Linus Torvalds [Thu, 22 Nov 2018 16:31:46 +0000 (08:31 -0800)]
Merge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Pull SCSI fixes from James Bottomley:
"Two small fixes.
The qla2xxx is a regression from 4.18 and the ufs one is a device
enablement fix"
* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
scsi: ufs: Fix hynix ufs bug with quirk on hi36xx SoC
scsi: qla2xxx: Timeouts occur on surprise removal of QLogic adapter
There is a HID feature report called "Resolution Multiplier"
Described in the "Enhanced Wheel Support in Windows" doc and
the "USB HID Usage Tables" page 30.
This was new for Windows Vista, so we're only a decade behind here. I only
accidentally found this a few days ago while debugging a stuck button on a
Microsoft mouse.
The docs above describe it like this: a wheel control by default sends
value 1 per notch. If the resolution multiplier is active, the wheel is
expected to send a value of $multiplier per notch (e.g. MS Sculpt mouse) or
just send events more often, i.e. for less physical motion (e.g. MS Comfort
mouse).
For the latter, you need the right HW of course. The Sculpt mouse has
tactile wheel clicks, so nothing really changes. The Comfort mouse has
continuous motion with no tactile clicks. Similar to the free-wheeling
Logitech mice but without any inertia.
Note that the doc also says that Vista and onwards *always* enable this
feature where available.
An example HID definition looks like this:
Usage Page Generic Desktop (0x01)
Usage Resolution Multiplier (0x48)
Logical Minimum 0
Logical Maximum 1
Physical Minimum 1
Physical Maximum 16
Report Size 2 # in bits
Report Count 1
Feature (Data, Var, Abs)
So the actual bits have values 0 or 1 and that reflects real values 1 or 16.
We've only seen single-bits so far, so there's low-res and hi-res, but
nothing in between.
The multiplier is available for HID usages "Wheel" and "AC Pan" (horiz wheel).
Microsoft suggests that
> Vendors should ship their devices with smooth scrolling disabled and allow
> Windows to enable it. This ensures that the device works like a regular HID
> device on legacy operating systems that do not support smooth scrolling.
(see the wheel doc linked above)
The mice that we tested so far do reset on unplug.
Device Support looks to be all (?) Microsoft mice but nothing else
Not supported:
- Logitech G500s, G303
- Roccat Kone XTD
- all the cheap Lenovo, HP, Dell, Logitech USB mice that come with a
workstation that I could find don't have it.
- Etekcity something something
- Razer Imperator
Supported:
- Microsoft Comfort Optical Mouse 3000 - yes, physical: 1:4
- Microsoft Sculpt Ergonomic Mouse - yes, physical: 1:12
- Microsoft Surface mouse - yes, physical: 1:4
So again, I think this is really just available on Microsoft mice, but
probably all decent MS mice released over the last decade.
Looking at the hardware itself:
- no noticeable notches in the weel
- low-res: 18 events per 360deg rotation (click angle 20 deg)
- high-res: 72 events per 360deg → matches multiplier of 4
- I can feel the notches during wheel turns
- low-res: 24 events per 360 deg rotation (click angle 15 deg)
- horiz wheel is tilt-based, continuous output value 1
- high-res: 24 events per 360deg with value 12 → matches multiplier of 12
- horiz wheel output rate doubles/triples?, values is 3
- It's a touch strip, not a wheel so no notches
- high-res: events have value 4 instead of 1
a bit strange given that it doesn't actually have notches.
Ok, why is this an issue for the current API? First, because the logitech
multiplier used in Harry's patches looks suspiciously like the Resolution
Multiplier so I think we should assume it's the same thing. Nestor, can you
shed some light on that?
- `REL_WHEEL` is defined as the number of notches, emulated where needed.
- `REL_WHEEL_HI_RES` is the movement of the user's finger in microns.
- `WM_MOUSEWHEEL` (Windows) is is a multiple of 120, defined as "the threshold
for action to be taken and one such action"
https://docs.microsoft.com/en-us/windows/desktop/inputdev/wm-mousewheel
If the multiplier is set to M, this means we need an accumulated value of M
until we can claim there was a wheel click. So after enabling the multiplier
and setting it to the maximum (like Windows):
- M units are 15deg rotation → 1 unit is 2620/M micron (see below). This is
the `REL_WHEEL_HI_RES` value.
- wheel diameter 20mm: 15 deg rotation is 2.62mm, 2620 micron (pi * 20mm /
(360deg/15deg))
- For every M units accumulated, send one `REL_WHEEL` event
The problem here is that we've now hardcoded 20mm/15 deg into the kernel and
we have no way of getting the size of the wheel or the click angle into the
kernel.
In userspace we now have to undo the kernel's calculation. If our click angle
is e.g. 20 degree we have to undo the (lossy) calculation from the kernel and
calculate the correct angle instead. This also means the 15 is a hardcoded
option forever and cannot be changed.
In hid-logitech-hidpp.c, the microns per unit is hardcoded per device.
Harry, did you measure those by hand? We'd need to update the kernel for
every device and there are 10 years worth of devices from MS alone.
The multiplier default is 8 which is in the right ballpark, so I'm pretty
sure this is the same as the Resolution Multiplier, just in HID++ lingo. And
given that the 120 magic factor is what Windows uses in the end, I can't
imagine Logitech rolling their own thing here. Nestor?
And we're already fairly inaccurate with the microns anyway. The MX Anywhere
2S has a click angle of 20 degrees (18 stops) and a 17mm wheel, so a wheel
notch is approximately 2.67mm, one event at multiplier 8 (1/8 of a notch)
would be 334 micron. That's only 80% of the fallback value of 406 in the
kernel. Multiplier 6 gives us 445micron (10% off). I'm assuming multiplier 7
doesn't exist because it's not a factor of 120.
Summary:
Best option may be to simply do what Windows is doing, all the HW manufacturers
have to use that approach after all. Switch `REL_WHEEL_HI_RES` to report in
fractions of 120, with 120 being one notch and divide that by the multiplier
for the actual events. So e.g. the Logitech multiplier 8 would send value 15
for each event in hi-res mode. This can be converted in userspace to
whatever userspace needs (combined with a hwdb there that tells you wheel
size/click angle/...).
Conflicts:
include/uapi/linux/input-event-codes.h -> I kept the new
reserved event in the code, so I had to adapt the revert
slightly
Dave Airlie [Thu, 22 Nov 2018 01:19:20 +0000 (11:19 +1000)]
Merge branch 'drm-fixes-4.20' of git://people.freedesktop.org/~agd5f/linux into drm-fixes
- OD fixes for powerplay
- Vega20 fixes
- KFD fix for Kaveri
- add missing firmware declaration for hainan (SI chip)
- Fix DC user experience regressions related to panels that support >8 bpc
Vincent Chen [Wed, 21 Nov 2018 01:38:11 +0000 (09:38 +0800)]
net: faraday: ftmac100: remove netif_running(netdev) check before disabling interrupts
In the original ftmac100_interrupt(), the interrupts are only disabled when
the condition "netif_running(netdev)" is true. However, this condition
causes kerenl hang in the following case. When the user requests to
disable the network device, kernel will clear the bit __LINK_STATE_START
from the dev->state and then call the driver's ndo_stop function. Network
device interrupts are not blocked during this process. If an interrupt
occurs between clearing __LINK_STATE_START and stopping network device,
kernel cannot disable the interrupts due to the condition
"netif_running(netdev)" in the ISR. Hence, kernel will hang due to the
continuous interruption of the network device.
In order to solve the above problem, the interrupts of the network device
should always be disabled in the ISR without being restricted by the
condition "netif_running(netdev)".
Y.C. Chen [Wed, 3 Oct 2018 06:57:47 +0000 (14:57 +0800)]
drm/ast: change resolution may cause screen blurred
The value of pitches is not correct while calling mode_set.
