LTP testcase mtest06 [1] can trigger a crash on s390x running 5.0.0-rc8.
This is a stress test, where one thread mmaps/writes/munmaps memory area
and other thread is trying to read from it:
page_table_free() is called with NULL mm parameter, but because "0" is a
valid address on s390 (see S390_lowcore), it keeps going until it
eventually crashes in lockdep's lock_acquire. This crash is
reproducible at least since 4.14.
Problem is that "vmf->vma" used in do_fault() can become stale. Because
mmap_sem may be released, other threads can come in, call munmap() and
cause "vma" be returned to kmem cache, and get zeroed/re-initialized and
re-used:
handle_mm_fault |
__handle_mm_fault |
do_fault |
vma = vmf->vma |
do_read_fault |
__do_fault |
vma->vm_ops->fault(vmf); |
mmap_sem is released |
|
| do_munmap()
| remove_vma_list()
| remove_vma()
| vm_area_free()
| # vma is released
| ...
| # same vma is allocated
| # from kmem cache
| do_mmap()
| vm_area_alloc()
| memset(vma, 0, ...)
|
pte_free(vma->vm_mm, ...); |
page_table_free |
spin_lock_bh(&mm->context.lock);|
<crash> |
Cache mm_struct to avoid using potentially stale "vma".
When VM_NO_GUARD is not set area->size includes adjacent guard page,
thus for correct size checking get_vm_area_size() should be used, but
not area->size.
This fixes possible kernel oops when userspace tries to mmap an area on
1 page bigger than was allocated by vmalloc_user() call: the size check
inside remap_vmalloc_range_partial() accounts non-existing guard page
also, so check successfully passes but vmalloc_to_page() returns NULL
(guard page does not physically exist).
The following code pattern example should trigger an oops:
soft_offline_in_use_page() passed refcount and page lock from tail page
to head page, which is not needed because we can pass any subpage to
split_huge_page().
Naoya had fixed a similar issue in c3901e722b29 ("mm: hwpoison: fix thp
split handling in memory_failure()"). But he missed fixing soft
offline.
While do swap, we should make sure there has no new dirty page since we
should swap i_data between two inode:
1.We should lock i_mmap_sem with write to avoid new pagecache from mmap
read/write;
2.Change filemap_flush to filemap_write_and_wait and move them to the
space protected by inode lock to avoid new pagecache from buffer read/write.
Before really do swap between inode and boot inode, something need to
check to avoid invalid or not permitted operation, like does this inode
has inline data. But the condition check should be protected by inode
lock to avoid change while swapping. Also some other condition will not
change between swapping, but there has no problem to do this under inode
lock.
pxa_cpufreq_init_voltages() is marked __init but usually inlined into
the non-__init pxa_cpufreq_init() function. When building with clang,
it can stay as a standalone function in a discarded section, and produce
this warning:
WARNING: vmlinux.o(.text+0x616a00): Section mismatch in reference from the function pxa_cpufreq_init() to the function .init.text:pxa_cpufreq_init_voltages()
The function pxa_cpufreq_init() references
the function __init pxa_cpufreq_init_voltages().
This is often because pxa_cpufreq_init lacks a __init
annotation or the annotation of pxa_cpufreq_init_voltages is wrong.
Fixes: 50e77fcd790e ("ARM: pxa: remove __init from cpufreq_driver->init()") Signed-off-by: Arnd Bergmann <arnd@arndb.de> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Reviewed-by: Nathan Chancellor <natechancellor@gmail.com> Acked-by: Robert Jarzmik <robert.jarzmik@free.fr> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
Commit 5ad7346b4ae2 ("cpufreq: kryo: Add module remove and exit") made
it possible to build the kryo cpufreq driver as a module, but it failed
to release all the resources, i.e. OPP tables, when the module is
unloaded.
This patch fixes it by releasing the OPP tables, by calling
dev_pm_opp_put_supported_hw() for them, from the
qcom_cpufreq_kryo_remove() routine. The array of pointers to the OPP
tables is also allocated dynamically now in qcom_cpufreq_kryo_probe(),
as the pointers will be required while releasing the resources.
Prohibit probing on optprobe template code, since it is not
a code but a template instruction sequence. If we modify
this template, copied template must be broken.
Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andrea Righi <righi.andrea@gmail.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: stable@vger.kernel.org Fixes: 9326638cbee2 ("kprobes, x86: Use NOKPROBE_SYMBOL() instead of __kprobes annotation") Link: http://lkml.kernel.org/r/154998787911.31052.15274376330136234452.stgit@devbox Signed-off-by: Ingo Molnar <mingo@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
Using the irq_gc_lock/irq_gc_unlock functions in the suspend and
resume functions creates the opportunity for a deadlock during
suspend, resume, and shutdown. Using the irq_gc_lock_irqsave/
irq_gc_unlock_irqrestore variants prevents this possible deadlock.
Cc: stable@vger.kernel.org Fixes: 7f646e92766e2 ("irqchip: brcmstb-l2: Add Broadcom Set Top Box Level-2 interrupt controller") Signed-off-by: Doug Berger <opendmb@gmail.com> Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
[maz: tidied up $SUBJECT] Signed-off-by: Marc Zyngier <marc.zyngier@arm.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
In current logic, its_parse_indirect_baser() will be invoked twice
when allocating Device tables. Add a *break* to omit the unnecessary
and annoying (might be ...) invoking.
Fixes: 32bd44dc19de ("irqchip/gic-v3-its: Fix the incorrect parsing of VCPU table size") Cc: stable@vger.kernel.org Signed-off-by: Zenghui Yu <yuzenghui@huawei.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
Using the batch API from the interconnect driver sometimes leads to a
KASAN error due to an access to freed memory. This is easier to trigger
with threadirqs on the kernel commandline.
BUG: KASAN: use-after-free in rpmh_tx_done+0x114/0x12c
Read of size 1 at addr fffffff51414ad84 by task irq/110-apps_rs/57
Memory state around the buggy address: fffffff51414ac80: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fffffff51414ad00: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
>fffffff51414ad80: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
^ fffffff51414ae00: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fffffff51414ae80: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
The batch API sets the same completion for each rpmh message that's sent
and then loops through all the messages and waits for that single
completion declared on the stack to be completed before returning from
the function and freeing the message structures. Unfortunately, some
messages may still be in process and 'stuck' in the TCS. At some later
point, the tcs_tx_done() interrupt will run and try to process messages
that have already been freed at the end of rpmh_write_batch(). This will
in turn access the 'needs_free' member of the rpmh_request structure and
cause KASAN to complain. Furthermore, if there's a message that's
completed in rpmh_tx_done() and freed immediately after the complete()
call is made we'll be racing with potentially freed memory when
accessing the 'needs_free' member:
Let's fix this by allocating a chunk of completions for each message and
waiting for all of them to be completed before returning from the batch
API. Alternatively, we could wait for the last message in the batch, but
that may be a more complicated change because it looks like
tcs_tx_done() just iterates through the indices of the queue and
completes each message instead of tracking the last inserted message and
completing that first.
Fixes: c8790cb6da58 ("drivers: qcom: rpmh: add support for batch RPMH request") Cc: Lina Iyer <ilina@codeaurora.org> Cc: "Raju P.L.S.S.S.N" <rplsssn@codeaurora.org> Cc: Matthias Kaehlcke <mka@chromium.org> Cc: Evan Green <evgreen@chromium.org> Cc: stable@vger.kernel.org Reviewed-by: Lina Iyer <ilina@codeaurora.org> Reviewed-by: Evan Green <evgreen@chromium.org> Signed-off-by: Stephen Boyd <swboyd@chromium.org> Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org> Signed-off-by: Andy Gross <andy.gross@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
Reflinking (clone/dedupe) and rename are operations that operate on two
inodes and therefore need to lock them in the same order to avoid ABBA
deadlocks. It happens that Btrfs' reflink implementation always locked
them in a different order from VFS's lock_two_nondirectories() helper,
which is used by the rename code in VFS, resulting in ABBA type deadlocks.
