Dan Williams [Thu, 13 Jun 2019 22:56:30 +0000 (15:56 -0700)]
PCI/P2PDMA: track pgmap references per resource, not globally
In preparation for fixing a race between devm_memremap_pages_release()
and the final put of a page from the device-page-map, allocate a
percpu-ref per p2pdma resource mapping.
Dan Williams [Thu, 13 Jun 2019 22:56:27 +0000 (15:56 -0700)]
lib/genalloc: introduce chunk owners
The p2pdma facility enables a provider to publish a pool of dma
addresses for a consumer to allocate. A genpool is used internally by
p2pdma to collect dma resources, 'chunks', to be handed out to
consumers. Whenever a consumer allocates a resource it needs to pin the
'struct dev_pagemap' instance that backs the chunk selected by
pci_alloc_p2pmem().
Currently that reference is taken globally on the entire provider
device. That sets up a lifetime mismatch whereby the p2pdma core needs
to maintain hacks to make sure the percpu_ref is not released twice.
This lifetime mismatch also stands in the way of a fix to
devm_memremap_pages() whereby devm_memremap_pages_release() must wait for
the percpu_ref ->release() callback to complete before it can proceed to
teardown pages.
So, towards fixing this situation, introduce the ability to store a 'chunk
owner' at gen_pool_add() time, and a facility to retrieve the owner at
gen_pool_{alloc,free}() time. For p2pdma this will be used to store and
recall individual dev_pagemap reference counter instances per-chunk.
Dan Williams [Thu, 13 Jun 2019 22:56:24 +0000 (15:56 -0700)]
PCI/P2PDMA: fix the gen_pool_add_virt() failure path
The pci_p2pdma_add_resource() implementation immediately frees the pgmap
if gen_pool_add_virt() fails. However, that means that when @dev
triggers a devres release devm_memremap_pages_release() will crash
trying to access the freed @pgmap.
Use the new devm_memunmap_pages() to manually free the mapping in the
error path.
Patch series "mm/devm_memremap_pages: Fix page release race", v2.
Logan audited the devm_memremap_pages() shutdown path and noticed that
it was possible to proceed to arch_remove_memory() before all potential
page references have been reaped.
Introduce a new ->cleanup() callback to do the work of waiting for any
straggling page references and then perform the percpu_ref_exit() in
devm_memremap_pages_release() context.
For p2pdma this involves some deeper reworks to reference count
resources on a per-instance basis rather than a per pci-device basis. A
modified genalloc api is introduced to convey a driver-private pointer
through gen_pool_{alloc,free}() interfaces. Also, a
devm_memunmap_pages() api is introduced since p2pdma does not
auto-release resources on a setup failure.
The dax and pmem changes pass the nvdimm unit tests, and the p2pdma
changes should now pass testing with the pci_p2pdma_release() fix.
Jrme, how does this look for HMM?
This patch (of 6):
The devm_add_action() facility allows a resource allocation routine to
add custom devm semantics. One such user is devm_memremap_pages().
There is now a need to manually trigger
devm_memremap_pages_release(). Introduce devm_release_action() so the
release action can be triggered via a new devm_memunmap_pages() api in a
follow-on change.
Minchan Kim [Thu, 13 Jun 2019 22:56:15 +0000 (15:56 -0700)]
mm/vmscan.c: fix trying to reclaim unevictable LRU page
There was the below bug report from Wu Fangsuo.
On the CMA allocation path, isolate_migratepages_range() could isolate
unevictable LRU pages and reclaim_clean_page_from_list() can try to
reclaim them if they are clean file-backed pages.
page:ffffffbf02f33b40 count:86 mapcount:84 mapping:ffffffc08fa7a810 index:0x24
flags: 0x19040c(referenced|uptodate|arch_1|mappedtodisk|unevictable|mlocked)
raw: 000000000019040cffffffc08fa7a81000000000000000240000005600000053
raw: ffffffc009b05b20ffffffc009b05b200000000000000000ffffffc09bf3ee80
page dumped because: VM_BUG_ON_PAGE(PageLRU(page) || PageUnevictable(page))
page->mem_cgroup:ffffffc09bf3ee80
------------[ cut here ]------------
kernel BUG at /home/build/farmland/adroid9.0/kernel/linux/mm/vmscan.c:1350!
