Re-order struct members so that hot data lies at the beginning of the
struct and cold data at the end. Improve the comments while we're here.
Signed-off-by: NFinn Thain <fthain@telegraphics.com.au>
Reviewed-by: NHannes Reinecke <hare@suse.com>
Tested-by: NOndrej Zary <linux@rainbow-software.org>
Tested-by: NMichael Schmitz <schmitzmic@gmail.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

If NCR5380_poll_politely() is called under irq lock, the polling time
limit is clamped to avoid a spike in interrupt latency. When not under
irq lock, the same polling time limit acts as the worst case delay
between schedule() calls.

During PDMA (under irq lock) I've found that the 10 ms time limit is
sometimes too short, and leads to the error message,
sd 0:0:0:0: [sda] tag#1 macscsi_pread: !REQ and !ACK

This particular target identifies itself as a QUANTUM DAYTONA514S. It
seems to be slower to assert ACK than the other targets I've tested.
This patch solves the problem by increasing the polling timeout.
Signed-off-by: NFinn Thain <fthain@telegraphics.com.au>
Reviewed-by: NHannes Reinecke <hare@suse.com>
Tested-by: NOndrej Zary <linux@rainbow-software.org>
Tested-by: NMichael Schmitz <schmitzmic@gmail.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

When polling a device register under irq lock the polling loop terminates
after a given number of jiffies. Make this timeout independent of the HZ
setting.

All 5380 drivers benefit from this patch, which optimizes the PIO fast
path, because they all use PIO transfers (for phases other than DATA IN
and DATA OUT). Some cards support only PIO transfers (even for DATA
phases). CPU cycles are scarce on some of these systems, so a small
improvement here makes a big difference.
Signed-off-by: NFinn Thain <fthain@telegraphics.com.au>
Reviewed-by: NHannes Reinecke <hare@suse.com>
Tested-by: NOndrej Zary <linux@rainbow-software.org>
Tested-by: NMichael Schmitz <schmitzmic@gmail.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

This patch fixes an old bug: accesses to device registers from the
interrupt handler (after reselection, DMA completion etc.) could mess
up a device register access elsewhere, if the latter takes place outside
of an irq lock (during selection etc.).
Signed-off-by: NFinn Thain <fthain@telegraphics.com.au>
Reviewed-by: NHannes Reinecke <hare@suse.com>
Tested-by: NMichael Schmitz <schmitzmic@gmail.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Signed-off-by: NFinn Thain <fthain@telegraphics.com.au>
Reviewed-by: NHannes Reinecke <hare@suse.com>
Acked-by: NRussell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Merge the port-mapped IO and memory-mapped IO support (with the help of
ioport_map) into the g_NCR5380 module and delete g_NCR5380_mmio.
Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Signed-off-by: NFinn Thain <fthain@telegraphics.com.au>
Reviewed-by: NHannes Reinecke <hare@suse.com>
Tested-by: NOndrej Zary <linux@rainbow-software.org>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Vendor specific setup_clocks callback may require the clocks managed by
ufshcd driver to be ON. So if the vendor specific setup_clocks callback
is called while the required clocks are turned off, it could result into
unclocked register access.

To prevent possible unclock register access, this change adds one more
argument to setup_clocks callback to let it know whether it is called
pre/post the clock changes by core driver.
Signed-off-by: NSubhash Jadavani <subhashj@codeaurora.org>
Reviewed-by: NKiwoong Kim <kwmad.kim@samsung.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

If we haven't logged into the fabric yet we want to be a little more nuanced
with our CVL handling than what we've been:

- If the FCF has been selected, check the source MAC to make sure the frame is
from the FCF we've selected.
- If a FCF is selected and the CVL is from the FCF but we have not logged in
yet, then reset everything and go back to solicitation.
Signed-off-by: NChad Dupuis <chad.dupuis@cavium.com>
Reviewed-by: NHannes Reinecke <hare@suse.com>
Acked-by: NJohannes Thumshirn <jth@kernel.org>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

When we receive an FLOGI but have already sent our own we should
not advance the state machine but rather wait for our FLOGI to
return before continuing with PLOGI.
Signed-off-by: NHannes Reinecke <hare@suse.com>
Acked-by: NJohannes Thumshirn <jth@kernel.org>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

When the port is already started we don't need to login; that
will only confuse the state machine.
Signed-off-by: NHannes Reinecke <hare@suse.com>
Acked-by: NJohannes Thumshirn <jth@kernel.org>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

