Both virtio net and napi_get_frags() allocate skbs
with a very small skb->head
While using page fragments instead of a kmalloc backed skb->head might give
a small performance improvement in some cases, there is a huge risk of
under estimating memory usage.
For both GOOD_COPY_LEN and GRO_MAX_HEAD, we can fit at least 32 allocations
per page (order-3 page in x86), or even 64 on PowerPC
We have been tracking OOM issues on GKE hosts hitting tcp_mem limits
but consuming far more memory for TCP buffers than instructed in tcp_mem[2]
Even if we force napi_alloc_skb() to only use order-0 pages, the issue
would still be there on arches with PAGE_SIZE >= 32768
This patch makes sure that small skb head are kmalloc backed, so that
other objects in the slab page can be reused instead of being held as long
as skbs are sitting in socket queues.
Note that we might in the future use the sk_buff napi cache,
instead of going through a more expensive __alloc_skb()
Another idea would be to use separate page sizes depending
on the allocated length (to never have more than 4 frags per page)
I would like to thank Greg Thelen for his precious help on this matter,
analysing crash dumps is always a time consuming task.
Fixes: fd11a83dd363 ("net: Pull out core bits of __netdev_alloc_skb and add __napi_alloc_skb") Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Paolo Abeni <pabeni@redhat.com> Cc: Greg Thelen <gthelen@google.com> Reviewed-by: Alexander Duyck <alexanderduyck@fb.com> Acked-by: Michael S. Tsirkin <mst@redhat.com> Link: https://lore.kernel.org/r/20210113161819.1155526-1-eric.dumazet@gmail.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
net/rxrpc/key.c:657:11: warning: Assigned value is garbage or undefined
toksize = toksizes[tok++];
^ ~~~~~~~~~~~~~~~
rxrpc_read() contains two consecutive loops. The first loop calculates the
token sizes and stores the results in toksizes[] and the second one uses
the array. When there is an error in identifying the token in the first
loop, the token is skipped, no change is made to the toksizes[] array.
When the same error happens in the second loop, the token is not skipped.
This will cause the toksizes[] array to be out of step and will overrun
past the calculated sizes.
Fix this by making both loops log a message and return an error in this
case. This should only happen if a new token type is incompletely
implemented, so it should normally be impossible to trigger this.
Fixes: 9a059cd5ca7d ("rxrpc: Downgrade the BUG() for unsupported token type in rxrpc_read()") Reported-by: Tom Rix <trix@redhat.com> Signed-off-by: David Howells <dhowells@redhat.com> Reviewed-by: Tom Rix <trix@redhat.com> Link: https://lore.kernel.org/r/161046503122.2445787.16714129930607546635.stgit@warthog.procyon.org.uk Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
We need to unregister the netdevice if config failed.
.ndo_uninit takes care of most of the heavy lifting.
This was uncovered by recent commit c269a24ce057 ("net: make
free_netdev() more lenient with unregistering devices").
Previously the partially-initialized device would be left
in the system.
Reported-and-tested-by: syzbot+2393580080a2da190f04@syzkaller.appspotmail.com Fixes: e2f1f072db8d ("sit: allow to configure 6rd tunnels via netlink") Acked-by: Nicolas Dichtel <nicolas.dichtel@6wind.com> Link: https://lore.kernel.org/r/20210114012947.2515313-1-kuba@kernel.org Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In commit 826f328e2b7e ("net: dcb: Validate netlink message in DCB
handler"), Linux started rejecting RTM_GETDCB netlink messages if they
contained a set-like DCB_CMD_ command.
The reason was that privileges were only verified for RTM_SETDCB messages,
but the value that determined the action to be taken is the command, not
the message type. And validation of message type against the DCB command
was the obvious missing piece.
Unfortunately it turns out that mlnx_qos, a somewhat widely deployed tool
for configuration of DCB, accesses the DCB set-like APIs through
RTM_GETDCB.
Therefore do not bounce the discrepancy between message type and command.
Instead, in addition to validating privileges based on the actual message
type, validate them also based on the expected message type. This closes
the loophole of allowing DCB configuration on non-admin accounts, while
maintaining backward compatibility.
Fixes: 2f90b8657ec9 ("ixgbe: this patch adds support for DCB to the kernel and ixgbe driver") Fixes: 826f328e2b7e ("net: dcb: Validate netlink message in DCB handler") Signed-off-by: Petr Machata <petrm@nvidia.com> Link: https://lore.kernel.org/r/a3edcfda0825f2aa2591801c5232f2bbf2d8a554.1610384801.git.me@pmachata.org Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
DCB uses the same handler function for both RTM_GETDCB and RTM_SETDCB
messages. dcb_doit() bounces RTM_SETDCB mesasges if the user does not have
the CAP_NET_ADMIN capability.
However, the operation to be performed is not decided from the DCB message
type, but from the DCB command. Thus DCB_CMD_*_GET commands are used for
reading DCB objects, the corresponding SET and DEL commands are used for
manipulation.
The assumption is that set-like commands will be sent via an RTM_SETDCB
message, and get-like ones via RTM_GETDCB. However, this assumption is not
enforced.
It is therefore possible to manipulate DCB objects without CAP_NET_ADMIN
capability by sending the corresponding command in an RTM_GETDCB message.
That is a bug. Fix it by validating the type of the request message against
the type used for the response.
MSFT ActiveSync implementation requires that the size of the response for
incoming query is to be provided in the request input length. Failure to
set the input size proper results in failed request transfer, where the
ActiveSync counterpart reports the NDIS_STATUS_INVALID_LENGTH (0xC0010014L)
error.
Set the input size for OID_GEN_PHYSICAL_MEDIUM query to the expected size
of the response in order for the ActiveSync to properly respond to the
request.
Fixes: 039ee17d1baa ("rndis_host: Add RNDIS physical medium checking into generic_rndis_bind()") Signed-off-by: Andrey Zhizhikin <andrey.zhizhikin@leica-geosystems.com> Link: https://lore.kernel.org/r/20210108095839.3335-1-andrey.zhizhikin@leica-geosystems.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
For all PCI functions on the netxen_nic adapter, interrupt
mode (INTx or MSI) configuration is dependent on what has
been configured by the PCI function zero in the shared
interrupt register, as these adapters do not support mixed
mode interrupts among the functions of a given adapter.
Logic for setting MSI/MSI-x interrupt mode in the shared interrupt
register based on PCI function id zero check is not appropriate for
all family of netxen adapters, as for some of the netxen family
adapters PCI function zero is not really meant to be probed/loaded
in the host but rather just act as a management function on the device,
which caused all the other PCI functions on the adapter to always use
legacy interrupt (INTx) mode instead of choosing MSI/MSI-x interrupt mode.
This patch replaces that check with port number so that for all
type of adapters driver attempts for MSI/MSI-x interrupt modes.