Internal error: Oops - BUG: 0 [#1] PREEMPT SMP
Modules linked in:
CPU: 0 PID: 7125 Comm: syz-executor Tainted: G S 4.14.81 #3
Hardware name: ASR AQUILAC EVB (DT)
task: ffffffc00a54cd00 task.stack: ffffffc009b00000
PC is at shrink_page_list+0x1998/0x3240
LR is at shrink_page_list+0x1998/0x3240
pc : [<ffffff90083a2158>] lr : [<ffffff90083a2158>] pstate: 60400045
sp : ffffffc009b05940
..
shrink_page_list+0x1998/0x3240
reclaim_clean_pages_from_list+0x3c0/0x4f0
alloc_contig_range+0x3bc/0x650
cma_alloc+0x214/0x668
ion_cma_allocate+0x98/0x1d8
ion_alloc+0x200/0x7e0
ion_ioctl+0x18c/0x378
do_vfs_ioctl+0x17c/0x1780
SyS_ioctl+0xac/0xc0
Wu found it's due to commit ad6b67041a45 ("mm: remove SWAP_MLOCK in
ttu"). Before that, unevictable pages go to cull_mlocked so that we
can't reach the VM_BUG_ON_PAGE line.
To fix the issue, this patch filters out unevictable LRU pages from the
reclaim_clean_pages_from_list in CMA.
Andrea Arcangeli [Thu, 13 Jun 2019 22:56:11 +0000 (15:56 -0700)]
coredump: fix race condition between collapse_huge_page() and core dumping
When fixing the race conditions between the coredump and the mmap_sem
holders outside the context of the process, we focused on
mmget_not_zero()/get_task_mm() callers in 04f5866e41fb70 ("coredump: fix
race condition between mmget_not_zero()/get_task_mm() and core
dumping"), but those aren't the only cases where the mmap_sem can be
taken outside of the context of the process as Michal Hocko noticed
while backporting that commit to older -stable kernels.
If mmgrab() is called in the context of the process, but then the
mm_count reference is transferred outside the context of the process,
that can also be a problem if the mmap_sem has to be taken for writing
through that mm_count reference.
khugepaged registration calls mmgrab() in the context of the process,
but the mmap_sem for writing is taken later in the context of the
khugepaged kernel thread.
collapse_huge_page() after taking the mmap_sem for writing doesn't
modify any vma, so it's not obvious that it could cause a problem to the
coredump, but it happens to modify the pmd in a way that breaks an
invariant that pmd_trans_huge_lock() relies upon. collapse_huge_page()
needs the mmap_sem for writing just to block concurrent page faults that
call pmd_trans_huge_lock().
Specifically the invariant that "!pmd_trans_huge()" cannot become a
"pmd_trans_huge()" doesn't hold while collapse_huge_page() runs.
The coredump will call __get_user_pages() without mmap_sem for reading,
which eventually can invoke a lockless page fault which will need a
functional pmd_trans_huge_lock().
So collapse_huge_page() needs to use mmget_still_valid() to check it's
not running concurrently with the coredump... as long as the coredump
can invoke page faults without holding the mmap_sem for reading.
This has "Fixes: khugepaged" to facilitate backporting, but in my view
it's more a bug in the coredump code that will eventually have to be
rewritten to stop invoking page faults without the mmap_sem for reading.
So the long term plan is still to drop all mmget_still_valid().
swkhack [Thu, 13 Jun 2019 22:56:08 +0000 (15:56 -0700)]
mm/mlock.c: change count_mm_mlocked_page_nr return type
On a 64-bit machine the value of "vma->vm_end - vma->vm_start" may be
negative when using 32 bit ints and the "count >> PAGE_SHIFT"'s result
will be wrong. So change the local variable and return value to
unsigned long to fix the problem.
Yang Shi [Thu, 13 Jun 2019 22:56:05 +0000 (15:56 -0700)]
mm: mmu_gather: remove __tlb_reset_range() for force flush
A few new fields were added to mmu_gather to make TLB flush smarter for
huge page by telling what level of page table is changed.