When fc_rport_login() is called while the rport is not
in RPORT_ST_INIT, RPORT_ST_READY, or RPORT_ST_DELETE
login is already in progress and there's no need to
drop down to FLOGI; doing so will only confuse the
other side.
Signed-off-by: NHannes Reinecke <hare@suse.com>
Acked-by: NJohannes Thumshirn <jth@kernel.org>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

When an ELS response handler receives a -FC_EX_CLOSED, the rdata->rp_mutex is
already held which can lead to a deadlock condition like the following stack trace:

[<ffffffffa04d8f18>] fc_rport_plogi_resp+0x28/0x200 [libfc]
[<ffffffffa04cfa1a>] fc_invoke_resp+0x6a/0xe0 [libfc]
[<ffffffffa04d0c08>] fc_exch_mgr_reset+0x1b8/0x280 [libfc]
[<ffffffffa04d87b3>] fc_rport_logoff+0x43/0xd0 [libfc]
[<ffffffffa04ce73d>] fc_disc_stop+0x6d/0xf0 [libfc]
[<ffffffffa04ce7ce>] fc_disc_stop_final+0xe/0x20 [libfc]
[<ffffffffa04d55f7>] fc_fabric_logoff+0x17/0x70 [libfc]

The other ELS handlers need to follow the FLOGI response handler and simply do
a kref_put against the fc_rport_priv struct and exit when receving a
-FC_EX_CLOSED response.
Signed-off-by: NChad Dupuis <chad.dupuis@cavium.com>
Reviewed-by: NHannes Reinecke <hare@suse.com>
Acked-by: NJohannes Thumshirn <jth@kernel.org>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

The list of attached 'rdata' remote port structures is RCU
protected, so there is no need to take the 'disc_mutex' when
traversing it.
Rather we should be using rcu_read_lock() and kref_get_unless_zero()
to validate the entries.
We need, however, take the disc_mutex when deleting an entry;
otherwise we risk clashes with list_add.
Signed-off-by: NHannes Reinecke <hare@suse.com>
Acked-by: NJohannes Thumshirn <jth@kernel.org>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

The kref handling in fc_rport is a mess. This patch updates
the kref handling according to the following rules:

- Take a reference whenever scheduling a workqueue
- Take a reference whenever an ELS command is send
- Drop the reference at the end of the workqueue function
- Drop the reference at the end of handling ELS replies
- Take a reference when allocating an rport
- Drop the reference when removing an rport
Signed-off-by: NHannes Reinecke <hare@suse.com>
Acked-by: NJohannes Thumshirn <jth@kernel.org>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

The hip06 D03 and hip07 D05 boards have different reference clock
frequencies for the SAS controller.

Register PHY_CTRL needs to be programmed differently according to this
frequency, so add support for this.

The default register setting in PHY_CTRL is for 50MHz, so only update
this register when the refclk frequency is 66MHz.

For ACPI we expect the _RST handler to set the correct value for
PHY_CTRL (we're forced to take different approach for DT and ACPI as
ACPI does not support fixed-clock device).
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NXiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Chipset hip07 incorporates v2 hw.
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NXiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Add support for hip07 chipset to hisi_sas controller.

Chipset hip07 has v2 hw.
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NXiang Chen <chenxiang66@hisilicon.com>
Acked-by: NRob Herring <robh@kernel.org>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Trace timestamps use struct timespec and CURRENT_TIME which are not
y2038 safe.  These timestamps are only part of the trace log on the
machine and are not shared with the fnic.  Replace then with y2038 safe
struct timespec64 and ktime_get_real_ts64(), respectively.
Signed-off-by: NDeepa Dinamani <deepa.kernel@gmail.com>
Reviewed-by: NArnd Bergmann <arnd@arndb.de>
Cc: Hiral Patel <hiralpat@cisco.com>
Cc: Suma Ramars <sramars@cisco.com>
Cc: Brian Uchino <buchino@cisco.com>
Cc: "James E.J. Bottomley" <jejb@linux.vnet.ibm.com>
Cc: "Martin K. Petersen" <martin.petersen@oracle.com>
Cc: linux-scsi@vger.kernel.org
Reviewed-by: NJohannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Switch the ipr driver to use pci_alloc_irq_vectors.  We need to two
calls to pci_alloc_irq_vectors as ipr only supports multiple MSI-X
vectors, but not multiple MSI vectors.

