Fix broken Tx ring validation for AF_XDP. The commit under the Fixes
tag, fixed an off-by-one error in the validation but introduced
another error. Descriptors are now let through even if they straddle a
chunk boundary which they are not allowed to do in aligned mode. Worse
is that they are let through even if they straddle the end of the umem
itself, tricking the kernel to read data outside the allowed umem
region which might or might not be mapped at all.
Fix this by reintroducing the old code, but subtract the length by one
to fix the off-by-one error that the original patch was
addressing. The test chunk != chunk_end makes sure packets do not
straddle chunk boundraries. Note that packets of zero length are
allowed in the interface, therefore the test if the length is
non-zero.
Fixes: ac31565c21 ("xsk: Fix for xp_aligned_validate_desc() when len == chunk_size")
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Acked-by: Björn Töpel <bjorn@kernel.org>
Link: https://lore.kernel.org/bpf/20210618075805.14412-1-magnus.karlsson@gmail.com
Fix a missing validation of a Tx descriptor when executing in skb mode
and the umem is in unaligned mode. A descriptor could point to a
buffer straddling the end of the umem, thus effectively tricking the
kernel to read outside the allowed umem region. This could lead to a
kernel crash if that part of memory is not mapped.
In zero-copy mode, the descriptor validation code rejects such
descriptors by checking a bit in the DMA address that tells us if the
next page is physically contiguous or not. For the last page in the
umem, this bit is not set, therefore any descriptor pointing to a
packet straddling this last page boundary will be rejected. However,
the skb path does not use this bit since it copies out data and can do
so to two different pages. (It also does not have the array of DMA
address, so it cannot even store this bit.) The code just returned
that the packet is always physically contiguous. But this is
unfortunately also returned for the last page in the umem, which means
that packets that cross the end of the umem are being allowed, which
they should not be.
Fix this by introducing a check for this in the SKB path only, not
penalizing the zero-copy path.
Fixes: 2b43470add ("xsk: Introduce AF_XDP buffer allocation API")
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Björn Töpel <bjorn@kernel.org>
Link: https://lore.kernel.org/bpf/20210617092255.3487-1-magnus.karlsson@gmail.com
Without calling loop_config_discard() the discard flag and parameters
aren't set/updated for the loop device and worst-case they could
indicate discard support when it isn't the case (ex: if the
LOOP_SET_STATUS ioctl was used with a different file prior to
LOOP_CONFIGURE).
Cc: <stable@vger.kernel.org> # 5.8.x-
Fixes: 3448914e8c ("loop: Add LOOP_CONFIGURE ioctl")
Signed-off-by: Kristian Klausen <kristian@klausen.dk>
Link: https://lore.kernel.org/r/20210618115157.31452-1-kristian@klausen.dk
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The insert_requests and dispatch_request elevator operations are
mandatory for the correct execution of an elevator, and all implemented
elevators (bfq, kyber and mq-deadline) implement them. As a result,
there is no need to check for these operations before calling them when
a queue has an elevator set. This simplifies the code in
__blk_mq_sched_dispatch_requests() and blk_mq_sched_insert_request().
To avoid out-of-tree elevators to crash the kernel in case of bad
implementation, add a check in elv_register() to verify that these
operations are implemented.
A small, probably not significant, IOPS improvement of 0.1% is observed
with this patch applied (4.117 MIOPS to 4.123 MIOPS, average of 20 fio
runs doing 4K random direct reads with psync and 32 jobs).
Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Link: https://lore.kernel.org/r/20210618015922.713999-1-damien.lemoal@wdc.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
tagset can't be used after blk_cleanup_queue() is returned because
freeing tagset usually follows blk_clenup_queue(). Commit d97e594c51
("blk-mq: Use request queue-wide tags for tagset-wide sbitmap") adds
check on q->tag_set->flags in blk_mq_exit_sched(), and causes
use-after-free.
Fixes it by using hctx->flags.
Reported-by: syzbot+77ba3d171a25c56756ea@syzkaller.appspotmail.com
Fixes: d97e594c51 ("blk-mq: Use request queue-wide tags for tagset-wide sbitmap")
Cc: John Garry <john.garry@huawei.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Tested-by: John Garry <john.garry@huawei.com>
Reviewed-by: John Garry <john.garry@huawei.com>
Link: https://lore.kernel.org/r/20210609063046.122843-1-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Two carveout reserved memory nodes each have been added for each of the
R5F remote processor devices within the MAIN domain on the TI AM642 EVM
and SK boards. These nodes are assigned to the respective rproc device
nodes as well. The first region will be used as the DMA pool for the rproc
devices, and the second region will furnish the static carveout regions
for the firmware memory.