In the past we had data corruption when reading compressed extents that
are shared within the same file and they are consecutive, this got fixed
by commit 005efedf2c7d0 ("Btrfs: fix read corruption of compressed and
shared extents") and by commit 808f80b46790f ("Btrfs: update fix for read
corruption of compressed and shared extents"). However there was a case
that was missing in those fixes, which is when the shared and compressed
extents are referenced with a non-zero offset. The following shell script
creates a reproducer for this issue:
#!/bin/bash
mkfs.btrfs -f /dev/sdc &> /dev/null
mount -o compress /dev/sdc /mnt/sdc
# Create a file with 3 consecutive compressed extents, each has an
# uncompressed size of 128Kb and a compressed size of 4Kb.
for ((i = 1; i <= 3; i++)); do
head -c 4096 /dev/zero
for ((j = 1; j <= 31; j++)); do
head -c 4096 /dev/zero | tr '\0' "\377"
done
done > /mnt/sdc/foobar
sync
echo "Digest after file creation: $(md5sum /mnt/sdc/foobar)"
# Clone the first extent into offsets 128K and 256K.
xfs_io -c "reflink /mnt/sdc/foobar 0 128K 128K" /mnt/sdc/foobar
xfs_io -c "reflink /mnt/sdc/foobar 0 256K 128K" /mnt/sdc/foobar
sync
echo "Digest after cloning: $(md5sum /mnt/sdc/foobar)"
# Punch holes into the regions that are already full of zeroes.
xfs_io -c "fpunch 0 4K" /mnt/sdc/foobar
xfs_io -c "fpunch 128K 4K" /mnt/sdc/foobar
xfs_io -c "fpunch 256K 4K" /mnt/sdc/foobar
sync
echo "Digest after hole punching: $(md5sum /mnt/sdc/foobar)"
When running the script we get the following output:
Digest after file creation: 5a0888d80d7ab1fd31c229f83a3bbcc8 /mnt/sdc/foobar
linked 131072/131072 bytes at offset 131072
128 KiB, 1 ops; 0.0033 sec (36.960 MiB/sec and 295.6830 ops/sec)
linked 131072/131072 bytes at offset 262144
128 KiB, 1 ops; 0.0015 sec (78.567 MiB/sec and 628.5355 ops/sec)
Digest after cloning: 5a0888d80d7ab1fd31c229f83a3bbcc8 /mnt/sdc/foobar
Digest after hole punching: 5a0888d80d7ab1fd31c229f83a3bbcc8 /mnt/sdc/foobar
Dropping page cache...
Digest after hole punching: fba694ae8664ed0c2e9ff8937e7f1484 /mnt/sdc/foobar
This happens because after reading all the pages of the extent in the
range from 128K to 256K for example, we read the hole at offset 256K
and then when reading the page at offset 260K we don't submit the
existing bio, which is responsible for filling all the page in the
range 128K to 256K only, therefore adding the pages from range 260K
to 384K to the existing bio and submitting it after iterating over the
entire range. Once the bio completes, the uncompressed data fills only
the pages in the range 128K to 256K because there's no more data read
from disk, leaving the pages in the range 260K to 384K unfilled. It is
just a slightly different variant of what was solved by commit 005efedf2c7d0 ("Btrfs: fix read corruption of compressed and shared
extents").
Fix this by forcing a bio submit, during readpages(), whenever we find a
compressed extent map for a page that is different from the extent map
for the previous page or has a different starting offset (in case it's
the same compressed extent), instead of the extent map's original start
offset.
A test case for fstests follows soon.
Reported-by: Zygo Blaxell <ce3g8jdj@umail.furryterror.org> Fixes: 808f80b46790f ("Btrfs: update fix for read corruption of compressed and shared extents") Fixes: 005efedf2c7d0 ("Btrfs: fix read corruption of compressed and shared extents") Cc: stable@vger.kernel.org # 4.3+ Tested-by: Zygo Blaxell <ce3g8jdj@umail.furryterror.org> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
The scrub_ctx csum_list member must be initialized before scrub_free_ctx
is called. If the csum_list is not initialized beforehand, the
list_empty call in scrub_free_csums will result in a null deref if the
allocation fails in the for loop.
Fixes: a2de733c78fa ("btrfs: scrub") CC: stable@vger.kernel.org # 3.0+ Reviewed-by: Nikolay Borisov <nborisov@suse.com> Signed-off-by: Dan Robertson <dan@dlrobertson.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
We recently had a customer issue with a corrupted filesystem. When
trying to mount this image btrfs panicked with a division by zero in
calc_stripe_length().
The corrupt chunk had a 'num_stripes' value of 1. calc_stripe_length()
takes this value and divides it by the number of copies the RAID profile
is expected to have to calculate the amount of data stripes. As a DUP
profile is expected to have 2 copies this division resulted in 1/2 = 0.
Later then the 'data_stripes' variable is used as a divisor in the
stripe length calculation which results in a division by 0 and thus a
kernel panic.
When encountering a filesystem with a DUP block group and a
'num_stripes' value unequal to 2, refuse mounting as the image is
corrupted and will lead to unexpected behaviour.
Code inspection showed a RAID1 block group has the same issues.
Fixes: e06cd3dd7cea ("Btrfs: add validadtion checks for chunk loading") CC: stable@vger.kernel.org # 4.4+ Reviewed-by: Qu Wenruo <wqu@suse.com> Reviewed-by: Nikolay Borisov <nborisov@suse.com> Signed-off-by: Johannes Thumshirn <jthumshirn@suse.de> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
We should drop the lock on this error path. This has been found by a
static tool.
The lock needs to be released, it's there to protect access to the
dev_replace members and is not supposed to be left locked. The value of
state that's being switched would need to be artifically changed to an
invalid value so the default: branch is taken.
Fixes: d189dd70e255 ("btrfs: fix use-after-free due to race between replace start and cancel") CC: stable@vger.kernel.org # 5.0+ Reviewed-by: Anand Jain <anand.jain@oracle.com> Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
This fixes a longstanding lockdep warning triggered by
fstests/btrfs/011.
Circular locking dependency check reports warning[1], that's because the
btrfs_scrub_dev() calls the stack #0 below with, the fs_info::scrub_lock
held. The test case leading to this warning:
In fact we have fs_info::scrub_workers_refcnt to track if the init and destroy
of the scrub workers are needed. So once we have incremented and decremented
the fs_info::scrub_workers_refcnt value in the thread, its ok to drop the
scrub_lock, and then actually do the btrfs_destroy_workqueue() part. So this
patch drops the scrub_lock before calling btrfs_destroy_workqueue().
We are holding a transaction handle when setting an acl, therefore we can
not allocate the xattr value buffer using GFP_KERNEL, as we could deadlock
if reclaim is triggered by the allocation, therefore setup a nofs context.
Fixes: 39a27ec1004e8 ("btrfs: use GFP_KERNEL for xattr and acl allocations") CC: stable@vger.kernel.org # 4.9+ Reviewed-by: Nikolay Borisov <nborisov@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
We are holding a transaction handle when creating a tree, therefore we can
not allocate the root using GFP_KERNEL, as we could deadlock if reclaim is
triggered by the allocation, therefore setup a nofs context.
Fixes: 74e4d82757f74 ("btrfs: let callers of btrfs_alloc_root pass gfp flags") CC: stable@vger.kernel.org # 4.9+ Reviewed-by: Nikolay Borisov <nborisov@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
This patch fixes a build failure when using GCC 8.1:
/usr/bin/ld: block/partitions/ldm.o: in function `ldm_parse_tocblock':
block/partitions/ldm.c:153: undefined reference to `strcmp'
This is caused by a new optimization which effectively replaces a
strncmp() call with a strcmp() call. This affects a number of strncmp()
call sites in the kernel.
The entire class of optimizations is avoided with -fno-builtin, which
gets enabled by -ffreestanding. This may avoid possible future build
failures in case new optimizations appear in future compilers.
I haven't done any performance measurements with this patch but I did
count the function calls in a defconfig build. For example, there are now
23 more sprintf() calls and 39 fewer strcpy() calls. The effect on the
other libc functions is smaller.
If this harms performance we can tackle that regression by optimizing
the call sites, ideally using semantic patches. That way, clang and ICC
builds might benfit too.
If a file has been copied up metadata only, and later data is copied up,
upper loses any security.capability xattr it has (underlying filesystem
clears it as upon file write).
From a user's point of view, this is just a file copy-up and that should
not result in losing security.capability xattr. Hence, before data copy
up, save security.capability xattr (if any) and restore it on upper after
data copy up is complete.