The issue we found so far on following system:
- Debian8 with XFCE Desktop
- Ubuntu with KDE Desktop
- SUSE15 with KDE Desktop
David S. Miller [Thu, 22 Nov 2018 00:14:56 +0000 (16:14 -0800)]
Merge branch 'smc-fixes'
Ursula Braun says:
====================
net/smc: fixes 2018-11-12
here is V4 of some net/smc fixes in different areas for the net tree.
v1->v2:
do not define 8-byte alignment for union smcd_cdc_cursor in
patch 4/5 "net/smc: atomic SMCD cursor handling"
v2->v3:
stay with 8-byte alignment for union smcd_cdc_cursor in
patch 4/5 "net/smc: atomic SMCD cursor handling", but get rid of
__packed for struct smcd_cdc_msg
v3->v4:
get rid of another __packed for struct smc_cdc_msg in
patch 4/5 "net/smc: atomic SMCD cursor handling"
====================
Ursula Braun [Tue, 20 Nov 2018 15:46:43 +0000 (16:46 +0100)]
net/smc: use after free fix in smc_wr_tx_put_slot()
In smc_wr_tx_put_slot() field pend->idx is used after being
cleared. That means always idx 0 is cleared in the wr_tx_mask.
This results in a broken administration of available WR send
payload buffers.
Hans Wippel [Tue, 20 Nov 2018 15:46:41 +0000 (16:46 +0100)]
net/smc: add SMC-D shutdown signal
When a SMC-D link group is freed, a shutdown signal should be sent to
the peer to indicate that the link group is invalid. This patch adds the
shutdown signal to the SMC code.
Karsten Graul [Tue, 20 Nov 2018 15:46:40 +0000 (16:46 +0100)]
net/smc: use queue pair number when matching link group
When searching for an existing link group the queue pair number is also
to be taken into consideration. When the SMC server sends a new number
in a CLC packet (keeping all other values equal) then a new link group
is to be created on the SMC client side.
Hans Wippel [Tue, 20 Nov 2018 15:46:39 +0000 (16:46 +0100)]
net/smc: abort CLC connection in smc_release
In case of a non-blocking SMC socket, the initial CLC handshake is
performed over a blocking TCP connection in a worker. If the SMC socket
is released, smc_release has to wait for the blocking CLC socket
operations (e.g., kernel_connect) inside the worker.
This patch aborts a CLC connection when the respective non-blocking SMC
socket is released to avoid waiting on socket operations or timeouts.
Eric Dumazet [Tue, 20 Nov 2018 13:53:59 +0000 (05:53 -0800)]
tcp: defer SACK compression after DupThresh
Jean-Louis reported a TCP regression and bisected to recent SACK
compression.
After a loss episode (receiver not able to keep up and dropping
packets because its backlog is full), linux TCP stack is sending
a single SACK (DUPACK).
Sender waits a full RTO timer before recovering losses.
While RFC 6675 says in section 5, "Algorithm Details",
(2) If DupAcks < DupThresh but IsLost (HighACK + 1) returns true --
indicating at least three segments have arrived above the current
cumulative acknowledgment point, which is taken to indicate loss
-- go to step (4).
...
(4) Invoke fast retransmit and enter loss recovery as follows:
there are old TCP stacks not implementing this strategy, and
still counting the dupacks before starting fast retransmit.
While these stacks probably perform poorly when receivers implement
LRO/GRO, we should be a little more gentle to them.
This patch makes sure we do not enable SACK compression unless
3 dupacks have been sent since last rcv_nxt update.
Ideally we should even rearm the timer to send one or two
more DUPACK if no more packets are coming, but that will
be work aiming for linux-4.21.
Many thanks to Jean-Louis for bisecting the issue, providing
packet captures and testing this patch.
Petr Machata [Tue, 20 Nov 2018 11:39:56 +0000 (11:39 +0000)]
net: skb_scrub_packet(): Scrub offload_fwd_mark
When a packet is trapped and the corresponding SKB marked as
already-forwarded, it retains this marking even after it is forwarded
across veth links into another bridge. There, since it ingresses the
bridge over veth, which doesn't have offload_fwd_mark, it triggers a
warning in nbp_switchdev_frame_mark().
Then nbp_switchdev_allowed_egress() decides not to allow egress from
this bridge through another veth, because the SKB is already marked, and
the mark (of 0) of course matches. Thus the packet is incorrectly
blocked.
Solve by resetting offload_fwd_mark() in skb_scrub_packet(). That
function is called from tunnels and also from veth, and thus catches the
cases where traffic is forwarded between bridges and transformed in a
way that invalidates the marking.
Fixes: 6bc506b4fb06 ("bridge: switchdev: Add forward mark support for stacked devices") Fixes: abf4bb6b63d0 ("skbuff: Add the offload_mr_fwd_mark field") Signed-off-by: Petr Machata <[email protected]> Suggested-by: Ido Schimmel <[email protected]> Acked-by: Jiri Pirko <[email protected]> Signed-off-by: David S. Miller <[email protected]>
Linus Torvalds [Wed, 21 Nov 2018 19:28:20 +0000 (11:28 -0800)]
Merge tag 'riscv-for-linus-4.20-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/palmer/riscv-linux
Pull RISC-V fixes from Palmer Dabbelt:
"This week is a bit bigger than I expected. That's my fault, as I
missed a few patches while I was at Plumbers last week. We have:
- A fix to a quite embarassing issue where raw_copy_to_user() was
implemented with asm_copy_from_user() (and vice versa).
- Improvements to our makefile to allow flat binaries to be
generated.
- A build fix that predeclares "struct module" at the top of
<asm/module.h>, which triggers warnings later in that header.
- The addition of our own <uapi/asm/unistd> header, which is
necessary to align our stat ABI on 32-bit systems.
- A fix to avoid printing a warning when the S or U bits are set in
print_isa().
I already have one patch in the queue for next week"
* tag 'riscv-for-linus-4.20-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/palmer/riscv-linux:
RISC-V: recognize S/U mode bits in print_isa
riscv: add asm/unistd.h UAPI header
riscv: fix warning in arch/riscv/include/asm/module.h
RISC-V: Build flat and compressed kernel images
RISC-V: Fix raw_copy_{to,from}_user()
Dave Chinner [Wed, 21 Nov 2018 16:06:37 +0000 (08:06 -0800)]
iomap: readpages doesn't zero page tail beyond EOF
When we read the EOF page of the file via readpages, we need
to zero the region beyond EOF that we either do not read or
should not contain data so that mmap does not expose stale data to
user applications.
However, iomap_adjust_read_range() fails to detect EOF correctly,
and so fsx on 1k block size filesystems fails very quickly with
mapreads exposing data beyond EOF. There are two problems here.
Firstly, when calculating the end block of the EOF byte, we have
to round the size by one to avoid a block aligned EOF from reporting
a block too large. i.e. a size of 1024 bytes is 1 block, which in
index terms is block 0. Therefore we have to calculate the end block
from (isize - 1), not isize.
The second bug is determining if the current page spans EOF, and so
whether we need split it into two half, one for the IO, and the
other for zeroing. Unfortunately, the code that checks whether
we should split the block doesn't actually check if we span EOF, it
just checks if the read spans the /offset in the page/ that EOF
sits on. So it splits every read into two if EOF is not page
aligned, regardless of whether we are reading the EOF block or not.
Hence we need to restrict the "does the read span EOF" check to
just the page that spans EOF, not every page we read.
This patch results in correct EOF detection through readpages:
As you can see, it now does full page reads until the last one which
is split correctly at the block aligned EOF, reading 3072 bytes and
zeroing the last 1024 bytes. The original version of the patch got
this right, but it got another case wrong.
The EOF detection crossing really needs to the the original length
as plen, while it starts at the end of the block, will be shortened
as up-to-date blocks are found on the page. This means "orig_pos +
plen" no longer points to the end of the page, and so will not
correctly detect EOF crossing. Hence we have to use the length
passed in to detect this partial page case:
Heere we see a trace where the first block on the EOF page is up to
date, hence poff = 1024 bytes. The offset into the page of EOF is
3072, so the range we want to read is 1024 - 3071, and the range we
want to zero is 3072 - 4095. You can see this is split correctly
now.
This fixes the stale data beyond EOF problem that fsx quickly
uncovers on 1k block size filesystems.
It returns EINVAL when the operation is not supported by the
filesystem. Fix it to return EOPNOTSUPP to be consistent with
the man page and clone_file_range().
Clean up the inconsistent error return handling while I'm there.
(I know, lipstick on a pig, but every little bit helps...)
Dave Chinner [Mon, 19 Nov 2018 21:31:11 +0000 (13:31 -0800)]
iomap: dio data corruption and spurious errors when pipes fill
When doing direct IO to a pipe for do_splice_direct(), then pipe is
trivial to fill up and overflow as it can only hold 16 pages. At
this point bio_iov_iter_get_pages() then returns -EFAULT, and we
abort the IO submission process. Unfortunately, iomap_dio_rw()
propagates the error back up the stack.