An additional reserved memory node is also added to reserve a portion of
the DDR memory to be used for performing inter-processor communication
between all the remote processors running RTOS or baremetal firmwares.
8 MB of memory is reserved for this purpose, and this accounts for all
the vrings and vring buffers between all the possible pairs of remote
processors.
The current carveout addresses and sizes are defined statically for each
rproc device. The R5F processors do not have an MMU, and as such require
the exact memory used by the firmwares to be set-aside. The firmware
images do not require any RSC_CARVEOUT entries in their resource tables
to allocate the memory for firmware memory segments.
NOTE:
1. The R5F1 carveouts are needed only if the R5F cluster is running in
Split (non Single-CPU) mode. The reserved memory nodes can be disabled
later on if there is no use-case defined to use the corresponding
remote processor.
2. The AM64x SoCs do not have any DSPs and one less R5F cluster compared
to J721E SoCs. So, while the carveout memories reserved for the R5F
clusters present on the SoC match to those on J721E, the overall
memory map reserved for firmwares is quite different. The number of
R5F clusters on AM64x SoCs are same as on J7200 SoCs, but the AM64x
SoCs also have an additional M4F core, so the RTOS IPC memory region
is 1 MB higher than on J7200 SoCs.
Signed-off-by: Suman Anna <s-anna@ti.com>
Reviewed-by: Praneeth Bajjuri <praneeth@ti.com>
Signed-off-by: Nishanth Menon <nm@ti.com>
Link: https://lore.kernel.org/r/20210615195718.15898-4-s-anna@ti.com
Add the required 'mboxes' property to all the R5F processors for the
TI AM642 EVM and SK boards. The mailboxes and some shared memory are
required for running the Remote Processor Messaging (RPMsg) stack
between the host processor and each of the R5Fs.
The chosen sub-mailboxes match the values used in the current firmware
images. This can be changed, if needed, as per the system integration
needs after making appropriate changes on the firmware side as well.
Note that any R5F Core1 resources are needed and used only when that
R5F cluster is configured for Split-mode.
Signed-off-by: Suman Anna <s-anna@ti.com>
Reviewed-by: Praneeth Bajjuri <praneeth@ti.com>
Signed-off-by: Nishanth Menon <nm@ti.com>
Link: https://lore.kernel.org/r/20210615195718.15898-3-s-anna@ti.com
The AM64x SoCs have 2 dual-core Arm Cortex-R5F processor (R5FSS)
subsystems/clusters. Both the R5F clusters are present within the
MAIN domain (MAIN_R5FSS0 & MAIN_R5FSS1). Each of these can be
configured at boot time to be either run in a new "Single-CPU" mode
or in an Asymmetric Multi Processing (AMP) fashion in Split-mode.
The mode is restricted to "Single-CPU" on some devices with the
appropriate eFuse bit set, but the most common devices support both
modes. These subsystems have 64 KB each Tightly-Coupled Memory (TCM)
internal memories for each core split between two banks - ATCM and
BTCM (further interleaved into two banks). The TCMs of both Cores
are combined in Single-CPU mode to provide a larger 128 KB of memory.
The other notable difference is that the TCMs are spaced 1 MB apart
on these SoCs unlike the existing SoCs.
Add the DT nodes for both these MAIN domain R5F cluster/subsystems,
the two R5F cores are added as child nodes to each of the corresponding
R5F cluster node. Both the clusters are configured to run in Split mode
by default, with the ATCMs enabled to allow the R5 cores to execute
code from DDR with boot-strapping code from ATCM. The inter-processor
communication between the main A72 cores and these processors is
achieved through shared memory and Mailboxes.
The following firmware names are used by default for these cores, and
can be overridden in a board dts file if desired:
MAIN R5FSS0 Core0: am64-main-r5f0_0-fw (both in Single-CPU & Split modes)
MAIN R5FSS0 Core1: am64-main-r5f0_1-fw (needed only in Split mode)
MAIN R5FSS1 Core0: am64-main-r5f1_0-fw (both in Single-CPU & Split modes)
MAIN R5FSS1 Core1: am64-main-r5f1_1-fw (needed only in Split mode)
NOTE:
A R5FSS cluster can be configured in "Single-CPU" mode by using a
value of 2 for the "ti,cluster-mode" property. Value of 1 is not
permitted (fails the dtbs_check).