Signed-off-by: Vivek Goyal <vgoyal@redhat.com> Reviewed-by: Amir Goldstein <amir73il@gmail.com> Fixes: 0c2888749363 ("ovl: A new xattr OVL_XATTR_METACOPY for file on upper") Cc: <stable@vger.kernel.org> # v4.19+ Signed-off-by: Miklos Szeredi <mszeredi@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
If a file with capability set (and hence security.capability xattr) is
written kernel clears security.capability xattr. For overlay, during file
copy up if xattrs are copied up first and then data is, copied up. This
means data copy up will result in clearing of security.capability xattr
file on lower has. And this can result into surprises. If a lower file has
CAP_SETUID, then it should not be cleared over copy up (if nothing was
actually written to file).
This also creates problems with chown logic where it first copies up file
and then tries to clear setuid bit. But by that time security.capability
xattr is already gone (due to data copy up), and caller gets -ENODATA.
This has been reported by Giuseppe here.
Fix this by copying up data first and then metadta. This is a regression
which has been introduced by my commit as part of metadata only copy up
patches.
TODO: There will be some corner cases where a file is copied up metadata
only and later data copy up happens and that will clear security.capability
xattr. Something needs to be done about that too.
Fixes: bd64e57586d3 ("ovl: During copy up, first copy up metadata and then data") Cc: <stable@vger.kernel.org> # v4.19+ Reported-by: Giuseppe Scrivano <gscrivan@redhat.com> Signed-off-by: Vivek Goyal <vgoyal@redhat.com> Signed-off-by: Miklos Szeredi <mszeredi@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
Before this patch, it was possible for two pipes to affect each other after
data had been transferred between them with tee():
============
$ cat tee_test.c
int main(void) {
int pipe_a[2];
if (pipe(pipe_a)) err(1, "pipe");
int pipe_b[2];
if (pipe(pipe_b)) err(1, "pipe");
if (write(pipe_a[1], "abcd", 4) != 4) err(1, "write");
if (tee(pipe_a[0], pipe_b[1], 2, 0) != 2) err(1, "tee");
if (write(pipe_b[1], "xx", 2) != 2) err(1, "write");
As suggested by Al Viro, fix it by creating a separate type for
non-mergeable pipe buffers, then changing the types of buffers in
splice_pipe_to_pipe() and link_pipe().
Cc: <stable@vger.kernel.org> Fixes: 7c77f0b3f920 ("splice: implement pipe to pipe splicing") Fixes: 70524490ee2e ("[PATCH] splice: add support for sys_tee()") Suggested-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
d_delete only unhashes an entry if it is reached with
dentry->d_lockref.count != 1. Prior to commit 8ead9dd54716 ("devpts:
more pty driver interface cleanups"), d_delete was called on a dentry
from devpts_pty_kill with two references held, which would trigger the
unhashing, and the subsequent dputs would release it.
Commit 8ead9dd54716 reworked devpts_pty_kill to stop acquiring the second
reference from d_find_alias, and the d_delete call left the dentries
still on the hashed list without actually ever being dropped from dcache
before explicit cleanup. This causes the number of negative dentries for
devpts to pile up, and an `ls /dev/pts` invocation can take seconds to
return.
Provide always_delete_dentry() from simple_dentry_operations
as .d_delete for devpts, to make the dentry be dropped from dcache.
Without this cleanup, the number of dentries in /dev/pts/ can be grown
arbitrarily as:
This patch fixes warning seen when BLK-MQ is enabled and hardware does not
support MQ. This will result into driver requesting MSIx vectors which are
equal or less than pre_desc via PCI IRQ Affinity infrastructure.
This patch fixes LUN discovery when loop ID is not yet assigned by the
firmware during driver load/sg_reset operations. Driver will now search for
new loop id before retrying login.
Fixes: 48acad099074 ("scsi: qla2xxx: Fix N2N link re-connect") Cc: stable@vger.kernel.org #4.19 Signed-off-by: Himanshu Madhani <hmadhani@marvell.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
When using SCSI passthrough in combination with the iSCSI target driver
then cmd->t_state_lock may be obtained from interrupt context. Hence, all
code that obtains cmd->t_state_lock from thread context must disable
interrupts first. This patch avoids that lockdep reports the following:
WARNING: inconsistent lock state
4.18.0-dbg+ #1 Not tainted
--------------------------------
inconsistent {HARDIRQ-ON-W} -> {IN-HARDIRQ-W} usage.
iscsi_ttx/1800 [HC1[1]:SC0[2]:HE0:SE0] takes: 000000006e7b0ceb (&(&cmd->t_state_lock)->rlock){?...}, at: target_complete_cmd+0x47/0x2c0 [target_core_mod]
{HARDIRQ-ON-W} state was registered at:
lock_acquire+0xd2/0x260
_raw_spin_lock+0x32/0x50
iscsit_close_connection+0x97e/0x1020 [iscsi_target_mod]
iscsit_take_action_for_connection_exit+0x108/0x200 [iscsi_target_mod]
iscsi_target_rx_thread+0x180/0x190 [iscsi_target_mod]
kthread+0x1cf/0x1f0
ret_from_fork+0x24/0x30
irq event stamp: 1281
hardirqs last enabled at (1279): [<ffffffff970ade79>] __local_bh_enable_ip+0xa9/0x160
hardirqs last disabled at (1281): [<ffffffff97a008a5>] interrupt_entry+0xb5/0xd0
softirqs last enabled at (1278): [<ffffffff977cd9a1>] lock_sock_nested+0x51/0xc0
softirqs last disabled at (1280): [<ffffffffc07a6e04>] ip6_finish_output2+0x124/0xe40 [ipv6]
other info that might help us debug this:
Possible unsafe locking scenario:
It was reported that some devices report an OPTIMAL TRANSFER LENGTH of
0xFFFF blocks. That looks bogus, especially for a device with a
4096-byte physical block size.
Ignore OPTIMAL TRANSFER LENGTH if it is not a multiple of the device's
reported physical block size.
To make the sanity checking conditionals more readable--and to
facilitate printing warnings--relocate the checking to a helper
function. No functional change aside from the printks.
Cc: <stable@vger.kernel.org>
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=199759 Reported-by: Christoph Anton Mitterer <calestyo@scientia.net> Reviewed-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
The virtio scsi spec defines struct virtio_scsi_ctrl_tmf as a set of
device-readable records and a single device-writable response entry:
struct virtio_scsi_ctrl_tmf
{
// Device-readable part
le32 type;
le32 subtype;
u8 lun[8];
le64 id;
// Device-writable part
u8 response;
}
The above should be organised as two descriptor entries (or potentially
more if using VIRTIO_F_ANY_LAYOUT), but without any extra data after "le64
id" or after "u8 response".
The Linux driver doesn't respect that, with virtscsi_abort() and
virtscsi_device_reset() setting cmd->sc before calling virtscsi_tmf(). It
results in the original scsi command payload (or writable buffers) added to
the tmf.
This fixes the problem by leaving cmd->sc zeroed out, which makes
virtscsi_kick_cmd() add the tmf to the control vq without any payload.
Cc: stable@vger.kernel.org Signed-off-by: Felipe Franciosi <felipe@nutanix.com> Reviewed-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
A queue with a capacity of zero is clearly not a valid virtio queue.
Some emulators report zero queue size if queried with an invalid queue
index. Instead of crashing in this case let us just return -ENOENT. To
make that work properly, let us fix the notifier cleanup logic as well.
Cc: stable@vger.kernel.org Signed-off-by: Halil Pasic <pasic@linux.ibm.com> Signed-off-by: Cornelia Huck <cohuck@redhat.com> Signed-off-by: Michael S. Tsirkin <mst@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
The setup_lowcore() function creates a new prefix page for the boot CPU.
The PSW mask for the system_call, external interrupt, i/o interrupt and
the program check handler have the DAT bit set in this new prefix page.
At the time setup_lowcore is called the system still runs without virtual
address translation, the paging_init() function creates the kernel page
table and loads the CR13 with the kernel ASCE.
Any code between setup_lowcore() and the end of paging_init() that has
a BUG or WARN statement will create a program check that can not be
handled correctly as there is no kernel page table yet.
To allow early WARN statements initially setup the lowcore with DAT off
and set the DAT bit only after paging_init() has completed.