The error is converted from the EFAULT to EAGAIN in
generic_file_splice_read() to tell the splice layers that the pipe
is full. do_splice_direct() completely fails to handle EAGAIN errors
(it aborts on error) and returns EAGAIN to the caller.
copy_file_write() then completely fails to handle EAGAIN as well,
and so returns EAGAIN to userspace, having failed to copy the data
it was asked to.
Avoid this whole steaming pile of fail by having iomap_dio_rw()
silently swallow EFAULT errors and so do short reads.
To make matters worse, iomap_dio_actor() has a stale data exposure
bug bio_iov_iter_get_pages() fails - it does not zero the tail block
that it may have been left uncovered by partial IO. Fix the error
handling case to drop to the sub-block zeroing rather than
immmediately returning the -EFAULT error.
Dave Chinner [Mon, 19 Nov 2018 21:31:10 +0000 (13:31 -0800)]
iomap: sub-block dio needs to zeroout beyond EOF
If we are doing sub-block dio that extends EOF, we need to zero
the unused tail of the block to initialise the data in it it. If we
do not zero the tail of the block, then an immediate mmap read of
the EOF block will expose stale data beyond EOF to userspace. Found
with fsx running sub-block DIO sizes vs MAPREAD/MAPWRITE operations.
Fix this by detecting if the end of the DIO write is beyond EOF
and zeroing the tail if necessary.
Dave Chinner [Mon, 19 Nov 2018 21:31:10 +0000 (13:31 -0800)]
iomap: FUA is wrong for DIO O_DSYNC writes into unwritten extents
When we write into an unwritten extent via direct IO, we dirty
metadata on IO completion to convert the unwritten extent to
written. However, when we do the FUA optimisation checks, the inode
may be clean and so we issue a FUA write into the unwritten extent.
This means we then bypass the generic_write_sync() call after
unwritten extent conversion has ben done and we don't force the
modified metadata to stable storage.
This violates O_DSYNC semantics. The window of exposure is a single
IO, as the next DIO write will see the inode has dirty metadata and
hence will not use the FUA optimisation. Calling
generic_write_sync() after completion of the second IO will also
sync the first write and it's metadata.
Fix this by avoiding the FUA optimisation when writing to unwritten
extents.
Dave Chinner [Tue, 20 Nov 2018 06:50:08 +0000 (22:50 -0800)]
xfs: delalloc -> unwritten COW fork allocation can go wrong
Long saga. There have been days spent following this through dead end
after dead end in multi-GB event traces. This morning, after writing
a trace-cmd wrapper that enabled me to be more selective about XFS
trace points, I discovered that I could get just enough essential
tracepoints enabled that there was a 50:50 chance the fsx config
would fail at ~115k ops. If it didn't fail at op 115547, I stopped
fsx at op 115548 anyway.
That gave me two traces - one where the problem manifested, and one
where it didn't. After refining the traces to have the necessary
information, I found that in the failing case there was a real
extent in the COW fork compared to an unwritten extent in the
working case.
Walking back through the two traces to the point where the CWO fork
extents actually diverged, I found that the bad case had an extra
unwritten extent in it. This is likely because the bug it led me to
had triggered multiple times in those 115k ops, leaving stray
COW extents around. What I saw was a COW delalloc conversion to an
unwritten extent (as they should always be through
xfs_iomap_write_allocate()) resulted in a /written extent/:
xfs_writepage: dev 259:0 ino 0x83 pgoff 0x17000 size 0x79a00 offset 0 length 0
xfs_iext_remove: dev 259:0 ino 0x83 state RC|LF|RF|COW cur 0xffff888247b899c0/2 offset 32 block 152 count 20 flag 1 caller xfs_bmap_add_extent_delay_real
xfs_bmap_pre_update: dev 259:0 ino 0x83 state RC|LF|RF|COW cur 0xffff888247b899c0/1 offset 1 block 4503599627239429 count 31 flag 0 caller xfs_bmap_add_extent_delay_real
xfs_bmap_post_update: dev 259:0 ino 0x83 state RC|LF|RF|COW cur 0xffff888247b899c0/1 offset 1 block 121 count 51 flag 0 caller xfs_bmap_add_ex
And the result according to the xfs_bmap_post_update trace was:
0 1 32 52
+H+wwwwwwwwwwwwwwwwwwwwwwww+
PREV
Which is clearly wrong - it should be a merged unwritten extent,
not an unwritten extent.
That lead me to look at the LEFT_FILLING|RIGHT_FILLING|RIGHT_CONTIG
case in xfs_bmap_add_extent_delay_real(), and sure enough, there's
the bug.
It takes the old delalloc extent (PREV) and adds the length of the
RIGHT extent to it, takes the start block from NEW, removes the
RIGHT extent and then updates PREV with the new extent.
What it fails to do is update PREV.br_state. For delalloc, this is
always XFS_EXT_NORM, while in this case we are converting the
delayed allocation to unwritten, so it needs to be updated to
XFS_EXT_UNWRITTEN. This LF|RF|RC case does not do this, and so
the resultant extent is always written.
And that's the bug I've been chasing for a week - a bmap btree bug,
not a reflink/dedupe/copy_file_range bug, but a BMBT bug introduced
with the recent in core extent tree scalability enhancements.
Dave Chinner [Mon, 19 Nov 2018 21:31:10 +0000 (13:31 -0800)]
xfs: flush removing page cache in xfs_reflink_remap_prep
On a sub-page block size filesystem, fsx is failing with a data
corruption after a series of operations involving copying a file
with the destination offset beyond EOF of the destination of the file:
8093(157 mod 256): TRUNCATE DOWN from 0x7a120 to 0x50000 ******WWWW
8094(158 mod 256): INSERT 0x25000 thru 0x25fff (0x1000 bytes)
8095(159 mod 256): COPY 0x18000 thru 0x1afff (0x3000 bytes) to 0x2f400
8096(160 mod 256): WRITE 0x5da00 thru 0x651ff (0x7800 bytes) HOLE
8097(161 mod 256): COPY 0x2000 thru 0x5fff (0x4000 bytes) to 0x6fc00
The second copy here is beyond EOF, and it is to sub-page (4k) but
block aligned (1k) offset. The clone runs the EOF zeroing, landing
in a pre-existing post-eof delalloc extent. This zeroes the post-eof
extents in the page cache just fine, dirtying the pages correctly.
The problem is that xfs_reflink_remap_prep() now truncates the page
cache over the range that it is copying it to, and rounds that down
to cover the entire start page. This removes the dirty page over the
delalloc extent from the page cache without having written it back.
Hence later, when the page cache is flushed, the page at offset
0x6f000 has not been written back and hence exposes stale data,
which fsx trips over less than 10 operations later.
Fix this by changing xfs_reflink_remap_prep() to use
xfs_flush_unmap_range().
Robin Murphy [Wed, 21 Nov 2018 16:00:51 +0000 (16:00 +0000)]
swiotlb: Skip cache maintenance on map error
If swiotlb_bounce_page() failed, calling arch_sync_dma_for_device() may
lead to such delights as performing cache maintenance on whatever
address phys_to_virt(SWIOTLB_MAP_ERROR) looks like, which is typically
outside the kernel memory map and goes about as well as expected.
Don't do that.
Fixes: a4a4330db46a ("swiotlb: add support for non-coherent DMA") Tested-by: John Stultz <[email protected]> Signed-off-by: Robin Murphy <[email protected]> Signed-off-by: Christoph Hellwig <[email protected]>
Robin Murphy [Wed, 21 Nov 2018 16:00:50 +0000 (16:00 +0000)]
dma-direct: Make DIRECT_MAPPING_ERROR viable for SWIOTLB
With the overflow buffer removed, we no longer have a unique address
which is guaranteed not to be a valid DMA target to use as an error
token. The DIRECT_MAPPING_ERROR value of 0 tries to at least represent
an unlikely DMA target, but unfortunately there are already SWIOTLB
users with DMA-able memory at physical address 0 which now gets falsely
treated as a mapping failure and leads to all manner of misbehaviour.
The best we can do to mitigate that is flip DIRECT_MAPPING_ERROR to the
other commonly-used error value of all-bits-set, since the last single
byte of memory is by far the least-likely-valid DMA target.