Signed-off-by: Suman Anna <s-anna@ti.com>
Reviewed-by: Praneeth Bajjuri <praneeth@ti.com>
Signed-off-by: Nishanth Menon <nm@ti.com>
Link: https://lore.kernel.org/r/20210615195718.15898-2-s-anna@ti.com
The tprot() inline asm temporarily changes the program check new psw
to redirect a potential program check on the diag instruction.
Restoring of the program check new psw is done in C code behind the
inline asm.
This can be problematic, especially if the function is inlined, since
the compiler can reorder instructions in such a way that a different
instruction, which may result in a program check, might be executed
before the program check new psw has been restored.
To avoid such a scenario move restoring into the inline asm. For
consistency reasons move also saving of the original program check new
psw into the inline asm.
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
The __diag260() inline asm temporarily changes the program check new
psw to redirect a potential program check on the diag instruction.
Restoring of the program check new psw is done in C code behind the
inline asm.
This can be problematic, especially if the function is inlined, since
the compiler can reorder instructions in such a way that a different
instruction, which may result in a program check, might be executed
before the program check new psw has been restored.
To avoid such a scenario move restoring into the inline asm. For
consistency reasons move also saving of the original program check new
psw into the inline asm.
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
The __diag308() inline asm temporarily changes the program check new
psw to redirect a potential program check on the diag instruction.
Restoring of the program check new psw is done in C code behind the
inline asm.
This can be problematic, especially if the function is inlined, since
the compiler can reorder instructions in such a way that a different
instruction, which may result in a program check, might be executed
before the program check new psw has been restored.
To avoid such a scenario move restoring into the inline asm. For
consistency reasons move also saving of the original program check new
psw into the inline asm.
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Power management support was removed for s390 with
commit 394216275c ("s390: remove broken hibernate / power management
support").
Remove the hvc_iucv_driver, but keep the device struct around so that
it can continue to provide the hvc_iucv_dev_attr_groups attributes.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Acked-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Enable ztsd support in s390/boot, to enable booting with zstd
compressed kernel when configured with CONFIG_KERNEL_ZSTD=y.
BOOT_HEAP_SIZE is defined to 0x30000 in this case. Actual decompressor
memory usage with allyesconfig is currently 0x26150.
BugLink: https://bugs.launchpad.net/bugs/1931725
Signed-off-by: Dimitri John Ledkov <dimitri.ledkov@canonical.com>
cc: Heiko Carstens <hca@linux.ibm.com>
cc: Vasily Gorbik <gor@linux.ibm.com>
cc: Christian Borntraeger <borntraeger@de.ibm.com>
cc: linux-s390@vger.kernel.org
Link: https://lore.kernel.org/r/20210615114150.325080-1-dimitri.ledkov@canonical.com
[gor: added BOOT_HEAP_SIZE for zstd]
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Currently BOOT_HEAP_SIZE is always defined as 0x400000 due to
bogus condition. Use CONFIG_KERNEL_BZIP2 instead of
CONFIG_HAVE_KERNEL_BZIP2 to correct that.
BOOT_HEAP_SIZE of 0x10000 is still good enough for every decompressor
algorithm but bzip2. Actual decompressor memory usage with allyesconfig
is the following:
gzip 0xbc28
bzip2 0x379518
xz 0x7410
lzma 0x3e6c
lzo 0
lz4 0
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Remove register asm usage from diag8_noresponse() since it wasn't
needed at all. There is no requirement for even/odd register pairs for
diag 0x8.
For diag_response() use register pairs to fulfill the rx+1 and ry+1
requirements as required if a response buffer is specified. Also
change the inline asm to return the condition code of the diagnose
instruction and do the conditional handling of response length
calculation in C.
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Get rid of register asm statement and use a register pair.
This allows the compiler to allocate registers on its own.
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Introduce a register pair union, which is supposed to be used for
inline assemblies where instructions require parameters in even/odd
numbered register pairs.