Libudev relies on having a subsystem link for non-root devices. To
avoid libudev (and potentially other userspace tools) choking on the
matrix device let us introduce a matrix bus and with it the matrix
bus subsytem. Also make the matrix device reside within the matrix
bus.
Doing this we remove the forced link from the matrix device to the
vfio_ap driver and the device_type we do not need anymore.
Since the associated matrix driver is not the vfio_ap driver any more,
we have to change the search for the devices on the vfio_ap driver in
the function vfio_ap_verify_queue_reserved. Fixes: 1fde573413b5 ("s390: vfio-ap: base implementation of VFIO AP device driver") Cc: stable@vger.kernel.org Reported-by: Marc Hartmayer <mhartmay@linux.ibm.com> Reported-by: Christian Borntraeger <borntraeger@de.ibm.com> Signed-off-by: Pierre Morel <pmorel@linux.ibm.com> Tested-by: Christian Borntraeger <borntraeger@de.ibm.com> Reviewed-by: Cornelia Huck <cohuck@redhat.com> Reviewed-by: Tony Krowiak <akrowiak@linux.ibm.com> Acked-by: Halil Pasic <pasic@linux.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
The Allwinner A64 SoC is known[1] to have an unstable architectural
timer, which manifests itself most obviously in the time jumping forward
a multiple of 95 years[2][3]. This coincides with 2^56 cycles at a
timer frequency of 24 MHz, implying that the time went slightly backward
(and this was interpreted by the kernel as it jumping forward and
wrapping around past the epoch).
Investigation revealed instability in the low bits of CNTVCT at the
point a high bit rolls over. This leads to power-of-two cycle forward
and backward jumps. (Testing shows that forward jumps are about twice as
likely as backward jumps.) Since the counter value returns to normal
after an indeterminate read, each "jump" really consists of both a
forward and backward jump from the software perspective.
Unless the kernel is trapping CNTVCT reads, a userspace program is able
to read the register in a loop faster than it changes. A test program
running on all 4 CPU cores that reported jumps larger than 100 ms was
run for 13.6 hours and reported the following:
This works out to a jump larger than 100 ms about every 5.5 seconds on
each CPU core.
The largest jump (almost an hour!) was the following sequence of reads:
0x0000007fffffffff → 0x00000093feffffff → 0x0000008000000000
Note that the middle bits don't necessarily all read as all zeroes or
all ones during the anomalous behavior; however the low 10 bits checked
by the function in this patch have never been observed with any other
value.
Also note that smaller jumps are much more common, with backward jumps
of 2048 (2^11) cycles observed over 400 times per second on each core.
(Of course, this is partially explained by lower bits rolling over more
frequently.) Any one of these could have caused the 95 year time skip.
Similar anomalies were observed while reading CNTPCT (after patching the
kernel to allow reads from userspace). However, the CNTPCT jumps are
much less frequent, and only small jumps were observed. The same program
as before (except now reading CNTPCT) observed after 72 hours:
Further investigation showed that the instability in CNTPCT/CNTVCT also
affected the respective timer's TVAL register. The following values were
observed immediately after writing CNVT_TVAL to 0x10000000:
I was also twice able to reproduce the issue covered by Allwinner's
workaround[4], that writing to TVAL sometimes fails, and both CVAL and
TVAL are left with entirely bogus values. One was the following values:
CNTVCT | CNTV_TVAL | CNTV_CVAL
--------------------+------------+--------------------------------------
0x000000d4a2d6014c | 0x8fbd5721 | 0x000000d132935fff (615s in the past) Reviewed-by: Marc Zyngier <marc.zyngier@arm.com>
========================================================================
Because the CPU can read the CNTPCT/CNTVCT registers faster than they
change, performing two reads of the register and comparing the high bits
(like other workarounds) is not a workable solution. And because the
timer can jump both forward and backward, no pair of reads can
distinguish a good value from a bad one. The only way to guarantee a
good value from consecutive reads would be to read _three_ times, and
take the middle value only if the three values are 1) each unique and
2) increasing. This takes at minimum 3 counter cycles (125 ns), or more
if an anomaly is detected.
However, since there is a distinct pattern to the bad values, we can
optimize the common case (1022/1024 of the time) to a single read by
simply ignoring values that match the error pattern. This still takes no
more than 3 cycles in the worst case, and requires much less code. As an
additional safety check, we still limit the loop iteration to the number
of max-frequency (1.2 GHz) CPU cycles in three 24 MHz counter periods.
For the TVAL registers, the simple solution is to not use them. Instead,
read or write the CVAL and calculate the TVAL value in software.
Although the manufacturer is aware of at least part of the erratum[4],
there is no official name for it. For now, use the kernel-internal name
"UNKNOWN1".
When shutting down the timer, ensure that after we have stopped the
timer any pending interrupts are cleared. This fixes a problem when
suspending, as interrupts are disabled before the timer is stopped,
so the timer interrupt may still be asserted, preventing the system
entering a low power state when the wfi is executed.
Signed-off-by: Stuart Menefy <stuart.menefy@mathembedded.com> Reviewed-by: Krzysztof Kozlowski <krzk@kernel.org> Tested-by: Marek Szyprowski <m.szyprowski@samsung.com> Cc: <stable@vger.kernel.org> # v4.3+ Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
When a timer tick occurs and the clock is in one-shot mode, the timer
needs to be stopped to prevent it triggering subsequent interrupts.
Currently this code is in exynos4_mct_tick_clear(), but as it is
only needed when an ISR occurs move it into exynos4_mct_tick_isr(),
leaving exynos4_mct_tick_clear() just doing what its name suggests it
should.
Signed-off-by: Stuart Menefy <stuart.menefy@mathembedded.com> Reviewed-by: Krzysztof Kozlowski <krzk@kernel.org> Tested-by: Marek Szyprowski <m.szyprowski@samsung.com> Cc: stable@vger.kernel.org # v4.3+ Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
The step values for some of the LDOs appears to be incorrect, resulting
in incorrect voltages (or at least, ones which are different from the
Samsung 3.4 vendor kernel).
Signed-off-by: Stuart Menefy <stuart.menefy@mathembedded.com> Reviewed-by: Krzysztof Kozlowski <krzk@kernel.org> Signed-off-by: Mark Brown <broonie@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
If regulator DT node doesn't exist, its of_parse_cb callback
function isn't called. Then all values for DT properties are
filled with zero. This leads to wrong register update for
FPS and POK settings.
Signed-off-by: Jinyoung Park <jinyoungp@nvidia.com> Signed-off-by: Mark Zhang <markz@nvidia.com> Signed-off-by: Mark Brown <broonie@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
LDO35 uses 25 mV step, not 50 mV. Bucks 7 and 8 use 12.5 mV step
instead of 6.25 mV. Wrong step caused over-voltage (LDO35) or
under-voltage (buck7 and 8) if regulators were used (e.g. on Exynos5420
Arndale Octa board).
Cc: <stable@vger.kernel.org> Fixes: cb74685ecb39 ("regulator: s2mps11: Add samsung s2mps11 regulator driver") Signed-off-by: Krzysztof Kozlowski <krzk@kernel.org> Signed-off-by: Mark Brown <broonie@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
spi-gpio is capable of dealing with active-high chip-selects.
Unfortunately, commit 4b859db2c606 ("spi: spi-gpio: add SPI_3WIRE
support") broke this by setting master->mode_bits, which overrides
the setting in the spi-bitbang code. Fix this.
[Fixed a trivial conflict with SPI_3WIRE_HIZ support -- broonie]
Fixes: 4b859db2c606 ("spi: spi-gpio: add SPI_3WIRE support") Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk> Signed-off-by: Mark Brown <broonie@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
Commit b682cffa3ac6 ("spi: omap2-mcspi: Set FIFO DMA trigger level to word length")
broke SPI transfers where bits_per_word != 8. This is because of
mimsatch between McSPI FIFO level event trigger size (SPI word length) and
DMA request size(word length * maxburst). This leads to data
corruption, lockup and errors like:
spi1.0: EOW timed out
Fix this by setting DMA maxburst size to 1 so that
McSPI FIFO level event trigger size matches DMA request size.