Aligning to tx_ndp_modulus is not sufficient because the next align
call can be cdc_ncm_align_tail, which can add up to ctx->tx_modulus +
ctx->tx_remainder - 1 bytes. This used to lead to occasional crashes
on a Huawei 909s-120 LTE module as follows:
- the condition marked /* if there is a remaining skb [...] */ is true
so the swaps happen
- skb_out is set from ctx->tx_curr_skb
- skb_out->len is exactly 0x3f52
- ctx->tx_curr_size is 0x4000 and delayed_ndp_size is 0xac
(note that the sum of skb_out->len and delayed_ndp_size is 0x3ffe)
- the for loop over n is executed once
- the cdc_ncm_align_tail call marked /* align beginning of next frame */
increases skb_out->len to 0x3f56 (the sum is now 0x4002)
- the condition marked /* check if we had enough room left [...] */ is
false so we break out of the loop
- the condition marked /* If requested, put NDP at end of frame. */ is
true so the NDP is written into skb_out
- now skb_out->len is 0x4002, so padding_count is minus two interpreted
as an unsigned number, which is used as the length argument to memset,
leading to a crash with various symptoms but usually including
The cdc_ncm_align_tail call first aligns on a ctx->tx_modulus
boundary (adding at most ctx->tx_modulus-1 bytes), then adds
ctx->tx_remainder bytes. Alternatively, the next alignment call can
occur in cdc_ncm_ndp16 or cdc_ncm_ndp32, in which case at most
ctx->tx_ndp_modulus-1 bytes are added.
A similar problem has occurred before, and the code is nontrivial to
reason about, so add a guard before the crashing call. By that time it
is too late to prevent any memory corruption (we'll have written past
the end of the buffer already) but we can at least try to get a warning
written into an on-disk log by avoiding the hard crash caused by padding
past the buffer with a huge number of zeros.
Signed-off-by: Jouni K. Seppänen <jks@iki.fi> Fixes: 4a0e3e989d66 ("cdc_ncm: Add support for moving NDP to end of NCM frame") Link: https://bugzilla.kernel.org/show_bug.cgi?id=209407 Reported-by: kernel test robot <lkp@intel.com> Reviewed-by: Bjørn Mork <bjorn@mork.no> Signed-off-by: David S. Miller <davem@davemloft.net>
[jks@iki.fi: backport to 4.4.y, 4.9.y] Signed-off-by: Jouni K. Seppänen <jks@iki.fi> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If you export a subdirectory of a filesystem, a READDIRPLUS on the root
of that export will return the filehandle of the parent with the ".."
entry.
The filehandle is optional, so let's just not return the filehandle for
".." if we're at the root of an export.
Note that once the client learns one filehandle outside of the export,
they can trivially access the rest of the export using further lookups.
However, it is also not very difficult to guess filehandles outside of
the export. So exporting a subdirectory of a filesystem should
considered equivalent to providing access to the entire filesystem. To
avoid confusion, we recommend only exporting entire filesystems.
Reported-by: Youjipeng <wangzhibei1999@gmail.com> Signed-off-by: J. Bruce Fields <bfields@redhat.com> Cc: stable@vger.kernel.org Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When updating the buffer demux, we will skip a scan element from the
device in the case `in_ind != out_ind` and we enter the while loop.
in_ind should only be refreshed with `find_next_bit()` in the end of the
loop.
Note, to cause problems we need a situation where we are skippig over
an element (channel not enabled) that happens to not have the same size
as the next element. Whilst this is a possible situation we haven't
actually identified any cases in mainline where it happens as most drivers
have consistent channel storage sizes with the exception of the timestamp
which is the last element and hence never skipped over.
Fixes: 5ada4ea9be16 ("staging:iio: add demux optionally to path from device to buffer") Signed-off-by: Nuno Sá <nuno.sa@analog.com> Link: https://lore.kernel.org/r/20201112144323.28887-1-nuno.sa@analog.com Cc: <Stable@vger.kernel.org> Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
[sudip: adjust context] Signed-off-by: Sudip Mukherjee <sudipm.mukherjee@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The 'distrust_firmware' module parameter dates from 2004 and the USB
subsystem is a lot more mature and reliable now than it was then.
Alter the default to false now.
acquire_slab() fails if there is contention on the freelist of the page
(probably because some other CPU is concurrently freeing an object from
the page). In that case, it might make sense to look for a different page
(since there might be more remote frees to the page from other CPUs, and
we don't want contention on struct page).
However, the current code accidentally stops looking at the partial list
completely in that case. Especially on kernels without CONFIG_NUMA set,
this means that get_partial() fails and new_slab_objects() falls back to
new_slab(), allocating new pages. This could lead to an unnecessary
increase in memory fragmentation.
Link: https://lkml.kernel.org/r/20201228130853.1871516-1-jannh@google.com Fixes: 7ced37197196 ("slub: Acquire_slab() avoid loop") Signed-off-by: Jann Horn <jannh@google.com> Acked-by: David Rientjes <rientjes@google.com> Acked-by: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Christoph Lameter <cl@linux.com> Cc: Pekka Enberg <penberg@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When setting password salt in the superblock, we forget to recompute the
superblock checksum so it will not match until the next superblock
modification which recomputes the checksum. Fix it.
CC: Michael Halcrow <mhalcrow@google.com> Reported-by: Andreas Dilger <adilger@dilger.ca> Fixes: 9bd8212f981e ("ext4 crypto: add encryption policy and password salt support") Signed-off-by: Jan Kara <jack@suse.cz> Link: https://lore.kernel.org/r/20201216101844.22917-8-jack@suse.cz Signed-off-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Before referencing the inode, we must ensure that the superblock can be
referenced. Otherwise, we can end up with iput() calling superblock
operations that are no longer valid or accessible.
Fixes: ea7c38fef0b7 ("NFSv4: Ensure we reference the inode for return-on-close in delegreturn") Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
We are not guaranteed the locking environment that would prevent
dentry getting renamed right under us. And it's possible for
old long name to be freed after rename, leading to UAF here.
Normally, when input device supporting force feedback effects is being
destroyed, we try to "flush" currently playing effects, so that the
physical device does not continue vibrating (or executing other effects).
Unfortunately this does not work well for uinput as flushing of the effects
deadlocks with the destroy action:
- if device is being destroyed because the file descriptor is being closed,
then there is noone to even service FF requests;
- if device is being destroyed because userspace sent UI_DEV_DESTROY,
while theoretically it could be possible to service FF requests,
userspace is unlikely to do so (they'd need to make sure FF handling
happens on a separate thread) even if kernel solves the issue with FF
ioctls deadlocking with UI_DEV_DESTROY ioctl on udev->mutex.
To avoid lockups like the one below, let's install a custom input device
flush handler, and avoid trying to flush force feedback effects when we
destroying the device, and instead rely on uinput to shut off the device
properly.
Since commit 1d6cd3929360 ("modpost: turn missing MODULE_LICENSE()
into error") the ppc32_allmodconfig build fails with:
ERROR: modpost: missing MODULE_LICENSE() in drivers/net/ethernet/freescale/fs_enet/mii-fec.o
ERROR: modpost: missing MODULE_LICENSE() in drivers/net/ethernet/freescale/fs_enet/mii-bitbang.o
Add the missing MODULE_LICENSEs to fix the build. Both files include a
copyright header indicating they are GPL v2.
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au> Reviewed-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
arm-linux-gnueabi-ld: drivers/isdn/mISDN/dsp_audio.o: in function `dsp_audio_generate_law_tables':
(.text+0x30c): undefined reference to `byte_rev_table'
arm-linux-gnueabi-ld: drivers/isdn/mISDN/dsp_audio.o:(.text+0x5e4): more undefined references to `byte_rev_table' follow
Signed-off-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
fs/dax.c uses copy_user_page() but ARC does not provide that interface,
resulting in a build error.
Provide copy_user_page() in <asm/page.h>.