Robbie Ko [Wed, 14 Nov 2018 18:32:37 +0000 (18:32 +0000)]
Btrfs: send, fix infinite loop due to directory rename dependencies
When doing an incremental send, due to the need of delaying directory move
(rename) operations we can end up in infinite loop at
apply_children_dir_moves().
An example scenario that triggers this problem is described below, where
directory names correspond to the numbers of their respective inodes.
When processing inode 257 we delay its move (rename) operation because its
new parent in the send snapshot, inode 272, was not yet processed. Then
when processing inode 272, we delay the move operation for that inode
because inode 274 is its ancestor in the send snapshot. Finally we delay
the move operation for inode 274 when processing it because inode 275 is
its new parent in the send snapshot and was not yet moved.
When finishing processing inode 275, we start to do the move operations
that were previously delayed (at apply_children_dir_moves()), resulting in
the following iterations:
1) We issue the move operation for inode 274;
2) Because inode 262 depended on the move operation of inode 274 (it was
delayed because 274 is its ancestor in the send snapshot), we issue the
move operation for inode 262;
3) We issue the move operation for inode 272, because it was delayed by
inode 274 too (ancestor of 272 in the send snapshot);
4) We issue the move operation for inode 269 (it was delayed by 262);
5) We issue the move operation for inode 257 (it was delayed by 272);
6) We issue the move operation for inode 260 (it was delayed by 272);
7) We issue the move operation for inode 258 (it was delayed by 269);
8) We issue the move operation for inode 264 (it was delayed by 257);
9) We issue the move operation for inode 271 (it was delayed by 258);
10) We issue the move operation for inode 263 (it was delayed by 264);
11) We issue the move operation for inode 268 (it was delayed by 271);
12) We verify if we can issue the move operation for inode 270 (it was
delayed by 271). We detect a path loop in the current state, because
inode 267 needs to be moved first before we can issue the move
operation for inode 270. So we delay again the move operation for
inode 270, this time we will attempt to do it after inode 267 is
moved;
13) We issue the move operation for inode 261 (it was delayed by 263);
14) We verify if we can issue the move operation for inode 266 (it was
delayed by 263). We detect a path loop in the current state, because
inode 270 needs to be moved first before we can issue the move
operation for inode 266. So we delay again the move operation for
inode 266, this time we will attempt to do it after inode 270 is
moved (its move operation was delayed in step 12);
15) We issue the move operation for inode 267 (it was delayed by 268);
16) We verify if we can issue the move operation for inode 266 (it was
delayed by 270). We detect a path loop in the current state, because
inode 270 needs to be moved first before we can issue the move
operation for inode 266. So we delay again the move operation for
inode 266, this time we will attempt to do it after inode 270 is
moved (its move operation was delayed in step 12). So here we added
again the same delayed move operation that we added in step 14;
17) We attempt again to see if we can issue the move operation for inode
266, and as in step 16, we realize we can not due to a path loop in
the current state due to a dependency on inode 270. Again we delay
inode's 266 rename to happen after inode's 270 move operation, adding
the same dependency to the empty stack that we did in steps 14 and 16.
The next iteration will pick the same move dependency on the stack
(the only entry) and realize again there is still a path loop and then
again the same dependency to the stack, over and over, resulting in
an infinite loop.
So fix this by preventing adding the same move dependency entries to the
stack by removing each pending move record from the red black tree of
pending moves. This way the next call to get_pending_dir_moves() will
not return anything for the current parent inode.
A test case for fstests, with this reproducer, follows soon.
Merge tag 'linux-cpupower-4.20-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux
Pull cpupower utility updates for 4.20-rc4 from Shuah Khan:
"This cpupower update for Linux 4.20-rc4 consists of compile fixes to allow
use of outside build flags and override of CFLAGS from Jiri Olsa, and fix
to compilation with STATIC=true from Konstantin Khlebnikov."
* tag 'linux-cpupower-4.20-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux:
tools cpupower: Override CFLAGS assignments
tools cpupower debug: Allow to use outside build flags
tools/power/cpupower: fix compilation with STATIC=true
Ville Syrjälä [Tue, 20 Nov 2018 13:54:50 +0000 (15:54 +0200)]
drm/i915: Add rotation readout for plane initial config
If we need to force a full plane update before userspace/fbdev
have given us a proper plane state we should try to maintain the
current plane state as much as possible (apart from the parts
of the state we're trying to fix up with the plane update).
To that end add basic readout for the plane rotation and
maintain it during the initial fb takeover.
Ville Syrjälä [Tue, 20 Nov 2018 13:54:49 +0000 (15:54 +0200)]
drm/i915: Force a LUT update in intel_initial_commit()
If we force a plane update to fix up our half populated plane state
we'll also force on the pipe gamma for the plane (since we always
enable pipe gamma currently). If the BIOS hasn't programmed a sensible
LUT into the hardware this will cause the image to become corrupted.
Typical symptoms are a purple/yellow/etc. flash when the driver loads.
To avoid this let's program something sensible into the LUT when
we do the plane update. In the future I plan to add proper plane
gamma enable readout so this is just a temporary measure.
Hans de Goede [Mon, 19 Nov 2018 18:06:01 +0000 (19:06 +0100)]
ACPI / platform: Add SMB0001 HID to forbidden_id_list
Many HP AMD based laptops contain an SMB0001 device like this:
Device (SMBD)
{
Name (_HID, "SMB0001") // _HID: Hardware ID
Name (_CRS, ResourceTemplate () // _CRS: Current Resource Settings
{
IO (Decode16,
0x0B20, // Range Minimum
0x0B20, // Range Maximum
0x20, // Alignment
0x20, // Length
)
IRQ (Level, ActiveLow, Shared, )
{7}
})
}
The legacy style IRQ resource here causes acpi_dev_get_irqresource() to
be called with legacy=true and this message to show in dmesg:
ACPI: IRQ 7 override to edge, high
This causes issues when later on the AMD0030 GPIO device gets enumerated:
Device (GPIO)
{
Name (_HID, "AMDI0030") // _HID: Hardware ID
Name (_CID, "AMDI0030") // _CID: Compatible ID
Name (_UID, Zero) // _UID: Unique ID
Method (_CRS, 0, NotSerialized) // _CRS: Current Resource Settings
{
Name (RBUF, ResourceTemplate ()
{
Interrupt (ResourceConsumer, Level, ActiveLow, Shared, ,, )
{
0x00000007,
}
Memory32Fixed (ReadWrite,
0xFED81500, // Address Base
0x00000400, // Address Length
)
})
Return (RBUF) /* \_SB_.GPIO._CRS.RBUF */
}
}
Now acpi_dev_get_irqresource() gets called with legacy=false, but because
of the earlier override of the trigger-type acpi_register_gsi() returns
-EBUSY (because we try to register the same interrupt with a different
trigger-type) and we end up setting IORESOURCE_DISABLED in the flags.
The setting of IORESOURCE_DISABLED causes platform_get_irq() to call
acpi_irq_get() which is not implemented on x86 and returns -EINVAL.
resulting in the following in dmesg:
amd_gpio AMDI0030:00: Failed to get gpio IRQ: -22
amd_gpio: probe of AMDI0030:00 failed with error -22
The SMB0001 is a "virtual" device in the sense that the only way the OS
interacts with it is through calling a couple of methods to do SMBus
transfers. As such it is weird that it has IO and IRQ resources at all,
because the driver for it is not expected to ever access the hardware
directly.
The Linux driver for the SMB0001 device directly binds to the acpi_device
through the acpi_bus, so we do not need to instantiate a platform_device
for this ACPI device. This commit adds the SMB0001 HID to the
forbidden_id_list, avoiding the instantiating of a platform_device for it.
Not instantiating a platform_device means we will no longer call
acpi_dev_get_irqresource() for the legacy IRQ resource fixing the probe of
the AMDI0030 device failing.
drm/fb-helper: Blacklist writeback when adding connectors to fbdev
Writeback connectors do not produce any on-screen output and require
special care for use. Such connectors are hidden from enumeration in
DRM resources by default, but they are still picked-up by fbdev.
This makes rather little sense since fbdev is not really adapted for
dealing with writeback.
Moreover, this is also a source of issues when userspace disables the
CRTC (and associated plane) without detaching the CRTC from the
connector (which is hidden by default). In this case, the connector is
still using the CRTC, leading to am "enabled/connectors mismatch" and
eventually the failure of the associated atomic commit. This situation
happens with VC4 testing under IGT GPU Tools.