__tlb_reset_range() is used to reset all these page table state to
unchanged, which is called by TLB flush for parallel mapping changes for
the same range under non-exclusive lock (i.e. read mmap_sem).
Before commit dd2283f2605e ("mm: mmap: zap pages with read mmap_sem in
munmap"), the syscalls (e.g. MADV_DONTNEED, MADV_FREE) which may update
PTEs in parallel don't remove page tables. But, the forementioned
commit may do munmap() under read mmap_sem and free page tables. This
may result in program hang on aarch64 reported by Jan Stancek. The
problem could be reproduced by his test program with slightly modified
below.
---8<---
static int map_size = 4096;
static int num_iter = 500;
static long threads_total;
static void *distant_area;
void *map_write_unmap(void *ptr)
{
int *fd = ptr;
unsigned char *map_address;
int i, j = 0;
for (i = 0; i < num_iter; i++) {
map_address = mmap(distant_area, (size_t) map_size, PROT_WRITE | PROT_READ,
MAP_SHARED | MAP_ANONYMOUS, -1, 0);
if (map_address == MAP_FAILED) {
perror("mmap");
exit(1);
}
tlb_finish_mmu()
if (mm_tlb_flush_nested(tlb->mm))
__tlb_reset_range()
__tlb_reset_range() would reset freed_tables and cleared_* bits, but this
may cause inconsistency for munmap() which do free page tables. Then it
may result in some architectures, e.g. aarch64, may not flush TLB
completely as expected to have stale TLB entries remained.
Use fullmm flush since it yields much better performance on aarch64 and
non-fullmm doesn't yields significant difference on x86.
The original proposed fix came from Jan Stancek who mainly debugged this
issue, I just wrapped up everything together.
Jan's testing results:
v5.2-rc2-24-gbec7550cca10
--------------------------
mean stddev
real 37.382 2.780
user 1.420 0.078
sys 54.658 1.855
v5.2-rc2-24-gbec7550cca10 + "mm: mmu_gather: remove __tlb_reset_range() for force flush"
---------------------------------------------------------------------------------------_
mean stddev
real 37.119 2.105
user 1.548 0.087
sys 55.698 1.357
Wengang Wang [Thu, 13 Jun 2019 22:56:01 +0000 (15:56 -0700)]
fs/ocfs2: fix race in ocfs2_dentry_attach_lock()
ocfs2_dentry_attach_lock() can be executed in parallel threads against the
same dentry. Make that race safe. The race is like this:
thread A thread B
(A1) enter ocfs2_dentry_attach_lock,
seeing dentry->d_fsdata is NULL,
and no alias found by
ocfs2_find_local_alias, so kmalloc
a new ocfs2_dentry_lock structure
to local variable "dl", dl1
.....
(B1) enter ocfs2_dentry_attach_lock,
seeing dentry->d_fsdata is NULL,
and no alias found by
ocfs2_find_local_alias so kmalloc
a new ocfs2_dentry_lock structure
to local variable "dl", dl2.
......
(A2) set dentry->d_fsdata with dl1,
call ocfs2_dentry_lock() and increase
dl1->dl_lockres.l_ro_holders to 1 on
success.
......
(B2) set dentry->d_fsdata with dl2
call ocfs2_dentry_lock() and increase
dl2->dl_lockres.l_ro_holders to 1 on
success.
......
(A3) call ocfs2_dentry_unlock()
and decrease
dl2->dl_lockres.l_ro_holders to 0
on success.
....
(B3) call ocfs2_dentry_unlock(),
decreasing
dl2->dl_lockres.l_ro_holders, but
see it's zero now, panic
Kirill Tkhai [Thu, 13 Jun 2019 22:55:58 +0000 (15:55 -0700)]
mm/vmscan.c: fix recent_rotated history
Johannes pointed out that after commit 886cf1901db9 ("mm: move
recent_rotated pages calculation to shrink_inactive_list()") we lost all
zone_reclaim_stat::recent_rotated history.