../fs/dax.c: In function 'copy_cow_page_dax':
../fs/dax.c:702:2: error: implicit declaration of function 'copy_user_page'; did you mean 'copy_to_user_page'? [-Werror=implicit-function-declaration]
Reported-by: kernel test robot <lkp@intel.com> Signed-off-by: Randy Dunlap <rdunlap@infradead.org> Cc: Vineet Gupta <vgupta@synopsys.com> Cc: linux-snps-arc@lists.infradead.org Cc: Dan Williams <dan.j.williams@intel.com>
#Acked-by: Vineet Gupta <vgupta@synopsys.com> # v1 Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Matthew Wilcox <willy@infradead.org> Cc: Jan Kara <jack@suse.cz> Cc: linux-fsdevel@vger.kernel.org Cc: linux-nvdimm@lists.01.org
#Reviewed-by: Ira Weiny <ira.weiny@intel.com> # v2 Signed-off-by: Vineet Gupta <vgupta@synopsys.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Table 8-53 in the QUICC Engine Reference manual shows definitions of
fields up to a size of 192 bytes, not just 128. But in table 8-111,
one does find the text
Base Address of the Global Transmitter Parameter RAM Page. [...]
The user needs to allocate 128 bytes for this page. The address must
be aligned to the page size.
I've checked both rev. 7 (11/2015) and rev. 9 (05/2018) of the manual;
they both have this inconsistency (and the table numbers are the
same).
Adding a bit of debug printing, on my board the struct
ucc_geth_tx_global_pram is allocated at offset 0x880, while
the (opaque) ucc_geth_thread_data_tx gets allocated immediately
afterwards, at 0x900. So whatever the engine writes into the thread
data overlaps with the tail of the global tx pram (and devmem says
that something does get written during a simple ping).
I haven't observed any failure that could be attributed to this, but
it seems to be the kind of thing that would be extremely hard to
debug. So extend the struct definition so that we do allocate 192
bytes.
The top-level boot_targets (uImage and uImage.*) should be phony
targets. They just let Kbuild descend into arch/arc/boot/ and create
files there.
If a file exists in the top directory with the same name, the boot
image will not be created.
You can confirm it by the following steps:
$ export CROSS_COMPILE=<your-arc-compiler-prefix>
$ make -s ARCH=arc defconfig all # vmlinux will be built
$ touch uImage.gz
$ make ARCH=arc uImage.gz
CALL scripts/atomic/check-atomics.sh
CALL scripts/checksyscalls.sh
CHK include/generated/compile.h
# arch/arc/boot/uImage.gz is not created
We got a "deleted inode referenced" warning cross our fsstress test. The
bug can be reproduced easily with following steps:
cd /dev/shm
mkdir test/
fallocate -l 128M img
mkfs.ext4 -b 1024 img
mount img test/
dd if=/dev/zero of=test/foo bs=1M count=128
mkdir test/dir/ && cd test/dir/
for ((i=0;i<1000;i++)); do touch file$i; done # consume all block
cd ~ && renameat2(AT_FDCWD, /dev/shm/test/dir/file1, AT_FDCWD,
/dev/shm/test/dir/dst_file, RENAME_WHITEOUT) # ext4_add_entry in
ext4_rename will return ENOSPC!!
cd /dev/shm/ && umount test/ && mount img test/ && ls -li test/dir/file1
We will get the output:
"ls: cannot access 'test/dir/file1': Structure needs cleaning"
and the dmesg show:
"EXT4-fs error (device loop0): ext4_lookup:1626: inode #2049: comm ls:
deleted inode referenced: 139"
ext4_rename will create a special inode for whiteout and use this 'ino'
to replace the source file's dir entry 'ino'. Once error happens
latter(the error above was the ENOSPC return from ext4_add_entry in
ext4_rename since all space has been consumed), the cleanup do drop the
nlink for whiteout, but forget to restore 'ino' with source file. This
will trigger the bug describle as above.
Signed-off-by: yangerkun <yangerkun@huawei.com> Reviewed-by: Jan Kara <jack@suse.cz> Cc: stable@vger.kernel.org Fixes: cd808deced43 ("ext4: support RENAME_WHITEOUT") Link: https://lore.kernel.org/r/20210105062857.3566-1-yangerkun@huawei.com Signed-off-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Sasha Levin <sashal@kernel.org>
A widget's "dirty" list_head, much like its "list" list_head, eventually
chains back to a list_head on the snd_soc_card itself. This means that
the list can stick around even after the widget (or all widgets) have
been freed. Currently, however, widgets that are in the dirty list when
freed remain there, corrupting the entire list and leading to memory
errors and undefined behavior when the list is next accessed or
modified.
I encountered this issue when a component failed to probe relatively
late in snd_soc_bind_card(), causing it to bail out and call
soc_cleanup_card_resources(), which eventually called
snd_soc_dapm_free() with widgets that were still dirty from when they'd
been added.
Fixes: db432b414e20 ("ASoC: Do DAPM power checks only for widgets changed since last run") Cc: stable@vger.kernel.org Signed-off-by: Thomas Hebb <tommyhebb@gmail.com> Reviewed-by: Charles Keepax <ckeepax@opensource.cirrus.com> Link: https://lore.kernel.org/r/f8b5f031d50122bf1a9bfc9cae046badf4a7a31a.1607822410.git.tommyhebb@gmail.com Signed-off-by: Mark Brown <broonie@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
syzbot reproduces BUG_ON in skb_checksum_help():
tun creates (bogus) skb with huge partial-checksummed area and
small ip packet inside. Then ip_rcv trims the skb based on size
of internal ip packet, after that csum offset points beyond of
trimmed skb. Then checksum_tg() called via netfilter hook
triggers BUG_ON:
To work around the problem this patch forces pskb_trim_rcsum_slow()
to return -EINVAL in described scenario. It allows its callers to
drop such kind of packets.
When irq_domain_get_irq_data() or irqd_cfg() fails
at i == 0, data allocated by kzalloc() has not been
freed before returning, which leads to memleak.
Fixes: b106ee63abcc ("irq_remapping/vt-d: Enhance Intel IR driver to support hierarchical irqdomains") Signed-off-by: Dinghao Liu <dinghao.liu@zju.edu.cn> Acked-by: Lu Baolu <baolu.lu@linux.intel.com> Link: https://lore.kernel.org/r/20210105051837.32118-1-dinghao.liu@zju.edu.cn Signed-off-by: Will Deacon <will@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
arm-linux-gnueabi-ld: drivers/net/wireless/ath/wil6210/fw.o: in function `wil_fw_verify':
fw.c:(.text+0x74c): undefined reference to `crc32_le'
arm-linux-gnueabi-ld: drivers/net/wireless/ath/wil6210/fw.o:fw.c:(.text+0x758): more undefined references to `crc32_le' follow
Fixes: 151a9706503f ("wil6210: firmware download") Signed-off-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Currently there is an unlikely case where cpufreq_cpu_get() returns a
NULL policy and this will cause a NULL pointer dereference later on.
Fix this by passing the policy to transition_frequency_fidvid() from
the caller and hence eliminating the need for the cpufreq_cpu_get()
and cpufreq_cpu_put().
Thanks to Viresh Kumar for suggesting the fix.
Addresses-Coverity: ("Dereference null return") Fixes: b43a7ffbf33b ("cpufreq: Notify all policy->cpus in cpufreq_notify_transition()") Suggested-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Colin Ian King <colin.king@canonical.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
pxa2xx_spi_remove() accesses the driver's private data after calling
spi_unregister_controller() even though that function releases the last
reference on the spi_controller and thereby frees the private data.