Filter out writeback connectors in the fbdev helper to solve this.
Chris Wilson [Mon, 19 Nov 2018 15:41:53 +0000 (15:41 +0000)]
drm/i915: Write GPU relocs harder with gen3
Under moderate amounts of GPU stress, we can observe on Bearlake and
Pineview (later gen3 models) that we execute the following batch buffer
before the write into the batch is coherent. Adding extra (tested with
upto 32x) MI_FLUSH to either the invalidation, flush or both phases does
not solve the incoherency issue with the relocations, but emitting the
MI_STORE_DWORD_IMM twice does. So be it.
Davide Caratti [Tue, 20 Nov 2018 21:18:44 +0000 (22:18 +0100)]
net/sched: act_police: fix race condition on state variables
after 'police' configuration parameters were converted to use RCU instead
of spinlock, the state variables used to compute the traffic rate (namely
'tcfp_toks', 'tcfp_ptoks' and 'tcfp_t_c') are erroneously read/updated in
the traffic path without any protection.
Use a dedicated spinlock to avoid race conditions on these variables, and
ensure proper cache-line alignment. In this way, 'police' is still faster
than what we observed when 'tcf_lock' was used in the traffic path _ i.e.
reverting commit 2d550dbad83c ("net/sched: act_police: don't use spinlock
in the data path"). Moreover, we preserve the throughput improvement that
was obtained after 'police' started using per-cpu counters, when 'avrate'
is used instead of 'rate'.
Changes since v1 (thanks to Eric Dumazet):
- call ktime_get_ns() before acquiring the lock in the traffic path
- use a dedicated spinlock instead of tcf_lock
- improve cache-line usage
Fixes: 2d550dbad83c ("net/sched: act_police: don't use spinlock in the data path") Reported-and-suggested-by: Eric Dumazet <[email protected]> Signed-off-by: Davide Caratti <[email protected]> Reviewed-by: Eric Dumazet <[email protected]>
Linus Torvalds [Tue, 20 Nov 2018 22:31:00 +0000 (14:31 -0800)]
Merge tag 'mips_fixes_4.20_3' of git://git.kernel.org/pub/scm/linux/kernel/git/mips/linux
Pull MIPS fixes from Paul Burton:
"A few MIPS fixes for 4.20:
- Re-enable the Cavium Octeon USB driver in its defconfig after it
was accidentally removed back in 4.14.
- Have early memblock allocations be performed bottom-up to more
closely match the behaviour we used to have with bootmem, which
seems a safer choice since we've seen fallout from the change made
in the 4.20 merge window.
- Simplify max_low_pfn calculation in the NUMA code for the Loongson3
and SGI IP27 platforms to both clean up the code & ensure
max_low_pfn has been set appropriately before it is used"
* tag 'mips_fixes_4.20_3' of git://git.kernel.org/pub/scm/linux/kernel/git/mips/linux:
MIPS: Loongson3,SGI-IP27: Simplify max_low_pfn calculation
MIPS: Let early memblock_alloc*() allocate memories bottom-up
MIPS: OCTEON: cavium_octeon_defconfig: re-enable OCTEON USB driver
Heiner Kallweit [Tue, 20 Nov 2018 20:22:50 +0000 (21:22 +0100)]
MAINTAINERS: add myself as co-maintainer for r8169
Meanwhile I know the driver quite well and I refactored bigger parts
of it. As a result people contact me already with r8169 questions.
Therefore I'd volunteer to become co-maintainer of the driver also
officially.
The extent shifting code uses a flush and invalidate mechainsm prior
to shifting extents around. This is similar to what
xfs_free_file_space() does, but it doesn't take into account things
like page cache vs block size differences, and it will fail if there
is a page that it currently busy.
xfs_flush_unmap_range() handles all of these cases, so just convert
xfs_prepare_shift() to us that mechanism rather than having it's own
special sauce.
Dave Chinner [Mon, 19 Nov 2018 21:31:08 +0000 (13:31 -0800)]
xfs: finobt AG reserves don't consider last AG can be a runt
The last AG may be very small comapred to all other AGs, and hence
AG reservations based on the superblock AG size may actually consume
more space than the AG actually has. This results on assert failures
like:
Hence we need to ensure the finobt per-ag space reservations take
into account the size of the last AG rather than treat it like all
the other full size AGs.
Note that both refcountbt and rmapbt already take the size of the AG
into account via reading the AGF length directly.
Dave Chinner [Mon, 19 Nov 2018 21:31:08 +0000 (13:31 -0800)]
xfs: fix transient reference count error in xfs_buf_resubmit_failed_buffers
When retrying a failed inode or dquot buffer,
xfs_buf_resubmit_failed_buffers() clears all the failed flags from
the inde/dquot log items. In doing so, it also drops all the
reference counts on the buffer that the failed log items hold. This
means it can drop all the active references on the buffer and hence
free the buffer before it queues it for write again.
Putting the buffer on the delwri queue takes a reference to the
buffer (so that it hangs around until it has been written and
completed), but this goes bang if the buffer has already been freed.
Hence we need to add the buffer to the delwri queue before we remove
the failed flags from the log items attached to the buffer to ensure
it always remains referenced during the resubmit process.
Stephen Mallon [Tue, 20 Nov 2018 08:15:02 +0000 (19:15 +1100)]
tcp: Fix SOF_TIMESTAMPING_RX_HARDWARE to use the latest timestamp during TCP coalescing
During tcp coalescing ensure that the skb hardware timestamp refers to the
highest sequence number data.
Previously only the software timestamp was updated during coalescing.
tg3: Add PHY reset for 5717/5719/5720 in change ring and flow control paths
This patch has the fix to avoid PHY lockup with 5717/5719/5720 in change
ring and flow control paths. This patch solves the RX hang while doing
continuous ring or flow control parameters with heavy traffic from peer.
Will Deacon [Mon, 19 Nov 2018 11:07:18 +0000 (11:07 +0000)]
Documentation/security-bugs: Postpone fix publication in exceptional cases
At the request of the reporter, the Linux kernel security team offers to
postpone the publishing of a fix for up to 5 business days from the date
of a report.
While it is generally undesirable to keep a fix private after it has
been developed, this short window is intended to allow distributions to
package the fix into their kernel builds and permits early inclusion of
the security team in the case of a co-ordinated disclosure with other
parties. Unfortunately, discussions with major Linux distributions and
cloud providers has revealed that 5 business days is not sufficient to
achieve either of these two goals.
As an example, cloud providers need to roll out KVM security fixes to a
global fleet of hosts with sufficient early ramp-up and monitoring. An
end-to-end timeline of less than two weeks dramatically cuts into the
amount of early validation and increases the chance of guest-visible
regressions.
The consequence of this timeline mismatch is that security issues are
commonly fixed without the involvement of the Linux kernel security team
and are instead analysed and addressed by an ad-hoc group of developers
across companies contributing to Linux. In some cases, mainline (and
therefore the official stable kernels) can be left to languish for
extended periods of time. This undermines the Linux kernel security
process and puts upstream developers in a difficult position should they
find themselves involved with an undisclosed security problem that they
are unable to report due to restrictions from their employer.
To accommodate the needs of these users of the Linux kernel and
encourage them to engage with the Linux security team when security
issues are first uncovered, extend the maximum period for which fixes
may be delayed to 7 calendar days, or 14 calendar days in exceptional
cases, where the logistics of QA and large scale rollouts specifically
need to be accommodated. This brings parity with the linux-distros@
maximum embargo period of 14 calendar days.
MAINTAINERS: Add Sasha as a stable branch maintainer
Sasha has somehow been convinced into helping me with the stable kernel
maintenance. Codify this slip in good judgement before he realizes what
he really signed up for :)
Linus Torvalds [Tue, 20 Nov 2018 15:37:15 +0000 (07:37 -0800)]
Merge tag 'media/v4.20-3' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media
Pull media fixes from Mauro Carvalho Chehab:
- add a missing include at v4l2-controls uAPI header
- minor kAPI update for the request API
- some fixes at CEC core
- use a lower minimum height for the virtual codec driver
- cleanup a gcc warning due to the lack of a fall though markup
- tc358743: Remove unnecessary self assignment
- fix the V4L event subscription logic
- docs: Document metadata format in struct v4l2_format
- omap3isp and ipu3-cio2: fix unbinding logic
* tag 'media/v4.20-3' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media:
media: ipu3-cio2: Use cio2_queues_exit
media: ipu3-cio2: Unregister device nodes first, then release resources
media: omap3isp: Unregister media device as first
media: docs: Document metadata format in struct v4l2_format
media: v4l: event: Add subscription to list before calling "add" operation
media: dm365_ipipeif: better annotate a fall though
media: Rename vb2_m2m_request_queue -> v4l2_m2m_request_queue
media: cec: increase debug level for 'queue full'
media: cec: check for non-OK/NACK conditions while claiming a LA
media: vicodec: lower minimum height to 360
media: tc358743: Remove unnecessary self assignment
media: v4l: fix uapi mpeg slice params definition
v4l2-controls: add a missing include
mtd: spi-nor: fix selection of uniform erase type in flexible conf
There are uniform, non-uniform and flexible erase flash configurations.