Potyra, Stefan [Thu, 13 Jun 2019 22:55:55 +0000 (15:55 -0700)]
mm/mlock.c: mlockall error for flag MCL_ONFAULT
If mlockall() is called with only MCL_ONFAULT as flag, it removes any
previously applied lockings and does nothing else.
This behavior is counter-intuitive and doesn't match the Linux man page.
For mlockall():
EINVAL Unknown flags were specified or MCL_ONFAULT was specified
without either MCL_FUTURE or MCL_CURRENT.
Consequently, return the error EINVAL, if only MCL_ONFAULT is passed.
That way, applications will at least detect that they are calling
mlockall() incorrectly.
Manuel Traut [Thu, 13 Jun 2019 22:55:52 +0000 (15:55 -0700)]
scripts/decode_stacktrace.sh: prefix addr2line with $CROSS_COMPILE
At least for ARM64 kernels compiled with the crosstoolchain from
Debian/stretch or with the toolchain from kernel.org the line number is
not decoded correctly by 'decode_stacktrace.sh':
Johannes Weiner [Thu, 13 Jun 2019 22:55:46 +0000 (15:55 -0700)]
mm: memcontrol: don't batch updates of local VM stats and events
The kernel test robot noticed a 26% will-it-scale pagefault regression
from commit 42a300353577 ("mm: memcontrol: fix recursive statistics
correctness & scalabilty"). This appears to be caused by bouncing the
additional cachelines from the new hierarchical statistics counters.
We can fix this by getting rid of the batched local counters instead.
Originally, there were *only* group-local counters, and they were fully
maintained per cpu. A reader of a stats file high up in the cgroup tree
would have to walk the entire subtree and collect each level's per-cpu
counters to get the recursive view. This was prohibitively expensive,
and so we switched to per-cpu batched updates of the local counters
during a983b5ebee57 ("mm: memcontrol: fix excessive complexity in
memory.stat reporting"), reducing the complexity from nr_subgroups *
nr_cpus to nr_subgroups.
With growing machines and cgroup trees, the tree walk itself became too
expensive for monitoring top-level groups, and this is when the culprit
patch added hierarchy counters on each cgroup level. When the per-cpu
batch size would be reached, both the local and the hierarchy counters
would get batch-updated from the per-cpu delta simultaneously.
This makes local and hierarchical counter reads blazingly fast, but it
unfortunately makes the write-side too cache line intense.
Since local counter reads were never a problem - we only centralized
them to accelerate the hierarchy walk - and use of the local counters
are becoming rarer due to replacement with hierarchical views (ongoing
rework in the page reclaim and workingset code), we can make those local
counters unbatched per-cpu counters again.
The scheme will then be as such:
when a memcg statistic changes, the writer will:
- update the local counter (per-cpu)
- update the batch counter (per-cpu). If the batch is full:
- spill the batch into the group's atomic_t
- spill the batch into all ancestors' atomic_ts
- empty out the batch counter (per-cpu)
when a local memcg counter is read, the reader will:
- collect the local counter from all cpus
when a hiearchy memcg counter is read, the reader will:
- read the atomic_t
We might be able to simplify this further and make the recursive
counters unbatched per-cpu counters as well (batch upward propagation,
but leave per-cpu collection to the readers), but that will require a
more in-depth analysis and testing of all the callsites. Deal with the
immediate regression for now.
Commit d491f2b75237 ("PCI: PM: Avoid possible suspend-to-idle issue")
attempted to avoid a problem with devices whose drivers want them to
stay in D0 over suspend-to-idle and resume, but it did not go as far
as it should with that.
Namely, first of all, the power state of a PCI bridge with a
downstream device in D0 must be D0 (based on the PCI PM spec r1.2,
sec 6, table 6-1, if the bridge is not in D0, there can be no PCI
transactions on its secondary bus), but that is not actively enforced
during system-wide PM transitions, so use the skip_bus_pm flag
introduced by commit d491f2b75237 for that.
Second, the configuration of devices left in D0 (whatever the reason)
during suspend-to-idle need not be changed and attempting to put them
into D0 again by force is pointless, so explicitly avoid doing that.