Fix by switching over to the new devm_spi_alloc_master/slave() helper
which keeps the private data accessible until the driver has unbound.
Write buffers use a kmalloc()'ed buffer, they can leak
up to seven bytes of kernel memory to flash if writes are not
aligned.
So use ubifs_pad() to fill these gaps with padding bytes.
This was never a problem while scanning because the scanner logic
manually aligns node lengths and skips over these gaps.
Basically, consider .text.{hot|unlikely|unknown}.* part of .text, too.
When compiling with profiling information (collected via PGO
instrumentations or AutoFDO sampling), Clang will separate code into
.text.hot, .text.unlikely, or .text.unknown sections based on profiling
information. After D79600 (clang-11), these sections will have a
trailing `.` suffix, ie. .text.hot., .text.unlikely., .text.unknown..
When using -ffunction-sections together with profiling infomation,
either explicitly (FGKASLR) or implicitly (LTO), code may be placed in
sections following the convention:
.text.hot.<foo>, .text.unlikely.<bar>, .text.unknown.<baz>
where <foo>, <bar>, and <baz> are functions. (This produces one section
per function; we generally try to merge these all back via linker script
so that we don't have 50k sections).
For the above cases, we need to teach our linker scripts that such
sections might exist and that we'd explicitly like them grouped
together, otherwise we can wind up with code outside of the
_stext/_etext boundaries that might not be mapped properly for some
architectures, resulting in boot failures.
If the linker script is not told about possible input sections, then
where the section is placed as output is a heuristic-laiden mess that's
non-portable between linkers (ie. BFD and LLD), and has resulted in many
hard to debug bugs. Kees Cook is working on cleaning this up by adding
--orphan-handling=warn linker flag used in ARCH=powerpc to additional
architectures. In the case of linker scripts, borrowing from the Zen of
Python: explicit is better than implicit.
Also, ld.bfd's internal linker script considers .text.hot AND
.text.hot.* to be part of .text, as well as .text.unlikely and
.text.unlikely.*. I didn't see support for .text.unknown.*, and didn't
see Clang producing such code in our kernel builds, but I see code in
LLVM that can produce such section names if profiling information is
missing. That may point to a larger issue with generating or collecting
profiles, but I would much rather be safe and explicit than have to
debug yet another issue related to orphan section placement.
For some reason ip_tunnel insist on setting the DF bit anyway when the
inner header has the DF bit set, EVEN if the tunnel was configured with
'nopmtudisc'.
This means that the script added in the previous commit
cannot be made to work by adding the 'nopmtudisc' flag to the
ip tunnel configuration. Doing so breaks connectivity even for the
without-conntrack/netfilter scenario.
When nopmtudisc is set, the tunnel will skip the mtu check, so no
icmp error is sent to client. Then, because inner header has DF set,
the outer header gets added with DF bit set as well.
IP stack then sends an error to itself because the packet exceeds
the device MTU.
Fixes: 23a3647bc4f93 ("ip_tunnels: Use skb-len to PMTU check.") Cc: Stefano Brivio <sbrivio@redhat.com> Signed-off-by: Florian Westphal <fw@strlen.de> Acked-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Conntrack reassembly records the largest fragment size seen in IPCB.
However, when this gets forwarded/transmitted, fragmentation will only
be forced if one of the fragmented packets had the DF bit set.
In that case, a flag in IPCB will force fragmentation even if the
MTU is large enough.
This should work fine, but this breaks with ip tunnels.
Consider client that sends a UDP datagram of size X to another host.
The client fragments the datagram, so two packets, of size y and z, are
sent. DF bit is not set on any of these packets.
Middlebox netfilter reassembles those packets back to single size-X
packet, before routing decision.
packet-size-vs-mtu checks in ip_forward are irrelevant, because DF bit
isn't set. At output time, ip refragmentation is skipped as well
because x is still smaller than the mtu of the output device.
If ttransmit device is an ip tunnel, the packet size increases to
x+overhead.
Also, tunnel might be configured to force DF bit on outer header.
In this case, packet will be dropped (exceeds MTU) and an ICMP error is
generated back to sender.
But sender already respects the announced MTU, all the packets that
it sent did fit the announced mtu.
Force refragmentation as per original sizes unconditionally so ip tunnel
will encapsulate the fragments instead.
The only other solution I see is to place ip refragmentation in
the ip_tunnel code to handle this case.
Fixes: d6b915e29f4ad ("ip_fragment: don't forward defragmented DF packet") Reported-by: Christian Perle <christian.perle@secunet.com> Signed-off-by: Florian Westphal <fw@strlen.de> Acked-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The placeholder for instruction selection should use the second
argument's operand, which is %1, not %0. This could generate incorrect
assembly code if the memory addressing of operand %0 is a different
form from that of operand %1.
Also remove the %Un placeholder because having %Un placeholders
for two operands which are based on the same local var (ptep) doesn't
make much sense. By the way, it doesn't change the current behaviour
because "<>" constraint is missing for the associated "=m".
[chleroy: revised commit log iaw segher's comments and removed %U0]
When attempting to match EXTENDED COPY CSCD descriptors with corresponding
se_devices, target_xcopy_locate_se_dev_e4() currently iterates over LIO's
global devices list which includes all configured backstores.
This change ensures that only initiator-accessible backstores are
considered during CSCD descriptor lookup, according to the session's
se_node_acl LUN list.
To avoid LUN removal race conditions, device pinning is changed from being
configfs based to instead using the se_node_acl lun_ref.
Reference: CVE-2020-28374 Fixes: cbf031f425fd ("target: Add support for EXTENDED_COPY copy offload emulation") Reviewed-by: Lee Duncan <lduncan@suse.com> Signed-off-by: David Disseldorp <ddiss@suse.de> Signed-off-by: Mike Christie <michael.christie@oracle.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
This converts the xcopy code to use the idr helper. The next patch
will drop the g_device_list and make g_device_mutex local to the
target_core_device.c file.
Signed-off-by: Mike Christie <mchristi@redhat.com> Reviewed-by: Bart Van Assche <bart.vanassche@wdc.com> Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The XCOPY specification in SPC4r37 states that the XCOPY source and
destination device(s) should be derived from the copy source and copy
destination (CSCD) descriptor IDs in the XCOPY segment descriptor.
The CSCD IDs are generally (for block -> block copies), indexes into
the corresponding CSCD descriptor list, e.g.
=================================
EXTENDED COPY Header
=================================
CSCD Descriptor List
- entry 0
+ LU ID <--------------<------------------\
- entry 1 |
+ LU ID <______________<_____________ |
================================= | |
Segment Descriptor List | |
- segment 0 | |
+ src CSCD ID = 0 --------->---------+----/
+ dest CSCD ID = 1 ___________>______|
+ len
+ src lba
+ dest lba
=================================
Currently LIO completely ignores the src and dest CSCD IDs in the
Segment Descriptor List, and instead assumes that the first entry in the
CSCD list corresponds to the source, and the second to the destination.
This commit removes this assumption, by ensuring that the Segment
Descriptor List is parsed prior to processing the CSCD Descriptor List.
CSCD Descriptor List processing is modified to compare the current list
index with the previously obtained src and dest CSCD IDs.
Additionally, XCOPY requests where the src and dest CSCD IDs refer to
the CSCD Descriptor List entry can now be successfully processed.