Otherwise this cleans up a lot of cruft and allows to use a common
request_irq loop for irq types, which happens to only iterate over a
single line in the non MSI-X case.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Acked-by: NBrian King <brking@linux.vnet.ibm.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Switch the arcmsr driver to use pci_alloc_irq_vectors.  We need to two
calls to pci_alloc_irq_vectors as arcmsr only supports multiple MSI-X
vectors, but not multiple MSI vectors.

Otherwise this cleans up a lot of cruft and allows to use a common
request_irq loop for irq types, which happens to only iterate over a
single line in the non MSI-X case.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Acked-by: NChing Huang <ching2048@areca.com.tw>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Pull IOMMU fixes from Joerg Roedel:

 - Four patches from Robin Murphy fix several issues with the recently
   merged generic DT-bindings support for arm-smmu drivers

 - A fix for a dead-lock issue in the VT-d driver, which shows up on
   iommu hotplug

* tag 'iommu-fixes-v4.9-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/joro/iommu:
  iommu/vt-d: Fix dead-locks in disable_dmar_iommu() path
  iommu/arm-smmu: Fix out-of-bounds dereference
  iommu/arm-smmu: Check that iommu_fwspecs are ours
  iommu/arm-smmu: Don't inadvertently reject multiple SMMUv3s
  iommu/arm-smmu: Work around ARM DMA configuration

It turns out that the disable_dmar_iommu() code-path tried
to get the device_domain_lock recursivly, which will
dead-lock when this code runs on dmar removal. Fix both
code-paths that could lead to the dead-lock.

Fixes: 55d94043 ('iommu/vt-d: Get rid of domain->iommu_lock')
Signed-off-by: NJoerg Roedel <jroedel@suse.de>

When we iterate a master's config entries, what we generally care
about is the entry's stream map index, rather than the entry index
itself, so it's nice to have the iterator automatically assign the
former from the latter. Unfortunately, booting with KASAN reveals
the oversight that using a simple comma operator results in the
entry index being dereferenced before being checked for validity,
so we always access one element past the end of the fwspec array.

Flip things around so that the check always happens before the index
may be dereferenced.

Fixes: adfec2e7 ("iommu/arm-smmu: Convert to iommu_fwspec")
Reported-by: NMark Rutland <mark.rutland@arm.com>
Signed-off-by: NRobin Murphy <robin.murphy@arm.com>
Acked-by: NWill Deacon <will.deacon@arm.com>
Signed-off-by: NJoerg Roedel <jroedel@suse.de>

We seem to have forgotten to check that iommu_fwspecs actually belong to
us before we go ahead and dereference their private data. Oops.

Fixes: 021bb842 ("iommu/arm-smmu: Wire up generic configuration support")
Signed-off-by: NRobin Murphy <robin.murphy@arm.com>
Signed-off-by: NJoerg Roedel <jroedel@suse.de>

We now delay installing our per-bus iommu_ops until we know an SMMU has
successfully probed, as they don't serve much purpose beforehand, and
doing so also avoids fights between multiple IOMMU drivers in a single
kernel. However, the upshot of passing the return value of bus_set_iommu()
back from our probe function is that if there happens to be more than
one SMMUv3 device in a system, the second and subsequent probes will
wind up returning -EBUSY to the driver core and getting torn down again.

Avoid re-setting ops if ours are already installed, so that any genuine
failures stand out.

Fixes: 08d4ca2a ("iommu/arm-smmu: Support non-PCI devices with SMMUv3")
CC: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
CC: Hanjun Guo <hanjun.guo@linaro.org>
Signed-off-by: NRobin Murphy <robin.murphy@arm.com>
Signed-off-by: NJoerg Roedel <jroedel@suse.de>

The 32-bit ARM DMA configuration code predates the IOMMU core's default
domain functionality, and instead relies on allocating its own domains
and attaching any devices using the generic IOMMU binding to them.
Unfortunately, it does this relatively early on in the creation of the
device, before we've seen our add_device callback, which leads us to
attempt to operate on a half-configured master.