Daniel Borkmann [Thu, 13 Jun 2019 21:07:01 +0000 (23:07 +0200)]
Merge branch 'bpf-ppc-div-fix'
Naveen N. Rao says:
====================
The first patch updates DIV64 overflow tests to properly detect error
conditions. The second patch fixes powerpc64 JIT to generate the proper
unsigned division instruction for BPF_ALU64.
====================
Naveen N. Rao [Wed, 12 Jun 2019 18:51:40 +0000 (00:21 +0530)]
powerpc/bpf: use unsigned division instruction for 64-bit operations
BPF_ALU64 div/mod operations are currently using signed division, unlike
BPF_ALU32 operations. Fix the same. DIV64 and MOD64 overflow tests pass
with this fix.
Fixes: 156d0e290e969c ("powerpc/ebpf/jit: Implement JIT compiler for extended BPF") Cc: [email protected] # v4.8+ Signed-off-by: Naveen N. Rao <[email protected]> Signed-off-by: Daniel Borkmann <[email protected]>
Naveen N. Rao [Wed, 12 Jun 2019 18:51:39 +0000 (00:21 +0530)]
bpf: fix div64 overflow tests to properly detect errors
If the result of the division is LLONG_MIN, current tests do not detect
the error since the return value is truncated to a 32-bit value and ends
up being 0.
bpf: simplify definition of BPF_FIB_LOOKUP related flags
Previously, the BPF_FIB_LOOKUP_{DIRECT,OUTPUT} flags in the BPF UAPI
were defined with the help of BIT macro. This had the following issues:
- In order to use any of the flags, a user was required to depend
on <linux/bits.h>.
- No other flag in bpf.h uses the macro, so it seems that an unwritten
convention is to use (1 << (nr)) to define BPF-related flags.
Fixes: 87f5fc7e48dd ("bpf: Provide helper to do forwarding lookups in kernel FIB table") Signed-off-by: Martynas Pumputis <[email protected]> Acked-by: Andrii Nakryiko <[email protected]> Signed-off-by: Daniel Borkmann <[email protected]>
We need to protect so that the call to smb2_reconnect() in
smb2_reconnect_server() does not end up freeing the session
because it can lead to a use after free and crash.
Thomas Gleixner [Thu, 13 Jun 2019 16:34:54 +0000 (18:34 +0200)]
Merge tag 'timers-v5.2-rc1' of https://git.linaro.org/people/daniel.lezcano/linux into timers/urgent
Pull timer fixes from Daniel Lezcano:
- Fix missing notrace leading to deadlock on arch_arm_timer (Julien Thierry)
- Fix compilation warning on timer-ti-dm (Philippe Mazenauer)
Linus Torvalds [Thu, 13 Jun 2019 15:59:05 +0000 (05:59 -1000)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/hid/hid
Pull HID fixes from Jiri Kosina:
- regression fixes (reverts) for module loading changes that turned out
to be incompatible with some userspace, from Benjamin Tissoires
- regression fix for special Logitech unifiying receiver 0xc52f, from
Hans de Goede
- a few device ID additions to logitech driver, from Hans de Goede
- fix for Bluetooth support on 2nd-gen Wacom Intuos Pro, from Jason
Gerecke
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/hid/hid:
HID: logitech-dj: Fix 064d:c52f receiver support
Revert "HID: core: Call request_module before doing device_add"
Revert "HID: core: Do not call request_module() in async context"
Revert "HID: Increase maximum report size allowed by hid_field_extract()"
HID: a4tech: fix horizontal scrolling
HID: hyperv: Add a module description line
HID: logitech-hidpp: Add support for the S510 remote control
HID: multitouch: handle faulty Elo touch device
HID: wacom: Sync INTUOSP2_BT touch state after each frame if necessary
HID: wacom: Correct button numbering 2nd-gen Intuos Pro over Bluetooth
HID: wacom: Send BTN_TOUCH in response to INTUOSP2_BT eraser contact
HID: wacom: Don't report anything prior to the tool entering range
HID: wacom: Don't set tool type until we're in range
HID: rmi: Use SET_REPORT request on control endpoint for Acer Switch 3 and 5
HID: logitech-hidpp: add support for the MX5500 keyboard
HID: logitech-dj: add support for the Logitech MX5500's Bluetooth Mini-Receiver
HID: i2c-hid: add iBall Aer3 to descriptor override
Takashi Iwai [Thu, 13 Jun 2019 15:33:34 +0000 (17:33 +0200)]
Merge tag 'asoc-fix-v5.2-rc4' of https://git.kernel.org/pub/scm/linux/kernel/git/broonie/sound into for-linus
ASoC: Fixes for v5.2
There's an awful lot of fixes here, almost all for the newly introduced
SoF DSP drivers (including a few things it turned up in shared code).