Fixes: cbf031f ("target: Add support for EXTENDED_COPY copy offload") Link: https://bugzilla.kernel.org/show_bug.cgi?id=191381 Signed-off-by: David Disseldorp <ddiss@suse.de> Reviewed-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
target_xcopy_locate_se_dev_e4() is used to locate an se_dev, based on
the WWN provided with the XCOPY request. Remove a couple of unneeded
arguments, and rely on the caller for the src/dst test.
Signed-off-by: David Disseldorp <ddiss@suse.de> Reviewed-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Check the length of the XCOPY request segment descriptor list against
the value advertised via the MAXIMUM SEGMENT DESCRIPTOR COUNT field in
the RECEIVE COPY OPERATING PARAMETERS response.
spc4r37 6.4.3.5 states:
If the number of segment descriptors exceeds the allowed number, the
copy manager shall terminate the command with CHECK CONDITION status,
with the sense key set to ILLEGAL REQUEST, and the additional sense
code set to TOO MANY SEGMENT DESCRIPTORS.
This functionality is testable using the libiscsi
ExtendedCopy.DescrLimits test.
Signed-off-by: David Disseldorp <ddiss@suse.de> Reviewed-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
In mtrr_type_lookup(), if the input memory address region is not in the
MTRR, over 4GB, and not over the top of memory, a write-back attribute
is returned. These condition checks are for ensuring the input memory
address region is actually mapped to the physical memory.
However, if the end address is just aligned with the top of memory,
the condition check treats the address is over the top of memory, and
write-back attribute is not returned.
And this hits in a real use case with NVDIMM: the nd_pmem module tries
to map NVDIMMs as cacheable memories when NVDIMMs are connected. If a
NVDIMM is the last of the DIMMs, the performance of this NVDIMM becomes
very low since it is aligned with the top of memory and its memory type
is uncached-minus.
Move the input end address change to inclusive up into
mtrr_type_lookup(), before checking for the top of memory in either
mtrr_type_lookup_{variable,fixed}() helpers.
htable_bits() can call jhash_size(32) and trigger shift-out-of-bounds
UBSAN: shift-out-of-bounds in net/netfilter/ipset/ip_set_hash_gen.h:151:6
shift exponent 32 is too large for 32-bit type 'unsigned int'
CPU: 0 PID: 8498 Comm: syz-executor519
Not tainted 5.10.0-rc7-next-20201208-syzkaller #0
Call Trace:
__dump_stack lib/dump_stack.c:79 [inline]
dump_stack+0x107/0x163 lib/dump_stack.c:120
ubsan_epilogue+0xb/0x5a lib/ubsan.c:148
__ubsan_handle_shift_out_of_bounds.cold+0xb1/0x181 lib/ubsan.c:395
htable_bits net/netfilter/ipset/ip_set_hash_gen.h:151 [inline]
hash_mac_create.cold+0x58/0x9b net/netfilter/ipset/ip_set_hash_gen.h:1524
ip_set_create+0x610/0x1380 net/netfilter/ipset/ip_set_core.c:1115
nfnetlink_rcv_msg+0xecc/0x1180 net/netfilter/nfnetlink.c:252
netlink_rcv_skb+0x153/0x420 net/netlink/af_netlink.c:2494
nfnetlink_rcv+0x1ac/0x420 net/netfilter/nfnetlink.c:600
netlink_unicast_kernel net/netlink/af_netlink.c:1304 [inline]
netlink_unicast+0x533/0x7d0 net/netlink/af_netlink.c:1330
netlink_sendmsg+0x907/0xe40 net/netlink/af_netlink.c:1919
sock_sendmsg_nosec net/socket.c:652 [inline]
sock_sendmsg+0xcf/0x120 net/socket.c:672
____sys_sendmsg+0x6e8/0x810 net/socket.c:2345
___sys_sendmsg+0xf3/0x170 net/socket.c:2399
__sys_sendmsg+0xe5/0x1b0 net/socket.c:2432
do_syscall_64+0x2d/0x70 arch/x86/entry/common.c:46
entry_SYSCALL_64_after_hwframe+0x44/0xa9
This patch replaces htable_bits() by simple fls(hashsize - 1) call:
it alone returns valid nbits both for round and non-round hashsizes.
It is normal to set any nbits here because it is validated inside
following htable_size() call which returns 0 for nbits>31.
While commit d5dcce0c414f ("device property: Keep secondary firmware
node secondary by type") describes everything correct in its commit
message, the change it made does the opposite and original commit c15e1bdda436 ("device property: Fix the secondary firmware node handling
in set_primary_fwnode()") was fully correct.
Revert the former one here and improve documentation in the next patch.
Fixes: d5dcce0c414f ("device property: Keep secondary firmware node secondary by type") Signed-off-by: Bard Liao <yung-chuan.liao@linux.intel.com> Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Reviewed-by: Heikki Krogerus <heikki.krogerus@linux.intel.com> Cc: 5.10+ <stable@vger.kernel.org> # 5.10+ Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
introduced a new location where a pmd was released, but neglected to
run the pmd page destructor. In fact, this happened previously for a
different pmd release path and was fixed by commit:
c283610e44ec ("x86, mm: do not leak page->ptl for pmd page tables").
This issue was hidden until recently because the failure mode is silent,
but commit:
b2b29d6d0119 ("mm: account PMD tables like PTE tables")
Given this is a repeat occurrence it seemed prudent to look for other
places where this destructor might be missing and whether a better
helper is needed. try_to_free_pmd_page() looks like a candidate, but
testing with setting up and tearing down pmd mappings via the dax unit
tests is thus far not triggering the failure.
As for a better helper pmd_free() is close, but it is a messy fit
due to requiring an @mm arg. Also, ___pmd_free_tlb() wants to call
paravirt_tlb_remove_table() instead of free_page(), so open-coded
pgtable_pmd_page_dtor() seems the best way forward for now.
Debugged together with Matthew Wilcox <willy@infradead.org>.
Fixes: 28ee90fe6048 ("x86/mm: implement free pmd/pte page interfaces") Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Borislav Petkov <bp@suse.de> Tested-by: Yi Zhang <yi.zhang@redhat.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: <stable@vger.kernel.org> Link: https://lkml.kernel.org/r/160697689204.605323.17629854984697045602.stgit@dwillia2-desk3.amr.corp.intel.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Johan Hovold [Fri, 8 Jan 2021 14:55:28 +0000 (15:55 +0100)]
USB: serial: keyspan_pda: remove unused variable
Remove an unused variable which was mistakingly left by commit 37faf5061541 ("USB: serial: keyspan_pda: fix write-wakeup
use-after-free") and only removed by a later change.
This is needed to suppress a W=1 warning about the unused variable in
the stable trees that the build bots triggers.
Reported-by: kernel test robot <lkp@intel.com> Signed-off-by: Johan Hovold <johan@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When binding the ConfigFS gadget to a UDC, the functions in each
configuration are added in list order. However, if usb_add_function()
fails, the failed function is put back on its configuration's
func_list and purge_configs_funcs() is called to further clean up.
purge_configs_funcs() iterates over the configurations and functions
in forward order, calling unbind() on each of the previously added
functions. But after doing so, each function gets moved to the
tail of the configuration's func_list. This results in reshuffling
the original order of the functions within a configuration such
that the failed function now appears first even though it may have
originally appeared in the middle or even end of the list. At this
point if the ConfigFS gadget is attempted to re-bind to the UDC,
the functions will be added in a different order than intended,
with the only recourse being to remove and relink the functions all
over again.