This is more or less the same register pair construct which was
available for 31 bit builds which was removed with commit 5a79859ae0
("s390: remove 31 bit support").
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Power management support was removed for s390 with
commit 394216275c ("s390: remove broken hibernate / power management
support").
Remove leftover vmlogrdr-related power management code.
Acked-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Power management support was removed for s390 with
commit 394216275c ("s390: remove broken hibernate / power management
support").
Remove leftover sclp base-related power management code. Note that we
keep the registration of the sclp platform driver since it is used to
externalize non-PM related attributes in sysfs.
Acked-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Power management support was removed for s390 with
commit 394216275c ("s390: remove broken hibernate / power management
support").
Remove leftover sclp quiesce-related power management code.
Acked-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Power management support was removed for s390 with
commit 394216275c ("s390: remove broken hibernate / power management
support").
Remove leftover sclp memory hotplug-related power management code.
Acked-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Power management support was removed for s390 with
commit 394216275c ("s390: remove broken hibernate / power management
support").
Remove leftover sclp vt220-related power management code.
Acked-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Power management support was removed for s390 with
commit 394216275c ("s390: remove broken hibernate / power management
support").
Remove leftover sclp console-related power management code.
Acked-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Power management support was removed for s390 with
commit 394216275c ("s390: remove broken hibernate / power management
support").
Remove leftover monwriter-related power management code.
Acked-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Power management support was removed for s390 with
commit 394216275c ("s390: remove broken hibernate / power management
support").
Remove leftover monreader-related power management code.
Acked-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Power management support was removed for s390 with
commit 394216275c ("s390: remove broken hibernate / power management
support").
Remove leftover xpram-related power management code.
Acked-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Power management support was removed for s390 with
commit 394216275c ("s390: remove broken hibernate / power management
support").
Remove leftover dcssblk-related power management code.
Acked-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Currently css_wait_for_slow_path() gets called inside the chp->lock.
The path-verification-loop of slowpath inside this lock could lead to
deadlock as reported by the lockdep validator.
The ccw_device_get_chp_desc() during the instance of a device-set-online
would try to acquire the same 'chp->lock' to read the chp->desc.
The instance of this function can get called from multiple scenario,
like probing or setting-device online manually. This could, in some
corner-cases lead to the deadlock.
lockdep validator reported this as,
CPU0 CPU1
---- ----
lock(&chp->lock);
lock(kn->active#43);
lock(&chp->lock);
lock((wq_completion)cio);
The chp->lock was introduced to serialize the access of struct
channel_path. This lock is not needed for the css_wait_for_slow_path()
function, so invoke the slow-path function outside this lock.
Fixes: b730f3a933 ("[S390] cio: add lock to struct channel_path")
Cc: <stable@vger.kernel.org>
Reviewed-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Signed-off-by: Vineeth Vijayan <vneethv@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
All s390 irqflags functions are very small and should be always inlined.
Therefore mark them __always_inline. This also allows to get rid of the
rather odd notrace attribute for these small functions, which was only
added to prevent tracing iff any of these functions would not be inlined.
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
s390 is the only architecture which makes use of the __no_kasan_or_inline
attribute for two functions. Given that both stap() and __load_psw_mask()
are very small functions they can and should be always inlined anyway.
Therefore get rid of __no_kasan_or_inline and always inline these
functions.
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
When read via debugfs, s390dbf debug-views print the kernel address of
the call-site that created a trace entry. The kernel's %p pointer
hashing feature obfuscates this address, and commit 860ec7c6e2
("s390/debug: use pK for kernel pointers") made this obfuscation
configurable via the kptr_restrict sysctl.
Obfuscation of kernel address data printed via s390dbf debug-views does
not add any additional protection since the associated debugfs files are
only accessible to the root user that typically has enough other means
to obtain kernel address data.
Also trace payload data may contain binary representations of kernel
addresses as part of logged data structues. Requiring such payload data
to be obfuscated as well would be impractical and greatly diminish the
use of s390dbf.
Therefore completely remove pointer obfuscation from s390dbf
debug-views.
Reviewed-by: Steffen Maier <maier@linux.ibm.com>
Acked-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Since OLDMEM_BASE/OLDMEM_SIZE is already taken into consideration and is
reflected in ident_map_size. reserve/remove_oldmem() is no longer needed
and could be removed.
Reviewed-by: Alexander Gordeev <agordeev@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>