This is a large and complex piece of code so it's not surprising that
there have been quite a few issues here, fortunately things seem to have
mostly calmed down now. Otherwise there's just a smattering of small fixes.
Vitaly Kuznetsov [Thu, 13 Jun 2019 11:35:02 +0000 (13:35 +0200)]
KVM: nVMX: use correct clean fields when copying from eVMCS
Unfortunately, a couple of mistakes were made while implementing
Enlightened VMCS support, in particular, wrong clean fields were
used in copy_enlightened_to_vmcs12():
- exception_bitmap is covered by CONTROL_EXCPN;
- vm_exit_controls/pin_based_vm_exec_control/secondary_vm_exec_control
are covered by CONTROL_GRP1.
block/ps3vram: Use %llu to format sector_t after LBDAF removal
The removal of CONFIG_LBDAF changed the type of sector_t from "unsigned
long" to "u64" aka "unsigned long long" on 64-bit platforms, leading to
a compiler warning regression:
drivers/block/ps3vram.c: In function ‘ps3vram_probe’:
drivers/block/ps3vram.c:770:23: warning: format ‘%lu’ expects argument of type ‘long unsigned int’, but argument 4 has type ‘sector_t {aka long long unsigned int}’ [-Wformat=]
Hans de Goede [Tue, 11 Jun 2019 14:32:59 +0000 (16:32 +0200)]
libata: Extend quirks for the ST1000LM024 drives with NOLPM quirk
We've received a bugreport that using LPM with ST1000LM024 drives leads
to system lockups. So it seems that these models are buggy in more then
1 way. Add NOLPM quirk to the existing quirks entry for BROKEN_FPDMA_AA.
Coly Li [Sun, 9 Jun 2019 22:13:35 +0000 (06:13 +0800)]
bcache: only set BCACHE_DEV_WB_RUNNING when cached device attached
When people set a writeback percent via sysfs file,
/sys/block/bcache<N>/bcache/writeback_percent
current code directly sets BCACHE_DEV_WB_RUNNING to dc->disk.flags
and schedules kworker dc->writeback_rate_update.
If there is no cache set attached to, the writeback kernel thread is
not running indeed, running dc->writeback_rate_update does not make
sense and may cause NULL pointer deference when reference cache set
pointer inside update_writeback_rate().
This patch checks whether the cache set point (dc->disk.c) is NULL in
sysfs interface handler, and only set BCACHE_DEV_WB_RUNNING and
schedule dc->writeback_rate_update when dc->disk.c is not NULL (it
means the cache device is attached to a cache set).
This problem might be introduced from initial bcache commit, but
commit 3fd47bfe55b0 ("bcache: stop dc->writeback_rate_update properly")
changes part of the original code piece, so I add 'Fixes: 3fd47bfe55b0'
to indicate from which commit this patch can be applied.
Coly Li [Sun, 9 Jun 2019 22:13:34 +0000 (06:13 +0800)]
bcache: fix stack corruption by PRECEDING_KEY()
Recently people report bcache code compiled with gcc9 is broken, one of
the buggy behavior I observe is that two adjacent 4KB I/Os should merge
into one but they don't. Finally it turns out to be a stack corruption
caused by macro PRECEDING_KEY().