[30133.118289] configfs-gadget gadget: adding 'Mass Storage Function'/ffffff810af87200 to config 'c'/ffffff817d6a2520
[30133.119875] configfs-gadget gadget: adding 'cdc_network'/ffffff80f48d1a00 to config 'c'/ffffff817d6a2520
[30133.119974] using random self ethernet address
[30133.120002] using random host ethernet address
[30133.139604] usb0: HOST MAC 3e:27:46:ba:3e:26
[30133.140015] usb0: MAC 6e:28:7e:42:66:6a
[30133.140062] configfs-gadget gadget: adding 'Function FS Gadget'/ffffff80f3868438 to config 'c'/ffffff817d6a2520
[30133.140081] configfs-gadget gadget: adding 'Function FS Gadget'/ffffff80f3868438 --> -19
[30133.140098] configfs-gadget gadget: unbind function 'Mass Storage Function'/ffffff810af87200
[30133.140119] configfs-gadget gadget: unbind function 'cdc_network'/ffffff80f48d1a00
[30133.173201] configfs-gadget a600000.dwc3: failed to start g1: -19
[30136.661933] init: starting service 'adbd'...
[30136.700126] read descriptors
[30136.700413] read strings
[30138.574484] configfs-gadget gadget: adding 'Function FS Gadget'/ffffff80f3868438 to config 'c'/ffffff817d6a2520
[30138.575497] configfs-gadget gadget: adding 'Mass Storage Function'/ffffff810af87200 to config 'c'/ffffff817d6a2520
[30138.575554] configfs-gadget gadget: adding 'cdc_network'/ffffff80f48d1a00 to config 'c'/ffffff817d6a2520
[30138.575631] using random self ethernet address
[30138.575660] using random host ethernet address
[30138.595338] usb0: HOST MAC 2e:cf:43:cd:ca:c8
[30138.597160] usb0: MAC 6a:f0:9f:ee:82:a0
[30138.791490] configfs-gadget gadget: super-speed config #1: c
Fix this by reversing the iteration order of the functions in
purge_config_funcs() when unbinding them, and adding them back to
the config's func_list at the head instead of the tail. This
ensures that we unbind and unwind back to the original list order.
Fixes: 88af8bbe4ef7 ("usb: gadget: the start of the configfs interface") Signed-off-by: Chandana Kishori Chiluveru <cchiluve@codeaurora.org> Signed-off-by: Jack Pham <jackp@codeaurora.org> Reviewed-by: Peter Chen <peter.chen@nxp.com> Link: https://lore.kernel.org/r/20201229224443.31623-1-jackp@codeaurora.org Cc: stable <stable@vger.kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
There is a spinlock lockup as part of composite_disconnect
when it tries to acquire cdev->lock as part of usb_gadget_deactivate.
This is because the usb_gadget_deactivate is called from
usb_function_deactivate with the same spinlock held.
This would result in the below call stack and leads to stall.
When printer driver is loaded, the printer_func_bind function is called, in
this function, the interface descriptor be allocated memory, if after that,
the error occurred, the interface descriptor memory need to be free.
With commit 913e4a90b6f9 ("usb: gadget: f_uac2: finalize wMaxPacketSize according to bandwidth")
wMaxPacketSize is computed dynamically but the value is never reset.
Because of this, the actual maximum packet size can only decrease each time
the audio gadget is instantiated.
Reset the endpoint maximum packet size and mark wMaxPacketSize as dynamic
to solve the problem.
arm-linux-gnueabi-ld: drivers/usb/gadget/function/f_eem.o: in function `eem_unwrap':
f_eem.c:(.text+0x11cc): undefined reference to `crc32_le'
arm-linux-gnueabi-ld: drivers/usb/gadget/function/f_ncm.o:f_ncm.c:(.text+0x1e40):
more undefined references to `crc32_le' follow
The calculation of in_cables and out_cables bitmaps are done with the
bit shift by the value from the descriptor, which is an arbitrary
value, and can lead to UBSAN shift-out-of-bounds warnings.
Fix it by filtering the bad descriptor values with the check of the
upper bound 0x10 (the cable bitmaps are 16 bits).
Stack-allocated buffers cannot be used for DMA (on all architectures).
Replace the HP-channel macro with a helper function that allocates a
dedicated transfer buffer so that it can continue to be used with
arguments from the stack.
Note that the buffer is cleared on allocation as usblp_ctrl_msg()
returns success also on short transfers (the buffer is only used for
debugging).
Make sure to always cancel the control URB in write() so that it can be
reused after a timeout or spurious CMD_ACK.
Currently any further write requests after a timeout would fail after
triggering a WARN() in usb_submit_urb() when attempting to submit the
already active URB.
Signed-off-by: Daniel Palmer <daniel@0x0f.com>
https://lore.kernel.org/r/20201227031716.1343300-1-daniel@0x0f.com
[ johan: drop id defines, only bind to vendor class ] Cc: stable@vger.kernel.org Signed-off-by: Johan Hovold <johan@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Here's another variant PNY Pro Elite USB 3.1 Gen 2 portable SSD that
hangs and doesn't respond to ATA_1x pass-through commands. If it doesn't
support these commands, it should respond properly to the host. Add it
to the unusual uas list to be able to move forward with other
operations.
The commit 0472bf06c6fd ("xhci: Prevent U1/U2 link pm states if exit
latency is too long") was constraining the xhci code not to allow U1/U2
sleep states if the latency to wake up from the U-states reached the
service interval of an periodic endpoint. This fix was not taking into
account that in case the quirk XHCI_INTEL_HOST is set, the wakeup time
will be calculated and configured differently.
It checks for u1_params.mel/u2_params.mel as a limit. But the code could
decide to write another MEL into the hardware. This leads to broken
cases where not enough bandwidth is available for other devices:
usb 1-2: can't set config #1, error -28
This patch is fixing that case by checking for timeout_ns after the
wakeup time was calculated depending on the quirks.
Fixes: 0472bf06c6fd ("xhci: Prevent U1/U2 link pm states if exit latency is too long") Signed-off-by: Michael Grzeschik <m.grzeschik@pengutronix.de> Cc: stable <stable@vger.kernel.org> Link: https://lore.kernel.org/r/20201215193147.11738-1-m.grzeschik@pengutronix.de Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
if of_find_device_by_node() succeed, usbmisc_get_init_data() doesn't have
a corresponding put_device(). Thus add put_device() to fix the exception
handling for this function implementation.
Enable Super speed plus in configfs to support USB3.1 Gen2.
This ensures that when a USB gadget is plugged in, it is
enumerated as Gen 2 and connected at 10 Gbps if the host and
cable are capable of it.
Many in-tree gadget functions (fs, midi, acm, ncm, mass_storage,
etc.) already have SuperSpeed Plus support.