The non-uniform erase types, are the erase types that can _not_ erase
the entire flash by their own.
As the code was, in case flashes had flexible erase capabilities
(support both uniform and non-uniform erase types in the same flash
configuration) and supported multiple uniform erase type sizes, the
code did not sort the uniform erase types, and could select a wrong
erase type size.
Sort the uniform erase mask in case of flexible erase flash
configurations, in order to select the best uniform erase type size.
Uniform, non-uniform, and flexible configurations with just a valid
uniform erase type, are not affected by this change.
Uniform erase tested on mx25l3273fm2i-08g and sst26vf064B-104i/sn.
Non uniform erase tested on sst26vf064B-104i/sn.
Fixes: 5390a8df769e ("mtd: spi-nor: add support to non-uniform SFDP SPI NOR flash memories") Signed-off-by: Tudor Ambarus <[email protected]> Signed-off-by: Boris Brezillon <[email protected]>
Patrick Stählin [Fri, 9 Nov 2018 21:42:16 +0000 (22:42 +0100)]
RISC-V: recognize S/U mode bits in print_isa
Removes the warning about an unsupported ISA when reading /proc/cpuinfo
on QEMU. The "S" extension is not being returned as it is not accessible
from userspace.
Marcin Juszkiewicz reported issues while generating syscall table for riscv
using 4.20-rc1. The patch refactors our unistd.h files to match some other
architectures.
- Add asm/unistd.h UAPI header, which has __ARCH_WANT_NEW_STAT only for 64-bit
- Remove asm/syscalls.h UAPI header and merge to asm/unistd.h
- Adjust kernel asm/unistd.h
So now asm/unistd.h UAPI header should show all syscalls for riscv.
Before this, Makefile simply put `#include <asm-generic/unistd.h>` into
generated asm/unistd.h UAPI header thus user didn't see:
Anup Patel [Mon, 12 Nov 2018 05:55:15 +0000 (11:25 +0530)]
RISC-V: Build flat and compressed kernel images
This patch extends Linux RISC-V build system to build and install:
Image - Flat uncompressed kernel image
Image.gz - Flat and GZip compressed kernel image
Quiet a few bootloaders (such as Uboot, UEFI, etc) are capable of
booting flat and compressed kernel images. In case of Uboot, booting
Image or Image.gz is achieved using bootm command.
The flat and uncompressed kernel image (i.e. Image) is very useful
in pre-silicon developent and testing because we can create back-door
HEX files for RAM on FPGAs from Image.
Olof Johansson [Thu, 15 Nov 2018 00:27:55 +0000 (16:27 -0800)]
RISC-V: Fix raw_copy_{to,from}_user()
Sparse highlighted it, and appears to be a pure bug (from vs to).
./arch/riscv/include/asm/uaccess.h:403:35: warning: incorrect type in argument 1 (different address spaces)
./arch/riscv/include/asm/uaccess.h:403:39: warning: incorrect type in argument 2 (different address spaces)
./arch/riscv/include/asm/uaccess.h:409:37: warning: incorrect type in argument 1 (different address spaces)
./arch/riscv/include/asm/uaccess.h:409:41: warning: incorrect type in argument 2 (different address spaces)
The PixArt OEM mice are known for disconnecting every minute in
runlevel 1 or 3 if they are not always polled. So add quirk
ALWAYS_POLL for two Primax mice as well.
0x4e22 is the Dell MS111-P and 0x4d0f is the unbranded HP Portia
mouse HP 697738-001. Both were built until approx. 2014.
Those were the standard mice from those vendors and are still
around - even as new old stock.
Maarten Jacobs [Mon, 19 Nov 2018 23:18:49 +0000 (23:18 +0000)]
usb: cdc-acm: add entry for Hiro (Conexant) modem
The cdc-acm kernel module currently does not support the Hiro (Conexant)
H05228 USB modem. The patch below adds the device specific information:
idVendor 0x0572
idProduct 0x1349
Chris Wilson [Fri, 2 Nov 2018 16:12:12 +0000 (16:12 +0000)]
drm/i915: Prevent machine hang from Broxton's vtd w/a and error capture
Since capturing the error state requires fiddling around with the GGTT
to read arbitrary buffers and is itself run under stop_machine(), it
deadlocks the machine (effectively a hard hang) when run in conjunction
with Broxton's VTd workaround to serialize GGTT access.
v2: Store the ERR_PTR in first_error so that the error can be reported
to the user via sysfs.
v3: Mention the quirk in dmesg (using info as per usual)
The following fixes are for mlx5 core and netdev driver.
For -stable v4.16 bc7fda7d4637 ('net/mlx5e: IPoIB, Reset QP after channels are closed')
For -stable v4.17 36917a270395 ('net/mlx5: IPSec, Fix the SA context hash key')
For -stable v4.18 6492a432be3a ('net/mlx5e: Always use the match level enum when parsing TC rule match') c3f81be236b1 ('net/mlx5e: Removed unnecessary warnings in FEC caps query') c5ce2e736b64 ('net/mlx5e: Fix selftest for small MTUs')
For -stable v4.19 effcd896b25e ('net/mlx5e: Adjust to max number of channles when re-attaching') 394cbc5acd68 ('net/mlx5e: RX, verify received packet size in Linear Striding RQ') 447cbb3613c8 ('net/mlx5e: Don't match on vlan non-existence if ethertype is wildcarded') c223c1574612 ('net/mlx5e: Claim TC hw offloads support only under a proper build config')
Please pull and let me know if there's any problem.
====================
Juliet Kim [Mon, 19 Nov 2018 21:59:22 +0000 (15:59 -0600)]
net/ibmnvic: Fix deadlock problem in reset
This patch changes to use rtnl_lock only during a reset to avoid
deadlock that could occur when a thread operating close is holding
rtnl_lock and waiting for reset_lock acquired by another thread,
which is waiting for rtnl_lock in order to set the number of tx/rx
queues during a reset.
Also, we now setting the number of tx/rx queues during a soft reset
for failover or LPM events.
This patch series fixes various queue manager getter functions. It is
important to make sure the getter's caller will receive a valid queue even
in error case to prevent more serious bugs.
Please consider applying to net.
====================
Denis Bolotin [Mon, 19 Nov 2018 14:28:31 +0000 (16:28 +0200)]
qed: Fix QM getters to always return a valid pq
The getter callers doesn't know the valid Physical Queues (PQ) values.
This patch makes sure that a valid PQ will always be returned.
The patch consists of 3 fixes:
- When qed_init_qm_get_idx_from_flags() receives a disabled flag, it
returned PQ 0, which can potentially be another function's pq. Verify
that flag is enabled, otherwise return default start_pq.
- When qed_init_qm_get_idx_from_flags() receives an unknown flag, it
returned NULL and could lead to a segmentation fault. Return default
start_pq instead.
- A modulo operation was added to MCOS/VFS PQ getters to make sure the
PQ returned is in range of the required flag.
Trond Myklebust [Tue, 20 Nov 2018 01:11:45 +0000 (20:11 -0500)]
NFSv4: Fix a NFSv4 state manager deadlock
Fix a deadlock whereby the NFSv4 state manager can get stuck in the
delegation return code, waiting for a layout return to complete in
another thread. If the server reboots before that other thread
completes, then we need to be able to start a second state
manager thread in order to perform recovery.
Shay Agroskin [Thu, 8 Nov 2018 12:23:36 +0000 (14:23 +0200)]
net/mlx5e: Fix failing ethtool query on FEC query error
If FEC caps query fails when executing 'ethtool <interface>'
the whole callback fails unnecessarily, fixed that by replacing the
error return code with debug logging only.