Fixes: b682cffa3ac6 ("spi: omap2-mcspi: Set FIFO DMA trigger level to word length") Cc: stable@vger.kernel.org Reported-by: David Lechner <david@lechnology.com> Tested-by: David Lechner <david@lechnology.com> Signed-off-by: Vignesh R <vigneshr@ti.com> Signed-off-by: Mark Brown <broonie@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
("spi: pxa2xx: Switch to SPI core DMA mapping functionality")
switches to SPI core provided DMA helpers, it missed to setup maximum
supported DMA transfer length for the controller and thus users
mistakenly try to send more data than supported with the following
warning:
ili9341 spi-PRP0001:01: DMA disabled for transfer length 153600 greater than 65536
Setup maximum supported DMA transfer length in order to make users know
the limit.
Fixes: b6ced294fb61 ("spi: pxa2xx: Switch to SPI core DMA mapping functionality") Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Mark Brown <broonie@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
Commit 4dea6c9b0b64 ("spi: spi-ti-qspi: add mmap mode read support") has
has got order of parameter wrong when calling regmap_update_bits() to
select CS for mmap access. Mask and value arguments are interchanged.
Code will work on a system with single slave, but fails when more than
one CS is in use. Fix this by correcting the order of parameters when
calling regmap_update_bits().
Fixes: 4dea6c9b0b64 ("spi: spi-ti-qspi: add mmap mode read support") Cc: stable@vger.kernel.org Signed-off-by: Vignesh R <vigneshr@ti.com> Signed-off-by: Mark Brown <broonie@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
The original purpose of the code I fix is to replace max_discard with
max_trim if max_trim is less than max_discard. When max_discard is 0
we should replace max_discard with max_trim as well, because
max_discard equals 0 happens only when the max_do_calc_max_discard
process is overflowed, so if mmc_can_trim(card) is true, max_discard
should be replaced by an available max_trim.
However, in the original code, there are two lines of code interfere
the right process.
1) if (max_discard && mmc_can_trim(card))
when max_discard is 0, it skips the process checking if max_discard
needs to be replaced with max_trim.
2) if (max_trim < max_discard)
the condition is false when max_discard is 0. it also skips the process
that replaces max_discard with max_trim, in fact, we should replace the
0-valued max_discard with max_trim.
This fixes card initialization failure in high speed mode.
If U-Boot uses SDR or HS200/400 mode before starting Linux and Linux
DT does not enable SDR/HS200/HS400 mode, card initialization fails in
high speed mode.
It is necessary to initialize SCC registers during card initialization
phase. HW reset function is registered only for a port with either of
SDR/HS200/HS400 properties in device tree. If SDR/HS200/HS400 properties
are not present in device tree, SCC registers will not be reset. In SoC
that support SCC registers, HW reset function should be registered
regardless of the configuration of device tree.
Reproduction procedure:
- Use U-Boot that support MMC HS200/400 mode.
- Delete HS200/HS400 properties in device tree.
(Delete mmc-hs200-1_8v and mmc-hs400-1_8v)
- MMC port works high speed mode and all commands fail.
Now tuning reset will be done when the timing is MMC_TIMING_LEGACY/
MMC_TIMING_MMC_HS/MMC_TIMING_SD_HS. But for timing MMC_TIMING_MMC_HS,
we can not do tuning reset, otherwise HS400 timing is not right.
Here is the process of init HS400, first finish tuning in HS200 mode,
then switch to HS mode and 8 bit DDR mode, finally switch to HS400
mode. If we do tuning reset in HS mode, this will cause HS400 mode
lost the tuning setting, which will cause CRC error.
Signed-off-by: Haibo Chen <haibo.chen@nxp.com> Cc: stable@vger.kernel.org # v4.12+ Acked-by: Adrian Hunter <adrian.hunter@intel.com> Fixes: d9370424c948 ("mmc: sdhci-esdhc-imx: reset tuning circuit when power on mmc card") Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
If SSDT overlay is loaded via ConfigFS and then unloaded the device,
we would like to have OF modalias for, already gone. Thus, acpi_get_name()
returns no allocated buffer for such case and kernel crashes afterwards:
Commit f7c90c2aa40048 ("x86/xen: don't write ptes directly in 32-bit
PV guests") introduced a regression for booting dom0 on huge systems
with lots of RAM (in the TB range).
Reason is that on those hosts the p2m list needs to be moved early in
the boot process and this requires temporary page tables to be created.
Said commit modified xen_set_pte_init() to use a hypercall for writing
a PTE, but this requires the page table being in the direct mapped
area, which is not the case for the temporary page tables used in
xen_relocate_p2m().
As the page tables are completely written before being linked to the
actual address space instead of set_pte() a plain write to memory can
be used in xen_relocate_p2m().
The first version of this method was missing the check for
`ret == PATH_MAX`; then such a check was added, but it didn't call kfree()
on error, so there was still a small memory leak in the error case.
Fix it by using strndup_user() instead of open-coding it.
Commit d716ff71dd12 ("tracing: Remove taking of trace_types_lock in
pipe files") use the current tracer instead of the copy in
tracing_open_pipe(), but it forget to remove the freeing sentence in
the error path.
There's an error path that can call kfree(iter->trace) after the iter->trace
was assigned to tr->current_trace, which would be bad to free.
Because there may be random garbage beyond a string's null terminator,
it's not correct to copy the the complete character array for use as a
hist trigger key. This results in multiple histogram entries for the
'same' string key.
So, in the case of a string key, use strncpy instead of memcpy to
avoid copying in the extra bytes.
Before, using the gdbus entries in the following hist trigger as an
example:
We negotiate rsize mounts (and it can be overridden by user) to
typically 4MB, so using larger default I/O sizes from userspace
(changing to 1MB default i/o size returned by stat) the
performance is much better (and not just for long latency
network connections) in most use cases for SMB3 than the default I/O
size (which ends up being 128K for cp and can be even smaller for cp).
This can be 4x slower or worse depending on network latency.
By changing inode->blocksize from 32K (which was perhaps ok
for very old SMB1/CIFS) to a larger value, 1MB (but still less than
max size negotiated with the server which is 4MB, in order to minimize
risk) it significantly increases performance for the
noncached case, and slightly increases it for the cached case.
This can be changed by the user on mount (specifying bsize=
values from 16K to 16MB) to tune better for performance
for applications that depend on blocksize.
Signed-off-by: Steve French <stfrench@microsoft.com> Reviewed-by: Ronnie Sahlberg <lsahlber@redhat.com> CC: Stable <stable@vger.kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
When we have a READ lease for a file and have just issued a write
operation to the server we need to purge the cache and set oplock/lease
level to NONE to avoid reading stale data. Currently we do that
only if a write operation succedeed thus not covering cases when
a request was sent to the server but a negative error code was
returned later for some other reasons (e.g. -EIOCBQUEUED or -EINTR).
Fix this by turning off caching regardless of the error code being
returned.
The patches fixes generic tests 075 and 112 from the xfs-tests.
Cc: <stable@vger.kernel.org> Signed-off-by: Pavel Shilovsky <pshilov@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com> Reviewed-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
When we hit failures during constructing MIDs or sending PDUs
through the network, we end up not using message IDs assigned
to the packet. The next SMB packet will skip those message IDs
and continue with the next one. This behavior may lead to a server
not granting us credits until we use the skipped IDs. Fix this by
reverting the current ID to the original value if any errors occur
before we push the packet through the network stack.
This patch fixes the generic/310 test from the xfs-tests.
Cc: <stable@vger.kernel.org> # 4.19.x Signed-off-by: Pavel Shilovsky <pshilov@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
Currently on lease break the client sets a caching level twice:
when oplock is detected and when oplock is processed. While the
1st attempt sets the level to the value provided by the server,
the 2nd one resets the level to None unconditionally.
This happens because the oplock/lease processing code was changed
to avoid races between page cache flushes and oplock breaks.
The commit c11f1df5003d534 ("cifs: Wait for writebacks to complete
before attempting write.") fixed the races for oplocks but didn't
apply the same changes for leases resulting in overwriting the
server granted value to None. Fix this by properly processing
lease breaks.
Signed-off-by: Pavel Shilovsky <pshilov@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com> CC: Stable <stable@vger.kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
If we don't find a writable file handle when retrying writepages
we break of the loop and do not unlock and put pages neither from
wdata2 nor from the original wdata. Fix this by walking through
all the remaining pages and cleanup them properly.