Tested: plugged gadget into Linux host and saw:
[284907.385986] usb 8-2: new SuperSpeedPlus Gen 2 USB device number 3 using xhci_hcd
Tested-by: Lorenzo Colitti <lorenzo@google.com> Acked-by: Felipe Balbi <balbi@kernel.org> Signed-off-by: taehyun.cho <taehyun.cho@samsung.com> Signed-off-by: Lorenzo Colitti <lorenzo@google.com> Link: https://lore.kernel.org/r/20210106154625.2801030-1-lorenzo@google.com Cc: stable <stable@vger.kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Dexuan Cui [Sat, 9 Jan 2021 22:53:58 +0000 (14:53 -0800)]
video: hyperv_fb: Fix the mmap() regression for v5.4.y and older
db49200b1dad is backported from the mainline commit 5f1251a48c17 ("video: hyperv_fb: Fix the cache type when mapping the VRAM"),
to v5.4.y and older stable branches, but unluckily db49200b1dad causes
mmap() to fail for /dev/fb0 due to EINVAL:
[ 5797.049560] x86/PAT: a.out:1910 map pfn expected mapping type
uncached-minus for [mem 0xf8200000-0xf85cbfff], got write-back
This means the v5.4.y kernel detects an incompatibility issue about the
mapping type of the VRAM: db49200b1dad changes to use Write-Back when
mapping the VRAM, while the mmap() syscall tries to use Uncached-minus.
That’s to say, the kernel thinks Uncached-minus is incompatible with
Write-Back: see drivers/video/fbdev/core/fbmem.c: fb_mmap() ->
vm_iomap_memory() -> io_remap_pfn_range() -> ... -> track_pfn_remap() ->
reserve_pfn_range().
Note: any v5.5 and newer kernel doesn't have the issue, because they
have commit d21987d709e8 ("video: hyperv: hyperv_fb: Support deferred IO for Hyper-V frame buffer driver")
, and when the hyperv_fb driver has the deferred_io support,
fb_deferred_io_init() overrides info->fbops->fb_mmap with
fb_deferred_io_mmap(), which doesn’t check the mapping type
incompatibility. Note: since it's VRAM here, the checking is not really
necessary.
Fix the regression by ioremap_wc(), which uses Write-combining. The kernel
thinks it's compatible with Uncached-minus. The VRAM mappped by
ioremap_wc() is slightly slower than mapped by ioremap_cache(), but is
still significantly faster than by ioremap().
Change the comment accordingly. Linux VM on ARM64 Hyper-V is still not
working in the latest mainline yet, and when it works in future, the ARM64
support is unlikely to be backported to v5.4 and older, so using
ioremap_wc() in v5.4 and older should be ok.
Note: this fix is only targeted at the stable branches:
v5.4.y, v4.19.y, v4.14.y, v4.9.y and v4.4.y.
Fixes: db49200b1dad ("video: hyperv_fb: Fix the cache type when mapping the VRAM") Signed-off-by: Dexuan Cui <decui@microsoft.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
ugeth is the netdiv_priv() part of the netdevice. Accessing the memory
pointed to by ugeth (such as done by ucc_geth_memclean() and the two
of_node_puts) after free_netdev() is thus use-after-free.
virtnet_set_channels can recursively call cpus_read_lock if CONFIG_XPS
and CONFIG_HOTPLUG are enabled.
The path is:
virtnet_set_channels - calls get_online_cpus(), which is a trivial
wrapper around cpus_read_lock()
netif_set_real_num_tx_queues
netif_reset_xps_queues_gt
netif_reset_xps_queues - calls cpus_read_lock()
This call chain and potential deadlock happens when the number of TX
queues is reduced.
This commit the removes netif_set_real_num_[tr]x_queues calls from
inside the get/put_online_cpus section, as they don't require that it
be held.
Fixes: 47be24796c13 ("virtio-net: fix the set affinity bug when CPU IDs are not consecutive") Signed-off-by: Jeff Dike <jdike@akamai.com> Acked-by: Jason Wang <jasowang@redhat.com> Acked-by: Michael S. Tsirkin <mst@redhat.com> Link: https://lore.kernel.org/r/20201223025421.671-1-jdike@akamai.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Check Scell_log shift size in red_check_params() and modify all callers
of red_check_params() to pass Scell_log.
This prevents a shift out-of-bounds as detected by UBSAN:
UBSAN: shift-out-of-bounds in ./include/net/red.h:252:22
shift exponent 72 is too large for 32-bit type 'int'
Fixes: 8afa10cbe281 ("net_sched: red: Avoid illegal values") Signed-off-by: Randy Dunlap <rdunlap@infradead.org> Reported-by: syzbot+97c5bd9cc81eca63d36e@syzkaller.appspotmail.com Cc: Nogah Frankel <nogahf@mellanox.com> Cc: Jamal Hadi Salim <jhs@mojatatu.com> Cc: Cong Wang <xiyou.wangcong@gmail.com> Cc: Jiri Pirko <jiri@resnulli.us> Cc: netdev@vger.kernel.org Cc: "David S. Miller" <davem@davemloft.net> Cc: Jakub Kicinski <kuba@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Currently the vhost_zerocopy_callback() maybe be called to decrease
the refcount when sendmsg fails in tun. The error handling in vhost
handle_tx_zerocopy() will try to decrease the same refcount again.
This is wrong. To fix this issue, we only call vhost_net_ubuf_put()
when vq->heads[nvq->desc].len == VHOST_DMA_IN_PROGRESS.
Fixes: bab632d69ee4 ("vhost: vhost TX zero-copy support") Signed-off-by: Yunjian Wang <wangyunjian@huawei.com> Acked-by: Willem de Bruijn <willemb@google.com> Acked-by: Michael S. Tsirkin <mst@redhat.com> Acked-by: Jason Wang <jasowang@redhat.com> Link: https://lore.kernel.org/r/1609207308-20544-1-git-send-email-wangyunjian@huawei.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The cdc_ncm driver passes network connection notifications up to
usbnet_link_change(), which is the right place for any logging.
Remove the netdev_info() duplicating this from the driver itself.
This stops devices such as my "TRENDnet USB 10/100/1G/2.5G LAN"
(ID 20f4:e02b) adapter from spamming the kernel log with
ppp_cp_event is called directly or indirectly by ppp_rx with "ppp->lock"
held. It may call mod_timer to add a new timer. However, at the same time
ppp_timer may be already running and waiting for "ppp->lock". In this
case, there's no need for ppp_timer to continue running and it can just
exit.
If we let ppp_timer continue running, it may call add_timer. This causes
kernel panic because add_timer can't be called with a timer pending.
This patch fixes this problem.
Fixes: e022c2f07ae5 ("WAN: new synchronous PPP implementation for generic HDLC.") Cc: Krzysztof Halasa <khc@pm.waw.pl> Signed-off-by: Xie He <xie.he.0141@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
RT_TOS() only clears one of the ECN bits. Therefore, when
fib_compute_spec_dst() resorts to a fib lookup, it can return
different results depending on the value of the second ECN bit.
For example, ECT(0) and ECT(1) packets could be treated differently.
$ ip netns add ns0
$ ip netns add ns1
$ ip link add name veth01 netns ns0 type veth peer name veth10 netns ns1
$ ip -netns ns0 link set dev lo up
$ ip -netns ns1 link set dev lo up
$ ip -netns ns0 link set dev veth01 up
$ ip -netns ns1 link set dev veth10 up
$ ip -netns ns0 address add 192.0.2.10/24 dev veth01
$ ip -netns ns1 address add 192.0.2.11/24 dev veth10
$ ip -netns ns1 address add 192.0.2.21/32 dev lo
$ ip -netns ns1 route add 192.0.2.10/32 tos 4 dev veth10 src 192.0.2.21
$ ip netns exec ns1 sysctl -wq net.ipv4.icmp_echo_ignore_broadcasts=0
With TOS 4 and ECT(1), ns1 replies using source address 192.0.2.21
(ping uses -Q to set all TOS and ECN bits):
$ ip netns exec ns0 ping -c 1 -b -Q 5 192.0.2.255
[...]
