:19:56 +0000 (20:19 +0100)]
irqchip/gic: Consolidate hotplug state space

Even if both drivers are compiled in only one instance can run on a given
system depending on the available GIC version.

So having seperate hotplug states for them is pointless.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Marc Zyngier <marc.zyngier@arm.com>
Cc: Sebastian Siewior <bigeasy@linutronix.de>
Link: http://lkml.kernel.org/r/20161221192112.252416267@linutronix.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>