Cc: <stable@vger.kernel.org> Signed-off-by: Pavel Shilovsky <pshilov@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
Commit 5092fcf34908 ("crypto: arm64/aes-ce-ccm: add non-SIMD generic
fallback") introduced C fallback code to replace the NEON routines
when invoked from a context where the NEON is not available (i.e.,
from the context of a softirq taken while the NEON is already being
used in kernel process context)
Fix two logical flaws in the MAC calculation of the associated data.
The NEON MAC calculation routine fails to handle the case correctly
where there is some data in the buffer, and the input fills it up
exactly. In this case, we enter the loop at the end with w8 == 0,
while a negative value is assumed, and so the loop carries on until
the increment of the 32-bit counter wraps around, which is quite
obviously wrong.
So omit the loop altogether in this case, and exit right away.
Reported-by: Eric Biggers <ebiggers@kernel.org> Fixes: a3fd82105b9d1 ("arm64/crypto: AES in CCM mode using ARMv8 Crypto ...") Cc: stable@vger.kernel.org Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
The x86 MORUS implementations all fail the improved AEAD tests because
they produce the wrong result with some data layouts. The issue is that
they assume that if the skcipher_walk API gives 'nbytes' not aligned to
the walksize (a.k.a. walk.stride), then it is the end of the data. In
fact, this can happen before the end.
Also, when the CRYPTO_TFM_REQ_MAY_SLEEP flag is given, they can
incorrectly sleep in the skcipher_walk_*() functions while preemption
has been disabled by kernel_fpu_begin().
Fix these bugs.
Fixes: 56e8e57fc3a7 ("crypto: morus - Add common SIMD glue code for MORUS") Cc: <stable@vger.kernel.org> # v4.18+ Cc: Ondrej Mosnacek <omosnace@redhat.com> Signed-off-by: Eric Biggers <ebiggers@google.com> Reviewed-by: Ondrej Mosnacek <omosnace@redhat.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
gcmaes_crypt_by_sg() dereferences the NULL pointer returned by
scatterwalk_ffwd() when encrypting an empty plaintext and the source
scatterlist ends immediately after the associated data.
Fix it by only fast-forwarding to the src/dst data scatterlists if the
data length is nonzero.
This bug is reproduced by the "rfc4543(gcm(aes))" test vectors when run
with the new AEAD test manager.
Fixes: e845520707f8 ("crypto: aesni - Update aesni-intel_glue to use scatter/gather") Cc: <stable@vger.kernel.org> # v4.17+ Cc: Dave Watson <davejwatson@fb.com> Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
The x86 AEGIS implementations all fail the improved AEAD tests because
they produce the wrong result with some data layouts. The issue is that
they assume that if the skcipher_walk API gives 'nbytes' not aligned to
the walksize (a.k.a. walk.stride), then it is the end of the data. In
fact, this can happen before the end.
Also, when the CRYPTO_TFM_REQ_MAY_SLEEP flag is given, they can
incorrectly sleep in the skcipher_walk_*() functions while preemption
has been disabled by kernel_fpu_begin().
Fix these bugs.
Fixes: 1d373d4e8e15 ("crypto: x86 - Add optimized AEGIS implementations") Cc: <stable@vger.kernel.org> # v4.18+ Cc: Ondrej Mosnacek <omosnace@redhat.com> Signed-off-by: Eric Biggers <ebiggers@google.com> Reviewed-by: Ondrej Mosnacek <omosnace@redhat.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
Instantiating "cryptd(crc32c)" causes a crypto self-test failure because
the crypto_alloc_shash() in alg_test_crc32c() fails. This is because
cryptd(crc32c) is an ahash algorithm, not a shash algorithm; so it can
only be accessed through the ahash API, unlike shash algorithms which
can be accessed through both the ahash and shash APIs.
As the test is testing the shash descriptor format which is only
applicable to shash algorithms, skip it for ahash algorithms.
(Note that it's still important to fix crypto self-test failures even
for weird algorithm instantiations like cryptd(crc32c) that no one
would really use; in fips_enabled mode unprivileged users can use them
to panic the kernel, and also they prevent treating a crypto self-test
failure as a bug when fuzzing the kernel.)
Fixes: 8e3ee85e68c5 ("crypto: crc32c - Test descriptor context format") Cc: stable@vger.kernel.org Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
Some algorithms have a ->setkey() method that is not atomic, in the
sense that setting a key can fail after changes were already made to the
tfm context. In this case, if a key was already set the tfm can end up
in a state that corresponds to neither the old key nor the new key.
For example, in lrw.c, if gf128mul_init_64k_bbe() fails due to lack of
memory, then priv::table will be left NULL. After that, encryption with
that tfm will cause a NULL pointer dereference.
It's not feasible to make all ->setkey() methods atomic, especially ones
that have to key multiple sub-tfms. Therefore, make the crypto API set
CRYPTO_TFM_NEED_KEY if ->setkey() fails and the algorithm requires a
key, to prevent the tfm from being used until a new key is set.
[Cc stable mainly because when introducing the NEED_KEY flag I changed
AF_ALG to rely on it; and unlike in-kernel crypto API users, AF_ALG
previously didn't have this problem. So these "incompletely keyed"
states became theoretically accessible via AF_ALG -- though, the
opportunities for causing real mischief seem pretty limited.]
Fixes: f8d33fac8480 ("crypto: skcipher - prevent using skciphers without setting key") Cc: <stable@vger.kernel.org> # v4.16+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
The memcpy()s in the PCBC implementation use walk->iv as both the source
and destination, which has undefined behavior. These memcpy()'s are
actually unneeded, because walk->iv is already used to hold the previous
plaintext block XOR'd with the previous ciphertext block. Thus,
walk->iv is already updated to its final value.
The generic MORUS implementations all fail the improved AEAD tests
because they produce the wrong result with some data layouts. The issue
is that they assume that if the skcipher_walk API gives 'nbytes' not
aligned to the walksize (a.k.a. walk.stride), then it is the end of the
data. In fact, this can happen before the end. Fix them.
Fixes: 396be41f16fd ("crypto: morus - Add generic MORUS AEAD implementations") Cc: <stable@vger.kernel.org> # v4.18+ Cc: Ondrej Mosnacek <omosnace@redhat.com> Signed-off-by: Eric Biggers <ebiggers@google.com> Reviewed-by: Ondrej Mosnacek <omosnace@redhat.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
Some algorithms have a ->setkey() method that is not atomic, in the
sense that setting a key can fail after changes were already made to the
tfm context. In this case, if a key was already set the tfm can end up
in a state that corresponds to neither the old key nor the new key.
It's not feasible to make all ->setkey() methods atomic, especially ones
that have to key multiple sub-tfms. Therefore, make the crypto API set
CRYPTO_TFM_NEED_KEY if ->setkey() fails and the algorithm requires a
key, to prevent the tfm from being used until a new key is set.
Note: we can't set CRYPTO_TFM_NEED_KEY for OPTIONAL_KEY algorithms, so
->setkey() for those must nevertheless be atomic. That's fine for now
since only the crc32 and crc32c algorithms set OPTIONAL_KEY, and it's
not intended that OPTIONAL_KEY be used much.
[Cc stable mainly because when introducing the NEED_KEY flag I changed
AF_ALG to rely on it; and unlike in-kernel crypto API users, AF_ALG
previously didn't have this problem. So these "incompletely keyed"
states became theoretically accessible via AF_ALG -- though, the
opportunities for causing real mischief seem pretty limited.]
Fixes: 9fa68f620041 ("crypto: hash - prevent using keyed hashes without setting key") Cc: stable@vger.kernel.org Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
The SIMD routine ported from x86 used to have a special code path
for inputs < 16 bytes, which got lost somewhere along the way.
Instead, the current glue code aligns the input pointer to 16 bytes,
which is not really necessary on this architecture (although it
could be beneficial to performance to expose aligned data to the
the NEON routine), but this could result in inputs of less than
16 bytes to be passed in. This not only fails the new extended
tests that Eric has implemented, it also results in the code
reading past the end of the input, which could potentially result
in crashes when dealing with less than 16 bytes of input at the
end of a page which is followed by an unmapped page.
So update the glue code to only invoke the NEON routine if the
input is at least 16 bytes.