64 bytes from 192.0.2.21: icmp_seq=1 ttl=64 time=0.544 ms
But with TOS 4 and ECT(0), ns1 replies using source address 192.0.2.11
because the "tos 4" route isn't matched:
$ ip netns exec ns0 ping -c 1 -b -Q 6 192.0.2.255
[...]
64 bytes from 192.0.2.11: icmp_seq=1 ttl=64 time=0.597 ms
After this patch the ECN bits don't affect the result anymore:
$ ip netns exec ns0 ping -c 1 -b -Q 6 192.0.2.255
[...]
64 bytes from 192.0.2.21: icmp_seq=1 ttl=64 time=0.591 ms
Fixes: 35ebf65e851c ("ipv4: Create and use fib_compute_spec_dst() helper.") Signed-off-by: Guillaume Nault <gnault@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Commit 436e980e2ed5 ("kbuild: don't hardcode depmod path") stopped
hard-coding the path of depmod, but in the process caused trouble for
distributions that had that /sbin location, but didn't have it in the
PATH (generally because /sbin is limited to the super-user path).
Work around it for now by just adding /sbin to the end of PATH in the
depmod.sh script.
In realtime scenario, We do not want to have interference on the
isolated cpu cores. but when invoking alloc_workqueue() for percpu wq
on the housekeeping cpu, it kick a kworker on the isolated cpu.
The comment in pwq_adjust_max_active() said:
"Need to kick a worker after thawed or an unbound wq's
max_active is bumped"
So it is unnecessary to kick a kworker for percpu's wq when invoking
alloc_workqueue(). this patch only kick a worker based on the actual
activation of delayed works.
Signed-off-by: Yunfeng Ye <yeyunfeng@huawei.com> Reviewed-by: Lai Jiangshan <jiangshanlai@gmail.com> Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
mwifiex_cmd_802_11_ad_hoc_start() calls memcpy() without checking
the destination size may trigger a buffer overflower,
which a local user could use to cause denial of service
or the execution of arbitrary code.
Fix it by putting the length check before calling memcpy().
One of a class of bugs pointed out by Lars in a recent review.
iio_push_to_buffers_with_timestamp() assumes the buffer used is aligned
to the size of the timestamp (8 bytes). This is not guaranteed in
this driver which uses an array of smaller elements on the stack.
As Lars also noted this anti pattern can involve a leak of data to
userspace and that indeed can happen here. We close both issues by
moving to a suitable structure in the iio_priv() data.
This data is allocated with kzalloc() so no data can leak apart from
previous readings.
The explicit alignment of ts is not necessary in this case but
does make the code slightly less fragile so I have included it.
Apparently there has been a longstanding race between udev/systemd and
the module loader. Currently, the module loader sends a uevent right
after sysfs initialization, but before the module calls its init
function. However, some udev rules expect that the module has
initialized already upon receiving the uevent.
This race has been triggered recently (see link in references) in some
systemd mount unit files. For instance, the configfs module creates the
/sys/kernel/config mount point in its init function, however the module
loader issues the uevent before this happens. sys-kernel-config.mount
expects to be able to mount /sys/kernel/config upon receipt of the
module loading uevent, but if the configfs module has not called its
init function yet, then this directory will not exist and the mount unit
fails. A similar situation exists for sys-fs-fuse-connections.mount, as
the fuse sysfs mount point is created during the fuse module's init
function. If udev is faster than module initialization then the mount
unit would fail in a similar fashion.
To fix this race, delay the module KOBJ_ADD uevent until after the
module has finished calling its init routine.
I noticed that iounmap() of msgr_block_addr before return from
mpic_msgr_probe() in the error handling case is missing. So use
devm_ioremap() instead of just ioremap() when remapping the message
register block, so the mapping will be automatically released on
probe failure.
The on-disk quota format supports quota files with upto 2^32 blocks. Be
careful when computing quota file offsets in the quota files from block
numbers as they can overflow 32-bit types. Since quota files larger than
4GB would require ~26 millions of quota users, this is mostly a
theoretical concern now but better be careful, fuzzers would find the
problem sooner or later anyway...
Reviewed-by: Andreas Dilger <adilger@dilger.ca> Signed-off-by: Jan Kara <jack@suse.cz> Signed-off-by: Sasha Levin <sashal@kernel.org>
If a module fails to load due to an error in prepare_coming_module(),
the following error handling in load_module() runs with
MODULE_STATE_COMING in module's state. Fix it by correctly setting
MODULE_STATE_GOING under "bug_cleanup" label.
The snd_seq_queue struct contains various flags in the bit fields.
Those are categorized to two different use cases, both of which are
protected by different spinlocks. That implies that there are still
potential risks of the bad operations for bit fields by concurrent
accesses.
For addressing the problem, this patch rearranges those flags to be
a standard bool instead of a bit field.
As reported on:
https://lore.kernel.org/linux-media/20190627222020.45909-1-willemdebruijn.kernel@gmail.com/
if gp8psk_usb_in_op() returns an error, the status var is not
initialized. Yet, this var is used later on, in order to
identify:
- if the device was already started;
- if firmware has loaded;
- if the LNBf was powered on.
Using status = 0 seems to ensure that everything will be
properly powered up.
So, instead of the proposed solution, let's just set
status = 0.
Reported-by: syzbot <syzkaller@googlegroups.com> Reported-by: Willem de Bruijn <willemb@google.com> Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
A kernel-infoleak was reported by syzbot, which was caused because
dbells was left uninitialized.
Using kzalloc() instead of kmalloc() fixes this issue.
Specify type alignment when declaring linker-section match-table entries
to prevent gcc from increasing alignment and corrupting the various
tables with padding (e.g. timers, irqchips, clocks, reserved memory).
This is specifically needed on x86 where gcc (typically) aligns larger
objects like struct of_device_id with static extent on 32-byte
boundaries which at best prevents matching on anything but the first
entry. Specifying alignment when declaring variables suppresses this
optimisation.
Here's a 64-bit example where all entries are corrupt as 16 bytes of
padding has been inserted before the first entry:
And here's a 32-bit example where the 8-byte-aligned table happens to be
placed on a 32-byte boundary so that all but the first entry are corrupt
due to the 28 bytes of padding inserted between entries:
and include <linux/const.h> in UAPI headers instead of <linux/kernel.h>.
The reason is to avoid indirect <linux/sysinfo.h> include when using
some network headers: <linux/netlink.h> or others -> <linux/kernel.h>
-> <linux/sysinfo.h>.
This indirect include causes on MUSL redefinition of struct sysinfo when
included both <sys/sysinfo.h> and some of UAPI headers:
In file included from x86_64-buildroot-linux-musl/sysroot/usr/include/linux/kernel.h:5,
from x86_64-buildroot-linux-musl/sysroot/usr/include/linux/netlink.h:5,
from ../include/tst_netlink.h:14,
from tst_crypto.c:13:
x86_64-buildroot-linux-musl/sysroot/usr/include/linux/sysinfo.h:8:8: error: redefinition of `struct sysinfo'
struct sysinfo {
^~~~~~~
In file included from ../include/tst_safe_macros.h:15,
from ../include/tst_test.h:93,
from tst_crypto.c:11:
x86_64-buildroot-linux-musl/sysroot/usr/include/sys/sysinfo.h:10:8: note: originally defined here