Shay Agroskin [Sun, 28 Oct 2018 07:06:11 +0000 (09:06 +0200)]
net/mlx5e: Removed unnecessary warnings in FEC caps query
Querying interface FEC caps with 'ethtool [int]' after link reset
throws warning regading link speed.
This warning is not needed as there is already an indication in
user space that the link is not up.
Shay Agroskin [Sun, 28 Oct 2018 14:13:46 +0000 (16:13 +0200)]
net/mlx5e: Fix a bug in turning off FEC policy in unsupported speeds
Some speeds don't support turning FEC policy off. In case a requested
FEC policy is not supported for a speed (including current speed), its new
FEC policy would be:
no FEC - if disabling FEC is supported for that speed
unchanged - else
====================
net: ena: hibernation and rmmod bug fixes
This patchset includes 2 bug fixes:
1. A fix to a crash during resume from hibernation.
2. A fix to an illegal memory access during driver removal (e.g. during rmmod)
which might cause a crash in certain systems.
The subminor number in the driver version is also promoted to indicate driver
was changed.
====================
In ena_remove() we have the following stack call:
ena_remove()
unregister_netdev()
ena_destroy_device()
netif_carrier_off()
Calling netif_carrier_off() causes linkwatch to try to handle the
link change event on the already unregistered netdev, which leads
to a read from an unreadable memory address.
This patch switches the order of the two functions, so that
netif_carrier_off() is called on a regiestered netdev.
To accomplish this fix we also had to:
1. Remove the set bit ENA_FLAG_TRIGGER_RESET
2. Add a sanitiy check in ena_close()
both to prevent double device reset (when calling unregister_netdev()
ena_close is called, but the device was already deleted in
ena_destroy_device()).
3. Set the admin_queue running state to false to avoid using it after
device was reset (for example when calling ena_destroy_all_io_queues()
right after ena_com_dev_reset() in ena_down)
Fixes: 944b28aa2982 ("net: ena: fix missing lock during device destruction") Signed-off-by: Arthur Kiyanovski <[email protected]> Signed-off-by: David S. Miller <[email protected]>
net: ena: fix crash during failed resume from hibernation
During resume from hibernation if ena_restore_device fails,
ena_com_dev_reset() is called, and uses the readless read mechanism,
which was already destroyed by the call to
ena_com_mmio_reg_read_request_destroy(). This causes a NULL pointer
reference.
In this commit we switch the call order of the above two functions
to avoid this crash.
Fixes: d7703ddbd7c9 ("net: ena: fix rare bug when failed restart/resume is followed by driver removal") Signed-off-by: Arthur Kiyanovski <[email protected]> Signed-off-by: David S. Miller <[email protected]>
Xin Long [Sun, 18 Nov 2018 13:59:49 +0000 (21:59 +0800)]
sctp: not increase stream's incnt before sending addstrm_in request
Different from processing the addstrm_out request, The receiver handles
an addstrm_in request by sending back an addstrm_out request to the
sender who will increase its stream's in and incnt later.
Now stream->incnt has been increased since it sent out the addstrm_in
request in sctp_send_add_streams(), with the wrong stream->incnt will
even cause crash when copying stream info from the old stream's in to
the new one's in sctp_process_strreset_addstrm_out().
This patch is to fix it by simply removing the stream->incnt change
from sctp_send_add_streams().
Fixes: 242bd2d519d7 ("sctp: implement sender-side procedures for Add Incoming/Outgoing Streams Request Parameter") Reported-by: Jianwen Ji <[email protected]> Signed-off-by: Xin Long <[email protected]> Signed-off-by: David S. Miller <[email protected]>
Valentine Fatiev [Wed, 17 Oct 2018 08:45:07 +0000 (11:45 +0300)]
net/mlx5e: Fix selftest for small MTUs
Loopback test had fixed packet size, which can be bigger than configured
MTU. Shorten the loopback packet size to be bigger than minimal MTU
allowed by the device. Text field removed from struct 'mlx5ehdr'
as redundant to allow send small packets as minimal allowed MTU.
Fixes: d605d66 ("net/mlx5e: Add support for ethtool self diagnostics test") Signed-off-by: Valentine Fatiev <[email protected]> Reviewed-by: Eran Ben Elisha <[email protected]> Signed-off-by: Saeed Mahameed <[email protected]>
Moshe Shemesh [Thu, 11 Oct 2018 04:31:10 +0000 (07:31 +0300)]
net/mlx5e: RX, verify received packet size in Linear Striding RQ
In case of striding RQ, we use MPWRQ (Multi Packet WQE RQ), which means
that WQE (RX descriptor) can be used for many packets and so the WQE is
much bigger than MTU. In virtualization setups where the port mtu can
be larger than the vf mtu, if received packet is bigger than MTU, it
won't be dropped by HW on too small receive WQE. If we use linear SKB in
striding RQ, since each stride has room for mtu size payload and skb
info, an oversized packet can lead to crash for crossing allocated page
boundary upon the call to build_skb. So driver needs to check packet
size and drop it.
Introduce new SW rx counter, rx_oversize_pkts_sw_drop, which counts the
number of packets dropped by the driver for being too large.
As a new field is added to the RQ struct, re-open the channels whenever
this field is being used in datapath (i.e., in the case of linear
Striding RQ).
Fixes: 619a8f2a42f1 ("net/mlx5e: Use linear SKB in Striding RQ") Signed-off-by: Moshe Shemesh <[email protected]> Reviewed-by: Tariq Toukan <[email protected]> Signed-off-by: Saeed Mahameed <[email protected]>
Roi Dayan [Tue, 23 Oct 2018 14:30:04 +0000 (17:30 +0300)]
net/mlx5e: Apply the correct check for supporting TC esw rules split
The mirror and not the output count is the one denoting a split.
Fix to condition the offload attempt on the mirror count being > 0
along the firmware to have the related capability.
Yuval Avnery [Tue, 16 Oct 2018 20:20:20 +0000 (15:20 -0500)]
net/mlx5e: Adjust to max number of channles when re-attaching
When core driver enters deattach/attach flow after pci reset,
Number of logical CPUs may have changed.
As a result we need to update the cpu affiliated resource tables.
1. indirect rqt list
2. eq table
Reproduction (PowerPC):
echo 1000 > /sys/kernel/debug/powerpc/eeh_max_freezes
ppc64_cpu --smt=on
# Restart driver
modprobe -r ... ; modprobe ...
# Link up
ifconfig ...
# Only physical CPUs
ppc64_cpu --smt=off
# Inject PCI errors so PCI will reset - calling the pci error handler
echo 0x8000000000000000 > /sys/kernel/debug/powerpc/<PCI BUS>/err_injct_inboundA
Call trace when trying to add non-existing rqs to an indirect rqt:
mlx5e_redirect_rqt+0x84/0x260 [mlx5_core] (unreliable)
mlx5e_redirect_rqts+0x188/0x190 [mlx5_core]
mlx5e_activate_priv_channels+0x488/0x570 [mlx5_core]
mlx5e_open_locked+0xbc/0x140 [mlx5_core]
mlx5e_open+0x50/0x130 [mlx5_core]
mlx5e_nic_enable+0x174/0x1b0 [mlx5_core]
mlx5e_attach_netdev+0x154/0x290 [mlx5_core]
mlx5e_attach+0x88/0xd0 [mlx5_core]
mlx5_attach_device+0x168/0x1e0 [mlx5_core]
mlx5_load_one+0x1140/0x1210 [mlx5_core]
mlx5_pci_resume+0x6c/0xf0 [mlx5_core]
Create cq will fail when trying to use non-existing EQ.
Or Gerlitz [Sun, 28 Oct 2018 10:27:29 +0000 (12:27 +0200)]
net/mlx5e: Always use the match level enum when parsing TC rule match
We get the match level (none, l2, l3, l4) while going over the match
dissectors of an offloaded tc rule. When doing this, the match level
enum and the not min inline enum values should be used, fix that.
This worked accidentally b/c both enums have the same numerical values.
Fixes: d708f902989b ('net/mlx5e: Get the required HW match level while parsing TC flow matches') Signed-off-by: Or Gerlitz <[email protected]> Reviewed-by: Roi Dayan <[email protected]> Signed-off-by: Saeed Mahameed <[email protected]>
Or Gerlitz [Thu, 18 Oct 2018 10:31:27 +0000 (12:31 +0200)]
net/mlx5e: Claim TC hw offloads support only under a proper build config
Currently, we are only supporting tc hw offloads when the eswitch
support is compiled in, but we are not gating the adevertizment
of the NETIF_F_HW_TC feature on this config being set.