Reported-by: Eric Biggers <ebiggers@kernel.org> Reviewed-by: Eric Biggers <ebiggers@kernel.org> Fixes: 6ef5737f3931 ("crypto: arm64/crct10dif - port x86 SSE implementation to arm64") Cc: <stable@vger.kernel.org> # v4.10+ Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
The arm64 NEON bit-sliced implementation of AES-CTR fails the improved
skcipher tests because it sometimes produces the wrong ciphertext. The
bug is that the final keystream block isn't returned from the assembly
code when the number of non-final blocks is zero. This can happen if
the input data ends a few bytes after a page boundary. In this case the
last bytes get "encrypted" by XOR'ing them with uninitialized memory.
Fix the assembly code to return the final keystream block when needed.
Fixes: 88a3f582bea9 ("crypto: arm64/aes - don't use IV buffer to return final keystream block") Cc: <stable@vger.kernel.org> # v4.11+ Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org> Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
The SIMD routine ported from x86 used to have a special code path
for inputs < 16 bytes, which got lost somewhere along the way.
Instead, the current glue code aligns the input pointer to permit
the NEON routine to use special versions of the vld1 instructions
that assume 16 byte alignment, but this could result in inputs of
less than 16 bytes to be passed in. This not only fails the new
extended tests that Eric has implemented, it also results in the
code reading past the end of the input, which could potentially
result in crashes when dealing with less than 16 bytes of input
at the end of a page which is followed by an unmapped page.
So update the glue code to only invoke the NEON routine if the
input is at least 16 bytes.
Reported-by: Eric Biggers <ebiggers@kernel.org> Reviewed-by: Eric Biggers <ebiggers@kernel.org> Fixes: 1d481f1cd892 ("crypto: arm/crct10dif - port x86 SSE implementation to ARM") Cc: <stable@vger.kernel.org> # v4.10+ Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
The generic AEGIS implementations all fail the improved AEAD tests
because they produce the wrong result with some data layouts. The issue
is that they assume that if the skcipher_walk API gives 'nbytes' not
aligned to the walksize (a.k.a. walk.stride), then it is the end of the
data. In fact, this can happen before the end. Fix them.
Fixes: f606a88e5823 ("crypto: aegis - Add generic AEGIS AEAD implementations") Cc: <stable@vger.kernel.org> # v4.18+ Cc: Ondrej Mosnacek <omosnace@redhat.com> Signed-off-by: Eric Biggers <ebiggers@google.com> Reviewed-by: Ondrej Mosnacek <omosnace@redhat.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
Some algorithms have a ->setkey() method that is not atomic, in the
sense that setting a key can fail after changes were already made to the
tfm context. In this case, if a key was already set the tfm can end up
in a state that corresponds to neither the old key nor the new key.
For example, in gcm.c, if the kzalloc() fails due to lack of memory,
then the CTR part of GCM will have the new key but GHASH will not.
It's not feasible to make all ->setkey() methods atomic, especially ones
that have to key multiple sub-tfms. Therefore, make the crypto API set
CRYPTO_TFM_NEED_KEY if ->setkey() fails, to prevent the tfm from being
used until a new key is set.
[Cc stable mainly because when introducing the NEED_KEY flag I changed
AF_ALG to rely on it; and unlike in-kernel crypto API users, AF_ALG
previously didn't have this problem. So these "incompletely keyed"
states became theoretically accessible via AF_ALG -- though, the
opportunities for causing real mischief seem pretty limited.]
Fixes: dc26c17f743a ("crypto: aead - prevent using AEADs without setting key") Cc: <stable@vger.kernel.org> # v4.16+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
same story as with last May fixes in sysfs (7b745a4e4051
"unfuck sysfs_mount()"); new_sb is left uninitialized
in case of early errors in kernfs_mount_ns() and papering
over it by treating any error from kernfs_mount_ns() as
equivalent to !new_ns ends up conflating the cases when
objects had never been transferred to a superblock with
ones when that has happened and resulting new superblock
had been dropped. Easily fixed (same way as in sysfs
case). Additionally, there's a superblock leak on
kernfs_node_dentry() failure *and* a dentry leak inside
kernfs_node_dentry() itself - the latter on probably
impossible errors, but the former not impossible to trigger
(as the matter of fact, injecting allocation failures
at that point *does* trigger it).
Libnvdimm reserves the first 8K of pfn and devicedax namespaces to
store a superblock describing the namespace. This 8K reservation
is contained within the altmap area which the kernel uses for the
vmemmap backing for the pages within the namespace. The altmap
allows for some pages at the start of the altmap area to be reserved
and that mechanism is used to protect the superblock from being
re-used as vmemmap backing.
The number of PFNs to reserve is calculated using:
So on systems where PAGE_SIZE is greater than 8K the reservation
size is truncated to zero and the superblock area is re-used as
vmemmap backing. As a result all the namespace information stored
in the superblock (i.e. if it's a PFN or DAX namespace) is lost
and the namespace needs to be re-created to get access to the
contents.
This patch fixes this by using PFN_UP() rather than PHYS_PFN() to ensure
that at least one page is reserved. On systems with a 4K pages size this
patch should have no effect.
Cc: stable@vger.kernel.org Cc: Dan Williams <dan.j.williams@intel.com> Fixes: ac515c084be9 ("libnvdimm, pmem, pfn: move pfn setup to the core") Signed-off-by: Oliver O'Halloran <oohall@gmail.com> Reviewed-by: Vishal Verma <vishal.l.verma@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
For recovery, where non-dax access is needed to a given physical address
range, and testing, allow the 'force_raw' attribute to override the
default establishment of a dev_pagemap.
Otherwise without this capability it is possible to end up with a
namespace that can not be activated due to corrupted info-block, and one
that can not be repaired due to a section collision.
Cc: <stable@vger.kernel.org> Fixes: 004f1afbe199 ("libnvdimm, pmem: direct map legacy pmem by default") Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
When trying to see whether current nd_region intersects with others,
trim_pfn_device() has already calculated the *size* to be expanded to
SECTION size.
Do not double append 'adjust' to 'size' when calculating whether the end
of a region collides with the next pmem region.
Fixes: ae86cbfef381 "libnvdimm, pfn: Pad pfn namespaces relative to other regions" Cc: <stable@vger.kernel.org> Signed-off-by: Wei Yang <richardw.yang@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
The UEFI 2.7 specification sets expectations that the 'updating' flag is
eventually cleared. To date, the libnvdimm core has never adhered to
that protocol. The policy of the core matches the policy of other
multi-device info-block formats like MD-Software-RAID that expect
administrator intervention on inconsistent info-blocks, not automatic
invalidation.
However, some pre-boot environments may unfortunately attempt to "clean
up" the labels and invalidate a set when it fails to find at least one
"non-updating" label in the set. Clear the updating flag after set
updates to minimize the window of vulnerability to aggressive pre-boot
environments.
Ideally implementations would not write to the label area outside of
creating namespaces.
Note that this only minimizes the window, it does not close it as the
system can still crash while clearing the flag and the set can be
subsequently deleted / invalidated by the pre-boot environment.
Fixes: f524bf271a5c ("libnvdimm: write pmem label set") Cc: <stable@vger.kernel.org> Cc: Kelly Couch <kelly.j.couch@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
Commit 11189c1089da "acpi/nfit: Fix command-supported detection" broke
ND_CMD_CALL for bus-level commands. The "func = cmd" assumption is only
valid for:
The implementation is broken in all the ways the unit test did not touch:
1/ The local definition of in_buf and in_obj violated C99 initializer
expectations for zeroing. By only initializing 2 out of the three
struct members the compiler was free to zero-initialize the remaining
entry even though the aliased location in the union was initialized.
2/ The implementation made assumptions about the state of the 'smart'
payload after command execution that are satisfied by
acpi_nfit_ctl(), but not acpi_evaluate_dsm().
3/ populate_shutdown_status() is skipped on Intel NVDIMMs due to the early
return for skipping the common _LS{I,R,W} enabling.
4/ The input length should be zero.
This breakage was missed due to the unit test implementation only
testing the case where nfit_intel_shutdown_status() returns a valid
payload.
Much of this complexity would be saved if acpi_nfit_ctl() could be used, but
that currently requires a 'struct nvdimm *' argument and one is not created
until later in the init process. The health result is needed before the device
is created because the payload gates whether the nmemX/nfit/dirty_shutdown
property is visible in sysfs.