To avoid a crash, check for this situation on attach, but refuse to
play, as there's nothing we can do. This at least allows VFIO to keep
working for people who update their 32-bit DTs to the generic binding,
albeit with a few (innocuous) warnings from the DMA layer on boot.
Signed-off-by: NRobin Murphy <robin.murphy@arm.com>
Signed-off-by: NWill Deacon <will.deacon@arm.com>
Signed-off-by: NJoerg Roedel <jroedel@suse.de>

Pull arm64 fix from Will Deacon:
 "It's been pretty quiet on the fixes side of things for us, but Artem
  reported a build failure introduced during the merge window that
  appears with older GCCs that do not support asm goto. The fix is
  bigger than I'd like, but it's a mechnical move of some constants to
  break an include dependency between atomic.h and jump_label.h when
  !HAVE_JUMP_LABEL.

  Summary:

   - Fix build failure on compilers without asm goto"

* tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
  arm64: Fix circular include of asm/lse.h through linux/jump_label.h

Merge tag 'openrisc-for-linus-v4.9-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/groeck/linux-staging

Pull openrisc fix from Guenter Roeck:
 "Fix openrisc crash caused by ro_init changes"

* tag 'openrisc-for-linus-v4.9-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/groeck/linux-staging:
  openrisc: Define __ro_after_init to avoid crash

Merge tag 'hwmon-for-linus-v4.9-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/groeck/linux-staging

Pull hwmon fix from Guenter Roeck:
 "Fix resource leak on devm_kcalloc failure"

* tag 'hwmon-for-linus-v4.9-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/groeck/linux-staging:
  hwmon: (core) fix resource leak on devm_kcalloc failure

Pull HID fixes from Jiri Kosina:

 - modprobe-after-rmmod load failure bugfix for intel-ish, from Even Xu

 - IRQ probing bugfix for intel-ish, from Srinivas Pandruvada

 - attribute parsing fix in hid-sensor, from Ooi, Joyce

 - other small misc fixes / quirky device additions

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/hid:
  HID: sensor: fix attributes in HID sensor interface
  HID: intel-ish-hid: request_irq failure
  HID: intel-ish-hid: Fix driver reinit failure
  HID: intel-ish-hid: Move DMA disable code to new function
  HID: intel-ish-hid: consolidate ish wake up operation
  HID: usbhid: add ATEN CS962 to list of quirky devices
  HID: intel-ish-hid: Fix !CONFIG_PM build warning
  HID: sensor-hub: Fix packing of result buffer for feature report

openrisc qemu tests fail with the following crash.

Unable to handle kernel access at virtual address 0xc0300c34

Oops#: 0001
CPU #: 0
   PC: c016c710    SR: 0000ae67    SP: c1017e04
   GPR00: 00000000 GPR01: c1017e04 GPR02: c0300c34 GPR03: c0300c34
   GPR04: 00000000 GPR05: c0300cb0 GPR06: c0300c34 GPR07: 000000ff
   GPR08: c107f074 GPR09: c0199ef4 GPR10: c1016000 GPR11: 00000000
   GPR12: 00000000 GPR13: c107f044 GPR14: c0473774 GPR15: 07ce0000
   GPR16: 00000000 GPR17: c107ed8a GPR18: 00009600 GPR19: c107f044
   GPR20: c107ee74 GPR21: 00000003 GPR22: c0473770 GPR23: 00000033
   GPR24: 000000bf GPR25: 00000019 GPR26: c046400c GPR27: 00000001
   GPR28: c0464028 GPR29: c1018000 GPR30: 00000006 GPR31: ccf37483
     RES: 00000000 oGPR11: ffffffff
     Process swapper (pid: 1, stackpage=c1001960)

     Stack: Stack dump [0xc1017cf8]:
     sp + 00: 0xc1017e04
     sp + 04: 0xc0300c34
     sp + 08: 0xc0300c34
     sp + 12: 0x00000000
...

Bisect points to commit d2ec3f77 ("pty: make ptmx file ops read-only
after init"). Fix by defining __ro_after_init for the openrisc
architecture, similar to parisc.

Fixes: d2ec3f77 ("pty: make ptmx file ops read-only after init")
Cc: Kees Cook <keescook@chromium.org>
Signed-off-by: NGuenter Roeck <linux@roeck-us.net>
Acked-by: NStafford Horne <shorne@gmail.com>

Pull i2c fix from Wolfram Sang:
 "A bugfix for the I2C core fixing a (rare) race condition"

* 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux:
  i2c: core: fix NULL pointer dereference under race condition

Commit efd9e03f ("arm64: Use static keys for CPU features")
introduced support for static keys in asm/cpufeature.h, including
linux/jump_label.h. When CC_HAVE_ASM_GOTO is not defined, this causes a
circular dependency via linux/atomic.h, asm/lse.h and asm/cpufeature.h.