Fix it, and while doing that, also avoid dealing with the feature
on ethtool when the config is not set.
Or Gerlitz [Thu, 25 Oct 2018 15:41:58 +0000 (15:41 +0000)]
net/mlx5e: Don't match on vlan non-existence if ethertype is wildcarded
For the "all" ethertype we should not care whether the packet has
vlans. Besides being wrong, the way we did it caused FW error
for rules such as:
tc filter add dev eth0 protocol all parent ffff: \
prio 1 flower skip_sw action drop
b/c the matching meta-data (outer headers bit in struct mlx5_flow_spec)
wasn't set. Fix that by matching on vlan non-existence only if we were
also told to match on the ethertype.
net/mlx5e: IPoIB, Reset QP after channels are closed
The mlx5e channels should be closed before mlx5i_uninit_underlay_qp
puts the QP into RST (reset) state during mlx5i_close. Currently QP
state incorrectly set to RST before channels got deactivated and closed,
since mlx5_post_send request expects QP in RTS (Ready To Send) state.
The fix is to keep QP in RTS state until mlx5e channels get closed
and to reset QP afterwards.
Also this fix is simply correct in order to keep the open/close flow
symmetric, i.e mlx5i_init_underlay_qp() is called first thing at open,
the correct thing to do is to call mlx5i_uninit_underlay_qp() last thing
at close, which is exactly what this patch is doing.
Fixes: dae37456c8ac ("net/mlx5: Support for attaching multiple underlay QPs to root flow table") Signed-off-by: Denis Drozdov <[email protected]> Signed-off-by: Saeed Mahameed <[email protected]>
Raed Salem [Thu, 18 Oct 2018 05:55:21 +0000 (08:55 +0300)]
net/mlx5: IPSec, Fix the SA context hash key
The commit "net/mlx5: Refactor accel IPSec code" introduced a
bug where asynchronous short time change in hash key value
by create/release SA context might happen during an asynchronous
hash resize operation this could cause a subsequent remove SA
context operation to fail as the key value used during resize is
not the same key value used when remove SA context operation is
invoked.
This commit fixes the bug by defining the SA context hash key
such that it includes only fields that never change during the
lifetime of the SA context object.
Brian Foster [Wed, 14 Nov 2018 15:46:40 +0000 (07:46 -0800)]
xfs: fix shared extent data corruption due to missing cow reservation
Page writeback indirectly handles shared extents via the existence
of overlapping COW fork blocks. If COW fork blocks exist, writeback
always performs the associated copy-on-write regardless if the
underlying blocks are actually shared. If the blocks are shared,
then overlapping COW fork blocks must always exist.
fstests shared/010 reproduces a case where a buffered write occurs
over a shared block without performing the requisite COW fork
reservation. This ultimately causes writeback to the shared extent
and data corruption that is detected across md5 checks of the
filesystem across a mount cycle.
The problem occurs when a buffered write lands over a shared extent
that crosses an extent size hint boundary and that also happens to
have a partial COW reservation that doesn't cover the start and end
blocks of the data fork extent.
For example, a buffered write occurs across the file offset (in FSB
units) range of [29, 57]. A shared extent exists at blocks [29, 35]
and COW reservation already exists at blocks [32, 34]. After
accommodating a COW extent size hint of 32 blocks and the existing
reservation at offset 32, xfs_reflink_reserve_cow() allocates 32
blocks of reservation at offset 0 and returns with COW reservation
across the range of [0, 34]. The associated data fork extent is
still [29, 35], however, which isn't fully covered by the COW
reservation.
This leads to a buffered write at file offset 35 over a shared
extent without associated COW reservation. Writeback eventually
kicks in, performs an overwrite of the underlying shared block and
causes the associated data corruption.
Update xfs_reflink_reserve_cow() to accommodate the fact that a
delalloc allocation request may not fully cover the extent in the
data fork. Trim the data fork extent appropriately, just as is done
for shared extent boundaries and/or existing COW reservations that
happen to overlap the start of the data fork extent. This prevents
shared/010 failures due to data corruption on reflink enabled
filesystems.
The current Cadence QSPI driver caused a kernel panic sporadically
when writing to QSPI. The problem was caused by writing more bytes
than needed because the QSPI operated on 4 bytes at a time.
<snip>
[ 11.202044] Unable to handle kernel paging request at virtual address bffd3000
[ 11.209254] pgd = e463054d
[ 11.211948] [bffd3000] *pgd=2fffb811, *pte=00000000, *ppte=00000000
[ 11.218202] Internal error: Oops: 7 [#1] SMP ARM
[ 11.222797] Modules linked in:
[ 11.225844] CPU: 1 PID: 1317 Comm: systemd-hwdb Not tainted 4.17.7-d0c45cd44a8f
[ 11.235796] Hardware name: Altera SOCFPGA Arria10
[ 11.240487] PC is at __raw_writesl+0x70/0xd4
[ 11.244741] LR is at cqspi_write+0x1a0/0x2cc
</snip>
On a page boundary limit the number of bytes copied from the tx buffer
to remain within the page.
This patch uses a temporary buffer to hold the 4 bytes to write and then
copies only the bytes required from the tx buffer.
drm/amd/pp: handle negative values when reading OD
Reading the sysfs files pp_sclk_od and pp_mclk_od return the
percentage difference between the VBIOS-provided default
frequency and the current (possibly user-set) frequency in
the highest SCLK and MCLK DPM states, respectively.
Writing to these files provides an easy mechanism for
setting a higher-than-default maximum frequency. We
normally only allow values >= 0 to be written here.
However, with the addition of pp_od_clk_voltage, we now
allow users to set custom DPM tables. If they then set
the maximum DPM state to something less than the default,
later reads of pp_*_od should return a negative value.
The highest DPM state is now less than the VBIOS-provided
default, so the percentage is negative.
The math to calculate this was originally performed with
unsigned values, meaning reads that should return negative
values returned meaningless data. This patch corrects that
issue and normalizes how all of the calculations are done
across the various hwmgr types.
The dst's mtu in transport can be updated by a non sctp place like
in xfrm where the MTU information didn't get synced between asoc,
transport and dst, so it is still needed to do the pmtu check
in sctp_packet_config.
Xin Long [Sun, 18 Nov 2018 07:21:53 +0000 (15:21 +0800)]
sctp: not allow to set asoc prsctp_enable by sockopt
As rfc7496#section4.5 says about SCTP_PR_SUPPORTED:
This socket option allows the enabling or disabling of the
negotiation of PR-SCTP support for future associations. For existing
associations, it allows one to query whether or not PR-SCTP support
was negotiated on a particular association.
It means only sctp sock's prsctp_enable can be set.
Note that for the limitation of SCTP_{CURRENT|ALL}_ASSOC, we will
add it when introducing SCTP_{FUTURE|CURRENT|ALL}_ASSOC for linux
sctp in another patchset.
v1->v2:
- drop the params.assoc_id check as Neil suggested.
Xin Long [Sun, 18 Nov 2018 07:07:38 +0000 (15:07 +0800)]
sctp: count sk_wmem_alloc by skb truesize in sctp_packet_transmit
Now sctp increases sk_wmem_alloc by 1 when doing set_owner_w for the
skb allocked in sctp_packet_transmit and decreases by 1 when freeing
this skb.
But when this skb goes through networking stack, some subcomponents
might change skb->truesize and add the same amount on sk_wmem_alloc.
However sctp doesn't know the amount to decrease by, it would cause
a leak on sk->sk_wmem_alloc and the sock can never be freed.
Xiumei found this issue when it hit esp_output_head() by using sctp
over ipsec, where skb->truesize is added and so is sk->sk_wmem_alloc.
Since sctp has used sk_wmem_queued to count for writable space since
Commit cd305c74b0f8 ("sctp: use sk_wmem_queued to check for writable
space"), it's ok to fix it by counting sk_wmem_alloc by skb truesize
in sctp_packet_transmit.
Fixes: cac2661c53f3 ("esp4: Avoid skb_cow_data whenever possible") Reported-by: Xiumei Mu <[email protected]> Signed-off-by: Xin Long <[email protected]> Signed-off-by: David S. Miller <[email protected]>
projects / linux.git / log