Cc: <stable@vger.kernel.org> Fixes: 0ead11181fe0 ("acpi, nfit: Collect shutdown status") Reported-by: Dexuan Cui <decui@microsoft.com> Reviewed-by: Dexuan Cui <decui@microsoft.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
The radix tree would rewind the index in an iterator to the lowest index
of a multi-slot entry. The XArray iterators instead leave the index
unchanged, but I overlooked that when converting DAX from the radix tree
to the XArray. Adjust the index that we use for flushing to the start
of the PMD range.
Fixes: c1901cd33cf4 ("page cache: Convert find_get_entries_tag to XArray") Cc: <stable@vger.kernel.org> Reported-by: Piotr Balcer <piotr.balcer@intel.com> Tested-by: Dan Williams <dan.j.williams@intel.com> Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Matthew Wilcox <willy@infradead.org> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
For chain mode in cipher(eg. AES-CBC/DES-CBC), the iv is continuously
updated in the operation. The new iv value should be written to device
register by software.
Hash algorithms with an alignmask set, e.g. "xcbc(aes-aesni)" and
"michael_mic", fail the improved hash tests because they sometimes
produce the wrong digest. The bug is that in the case where a
scatterlist element crosses pages, not all the data is actually hashed
because the scatterlist walk terminates too early. This happens because
the 'nbytes' variable in crypto_hash_walk_done() is assigned the number
of bytes remaining in the page, then later interpreted as the number of
bytes remaining in the scatterlist element. Fix it.
Fixes: 900a081f6912 ("crypto: ahash - Fix early termination in hash walk") Cc: stable@vger.kernel.org Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
1. It stored the per-request state 'cnt' in the tfm context, which can be
used by multiple threads concurrently (e.g. via AF_ALG).
2. It didn't support messages not a multiple of the block cipher size,
despite being a stream cipher.
3. It didn't set cra_blocksize to 1 to indicate it is a stream cipher.
To fix these, set the 'chunksize' property to the cipher block size to
guarantee that when walking through the scatterlist, a partial block can
only occur at the end. Then change the implementation to XOR a block at
a time at first, then XOR the partial block at the end if needed. This
is the same way CTR and CFB are implemented. As a bonus, this also
improves performance in most cases over the current approach.
The memcpy() in crypto_cfb_decrypt_inplace() uses walk->iv as both the
source and destination, which has undefined behavior. It is unneeded
because walk->iv is already used to hold the previous ciphertext block;
thus, walk->iv is already updated to its final value. So, remove it.
Also, note that in-place decryption is the only case where the previous
ciphertext block is not directly available. Therefore, as a related
cleanup I also updated crypto_cfb_encrypt_segment() to directly use the
previous ciphertext block rather than save it into walk->iv. This makes
it consistent with in-place encryption and out-of-place decryption; now
only in-place decryption is different, because it has to be.
Fixes: a7d85e06ed80 ("crypto: cfb - add support for Cipher FeedBack mode") Cc: <stable@vger.kernel.org> # v4.17+ Cc: James Bottomley <James.Bottomley@HansenPartnership.com> Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
Like some other block cipher mode implementations, the CFB
implementation assumes that while walking through the scatterlist, a
partial block does not occur until the end. But the walk is incorrectly
being done with a blocksize of 1, as 'cra_blocksize' is set to 1 (since
CFB is a stream cipher) but no 'chunksize' is set. This bug causes
incorrect encryption/decryption for some scatterlist layouts.
Fix it by setting the 'chunksize'. Also extend the CFB test vectors to
cover this bug as well as cases where the message length is not a
multiple of the block size.
Fixes: a7d85e06ed80 ("crypto: cfb - add support for Cipher FeedBack mode") Cc: <stable@vger.kernel.org> # v4.17+ Cc: James Bottomley <James.Bottomley@HansenPartnership.com> Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
For decryption in CBC mode we need to save the last ciphertext block
for use as the next IV. However, we were trying to do this also with
zero sized ciphertext resulting in a panic.
Fix this by only doing the copy if the ciphertext length is at least
of IV size.
We were copying the last ciphertext block into the IV field
for CBC before removing the DMA mapping of the output buffer
with the result of the buffer sometime being out-of-sync cache
wise and were getting intermittent cases of bad output IV.
Fix it by moving the DMA buffer unmapping before the copy.
In cc_unmap_aead_request(), call dma_pool_free() for mlli buffer only
if an item is allocated from the pool and not always if there is a
pool allocated.
This fixes a kernel panic when trying to free a non-allocated item.
Roland reports the following issue and provides a root cause analysis:
"On a v4.19 i.MX6 system with IMA and CONFIG_DMA_API_DEBUG enabled, a
warning is generated when accessing files on a filesystem for which IMA
measurement is enabled:
------------[ cut here ]------------
WARNING: CPU: 0 PID: 1 at kernel/dma/debug.c:1181 check_for_stack.part.9+0xd0/0x120
caam_jr 2101000.jr0: DMA-API: device driver maps memory from stack [addr=b668049e]
Modules linked in:
CPU: 0 PID: 1 Comm: switch_root Not tainted 4.19.0-20181214-1 #2
Hardware name: Freescale i.MX6 Quad/DualLite (Device Tree)
Backtrace:
[<c010efb8>] (dump_backtrace) from [<c010f2d0>] (show_stack+0x20/0x24)
[<c010f2b0>] (show_stack) from [<c08b04f4>] (dump_stack+0xa0/0xcc)
[<c08b0454>] (dump_stack) from [<c012b610>] (__warn+0xf0/0x108)
[<c012b520>] (__warn) from [<c012b680>] (warn_slowpath_fmt+0x58/0x74)
[<c012b62c>] (warn_slowpath_fmt) from [<c0199acc>] (check_for_stack.part.9+0xd0/0x120)
[<c01999fc>] (check_for_stack.part.9) from [<c019a040>] (debug_dma_map_page+0x144/0x174)
[<c0199efc>] (debug_dma_map_page) from [<c065f7f4>] (ahash_final_ctx+0x5b4/0xcf0)
[<c065f240>] (ahash_final_ctx) from [<c065b3c4>] (ahash_final+0x1c/0x20)
[<c065b3a8>] (ahash_final) from [<c03fe278>] (crypto_ahash_op+0x38/0x80)
[<c03fe240>] (crypto_ahash_op) from [<c03fe2e0>] (crypto_ahash_final+0x20/0x24)
[<c03fe2c0>] (crypto_ahash_final) from [<c03f19a8>] (ima_calc_file_hash+0x29c/0xa40)
[<c03f170c>] (ima_calc_file_hash) from [<c03f2b24>] (ima_collect_measurement+0x1dc/0x240)
[<c03f2948>] (ima_collect_measurement) from [<c03f0a60>] (process_measurement+0x4c4/0x6b8)
[<c03f059c>] (process_measurement) from [<c03f0cdc>] (ima_file_check+0x88/0xa4)
[<c03f0c54>] (ima_file_check) from [<c02d8adc>] (path_openat+0x5d8/0x1364)
[<c02d8504>] (path_openat) from [<c02dad24>] (do_filp_open+0x84/0xf0)
[<c02daca0>] (do_filp_open) from [<c02cf50c>] (do_open_execat+0x84/0x1b0)
[<c02cf488>] (do_open_execat) from [<c02d1058>] (__do_execve_file+0x43c/0x890)
[<c02d0c1c>] (__do_execve_file) from [<c02d1770>] (sys_execve+0x44/0x4c)
[<c02d172c>] (sys_execve) from [<c0101000>] (ret_fast_syscall+0x0/0x28)
---[ end trace 3455789a10e3aefd ]---
The cause is that the struct ahash_request *req is created as a
stack-local variable up in the stack (presumably somewhere in the IMA
implementation), then passed down into the CAAM driver, which tries to
dma_single_map the req->result (indirectly via map_seq_out_ptr_result)
in order to make that buffer available for the CAAM to store the result
of the following hash operation.
The calling code doesn't know how req will be used by the CAAM driver,
and there could be other such occurrences where stack memory is passed
down to the CAAM driver. Therefore we should rather fix this issue in
the CAAM driver where the requirements are known."
Fix this problem by:
-instructing the crypto engine to write the final hash in state->caam_ctx
-subsequently memcpy-ing the final hash into req->result
Cc: <stable@vger.kernel.org> # v4.19+ Reported-by: Roland Hieber <rhi@pengutronix.de> Signed-off-by: Horia Geantă <horia.geanta@nxp.com> Tested-by: Roland Hieber <rhi@pengutronix.de> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>