See how PRECEDING_KEY() is defined in bset.h,
437 #define PRECEDING_KEY(_k) \
438 ({ \
439 struct bkey *_ret = NULL; \
440 \
441 if (KEY_INODE(_k) || KEY_OFFSET(_k)) { \
442 _ret = &KEY(KEY_INODE(_k), KEY_OFFSET(_k), 0); \
443 \
444 if (!_ret->low) \
445 _ret->high--; \
446 _ret->low--; \
447 } \
448 \
449 _ret; \
450 })
At line 442, _ret points to address of a on-stack variable combined by
KEY(), the life range of this on-stack variable is in line 442-446,
once _ret is returned to bch_btree_insert_key(), the returned address
points to an invalid stack address and this address is overwritten in
the following called bch_btree_iter_init(). Then argument 'search' of
bch_btree_iter_init() points to some address inside stackframe of
bch_btree_iter_init(), exact address depends on how the compiler
allocates stack space. Now the stack is corrupted.
The in-memory representation of SVE and FPSIMD registers is
different: the FPSIMD V-registers are stored as single 128-bit
host-endian values, whereas SVE registers are stored in an
endianness-invariant byte order.
This means that the two representations differ when running on a
big-endian host. But we blindly copy data from one representation
to another when converting between the two, resulting in the
register contents being unintentionally byteswapped in certain
situations. Currently this can be triggered by the first SVE
instruction after a syscall, for example (though the potential
trigger points may vary in future).
So, fix the conversion functions fpsimd_to_sve(), sve_to_fpsimd()
and sve_sync_from_fpsimd_zeropad() to swab where appropriate.
There is no common swahl128() or swab128() that we could use here.
Maybe it would be worth making this generic, but for now add a
simple local hack.
Since the byte order differences are exposed in ABI, also clarify
the documentation.
Cc: Alex Bennée <[email protected]> Cc: Peter Maydell <[email protected]> Cc: Alan Hayward <[email protected]> Cc: Julien Grall <[email protected]> Fixes: bc0ee4760364 ("arm64/sve: Core task context handling") Fixes: 8cd969d28fd2 ("arm64/sve: Signal handling support") Fixes: 43d4da2c45b2 ("arm64/sve: ptrace and ELF coredump support") Signed-off-by: Dave Martin <[email protected]>
[will: Fix typos in comments and docs spotted by Julien] Signed-off-by: Will Deacon <[email protected]>
Ming Lei [Tue, 11 Jun 2019 09:31:53 +0000 (17:31 +0800)]
blk-mq: remove WARN_ON(!q->elevator) from blk_mq_sched_free_requests
blk_mq_sched_free_requests() may be called in failure path in which
q->elevator may not be setup yet, so remove WARN_ON(!q->elevator) from
blk_mq_sched_free_requests for avoiding the false positive.
This function is actually safe to call in case of !q->elevator because
hctx->sched_tags is checked.
This patch removes the check in the null_blk_zoned for report zone
command, where it checks for the dev-,>zoned before executing the report
zone.
The null_zone_report() function is a block_device operation callback
which is initialized in the null_blk_main.c and gets called as a part
of blkdev for report zone IOCTL (BLKREPORTZONE).
The null_zone_report() will never get executed on the non-zoned block
device, in the non zoned block device blk_queue_is_zoned() will always
be false which is first check the blkdev_report_zones_ioctl()
before actual low level driver report zone callback is executed.
Here is the detailed scenario:-
1. modprobe null_blk
null_init
null_alloc_dev
dev->zoned = 0
null_add_dev
dev->zoned == 0
so we don't set the q->limits.zoned = BLK_ZONED_HR
blk-mq: no need to check return value of debugfs_create functions
When calling debugfs functions, there is no need to ever check the
return value. The function can work or not, but the code logic should
never do something different based on this.
When all of these checks are cleaned up, lots of the functions used in
the blk-mq-debugfs code can now return void, as no need to check the
return value of them either.
Overall, this ends up cleaning up the code and making it smaller, always
a nice win.
Eric Biggers [Wed, 12 Jun 2019 21:58:43 +0000 (14:58 -0700)]
io_uring: fix memory leak of UNIX domain socket inode
Opening and closing an io_uring instance leaks a UNIX domain socket
inode. This is because the ->file of the io_uring instance's internal
UNIX domain socket is set to point to the io_uring file, but then
sock_release() sees the non-NULL ->file and assumes the inode reference
is held by the file so doesn't call iput(). That's not the case here,
since the reference is still meant to be held by the socket; the actual
inode of the io_uring file is different.