This patch moves the capability macros out out of asm/cpufeature.h into
a separate asm/cpucaps.h and modifies some of the #includes accordingly.

Fixes: efd9e03f ("arm64: Use static keys for CPU features")
Reported-by: NArtem Savkov <asavkov@redhat.com>
Tested-by: NArtem Savkov <asavkov@redhat.com>
Signed-off-by: NCatalin Marinas <catalin.marinas@arm.com>
Signed-off-by: NWill Deacon <will.deacon@arm.com>

Merge branches 'sched-urgent-for-linus' and 'core-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull stack vmap fixups from Thomas Gleixner:
 "Two small patches related to sched_show_task():

   - make sure to hold a reference on the task stack while accessing it

   - remove the thread_saved_pc printout

  .. and add a sanity check into release_task_stack() to catch problems
  with task stack references"

* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  sched/core: Remove pointless printout in sched_show_task()
  sched/core: Fix oops in sched_show_task()

* 'core-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  fork: Add task stack refcounting sanity check and prevent premature task stack freeing

Pull MD fixes from Shaohua Li:
 "There are several bug fixes queued:

   - fix raid5-cache recovery bugs

   - fix discard IO error handling for raid1/10

   - fix array sync writes bogus position to superblock

   - fix IO error handling for raid array with external metadata"

* tag 'md/4.9-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/shli/md:
  md: be careful not lot leak internal curr_resync value into metadata. -- (all)
  raid1: handle read error also in readonly mode
  raid5-cache: correct condition for empty metadata write
  md: report 'write_pending' state when array in sync
  md/raid5: write an empty meta-block when creating log super-block
  md/raid5: initialize next_checkpoint field before use
  RAID10: ignore discard error
  RAID1: ignore discard error

Pull SCSI fixes from James Bottomley:
 "Two more important data integrity fixes related to RAID device drivers
  which wrongly throw away the SYNCHRONIZE CACHE command in the non-RAID
  path and a memory leak in the scsi_debug driver"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  scsi: arcmsr: Send SYNCHRONIZE_CACHE command to firmware
  scsi: scsi_debug: Fix memory leak if LBP enabled and module is unloaded
  scsi: megaraid_sas: Fix data integrity failure for JBOD (passthrough) devices

Pull input subsystem updates from Dmitry Torokhov.

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
  Input: psmouse - cleanup Focaltech code
  Input: i8042 - add XMG C504 to keyboard reset table

Pull FireWire (IEEE 1394) fixes from Stefan Richter:

 - add missing input validation to the firewire-net driver. Invalid
   IP-over-1394 encapsulation headers could trigger buffer overflows
   (CVE 2016-8633).

 - IP-over-1394 link fragmentation headers were read and written
   incorrectly, breaking fragmented RX/TX with other OS's stacks.

* tag 'firewire-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394/linux1394:
  firewire: net: fix fragmented datagram_size off-by-one
  firewire: net: guard against rx buffer overflows

Pull media fixes from Mauro Carvalho Chehab:
 "A series of fixup patches meant to fix the usage of DMA on stack, plus
  one warning fixup"

* tag 'media/v4.9-3' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media: (32 commits)
  [media] radio-bcm2048: don't ignore errors
  [media] pctv452e: fix semicolon.cocci warnings
  [media] flexcop-usb: don't use stack for DMA
  [media] stk-webcam: don't use stack for DMA
  [media] s2255drv: don't use stack for DMA
  [media] cpia2_usb: don't use stack for DMA
  [media] digitv: handle error code on RC query
  [media] dw2102: return error if su3000_power_ctrl() fails
  [media] nova-t-usb2: handle error code on RC query
  [media] technisat-usb2: use DMA buffers for I2C transfers
  [media] pctv452e: don't call BUG_ON() on non-fatal error
  [media] pctv452e: don't do DMA on stack
  [media] nova-t-usb2: don't do DMA on stack
  [media] gp8psk: don't go past the buffer size
  [media] gp8psk: don't do DMA on stack
  [media] dtv5100: don't do DMA on stack
  [media] dtt200u: handle USB control message errors
  [media] dtt200u: don't do DMA on stack
  [media] dtt200u-fe: handle errors on USB control messages
  [media] dtt200u-fe: don't do DMA on stack
  ...