Fix this leak by NULL-ing out ->file before releasing the socket.
Damien Le Moal [Tue, 4 Jun 2019 07:23:40 +0000 (16:23 +0900)]
block: force select mq-deadline for zoned block devices
In most use cases of zoned block devices (aka SMR disks), the
mq-deadline scheduler is mandatory as it implements sequential write
command processing guarantees with zone write locking. So make sure that
this scheduler is always enabled if CONFIG_BLK_DEV_ZONED is selected.
Todd Kjos [Wed, 12 Jun 2019 20:29:27 +0000 (13:29 -0700)]
binder: fix possible UAF when freeing buffer
There is a race between the binder driver cleaning
up a completed transaction via binder_free_transaction()
and a user calling binder_ioctl(BC_FREE_BUFFER) to
release a buffer. It doesn't matter which is first but
they need to be protected against running concurrently
which can result in a UAF.
Michael Ellerman [Thu, 13 Jun 2019 05:00:34 +0000 (15:00 +1000)]
Merge branch 'context-id-fix' into fixes
This merges a fix for a bug in our context id handling on 64-bit hash
CPUs.
The fix was written against v5.1 to ease backporting to stable
releases. Here we are merging it up to a v5.2-rc2 base, which involves
a bit of manual resolution.
Linus Torvalds [Thu, 13 Jun 2019 02:10:57 +0000 (16:10 -1000)]
Merge tag 'selinux-pr-20190612' of git://git.kernel.org/pub/scm/linux/kernel/git/pcmoore/selinux
Pull selinux fixes from Paul Moore:
"Three patches for v5.2.
One fixes a problem where we weren't correctly logging raw SELinux
labels, the other two fix problems where we weren't properly checking
calls to kmemdup()"
* tag 'selinux-pr-20190612' of git://git.kernel.org/pub/scm/linux/kernel/git/pcmoore/selinux:
selinux: fix a missing-check bug in selinux_sb_eat_lsm_opts()
selinux: fix a missing-check bug in selinux_add_mnt_opt( )
selinux: log raw contexts as untrusted strings
Michael Ellerman [Thu, 13 Jun 2019 02:07:59 +0000 (12:07 +1000)]
selftests/powerpc: Add test of fork with mapping above 512TB
This tests that when a process with a mapping above 512TB forks we
correctly separate the parent and child address spaces. This exercises
the bug in the context id handling fixed in the previous commit.
Dan Carpenter [Sat, 8 Jun 2019 09:23:57 +0000 (12:23 +0300)]
drm/amdgpu: Fix bounds checking in amdgpu_ras_is_supported()
The "block" variable can be set by the user through debugfs, so it can
be quite large which leads to shift wrapping here. This means we report
a "block" as supported when it's not, and that leads to array overflows
later on.
This bug is not really a security issue in real life, because debugfs is
generally root only.
Fixes: 36ea1bd2d084 ("drm/amdgpu: add debugfs ctrl node") Signed-off-by: Dan Carpenter <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
* tag 'clk-meson-5.2-1-fixes' of https://github.com/BayLibre/clk-meson:
clk: meson: meson8b: fix a typo in the VPU parent names array variable
clk: meson: fix MPLL 50M binding id typo
====================
net: mvpp2: prs: Fixes for VID filtering
This series fixes some issues with VID filtering offload, mainly due to
the wrong ranges being used in the TCAM header parser.
The first patch fixes a bug where removing a VLAN from a port's
whitelist would also remove it from other port's, if they are on the
same PPv2 instance.
The second patch makes so that we don't invalidate the wrong TCAM
entries when clearing the whole whitelist.
====================
net: mvpp2: prs: Use the correct helpers when removing all VID filters
When removing all VID filters, the mvpp2_prs_vid_entry_remove would be
called with the TCAM id incorrectly used as a VID, causing the wrong
TCAM entries to be invalidated.
Fix this by directly invalidating entries in the VID range.
Fixes: 56beda3db602 ("net: mvpp2: Add hardware offloading for VLAN filtering") Suggested-by: Yuri Chipchev <[email protected]> Signed-off-by: Maxime Chevallier <[email protected]> Signed-off-by: David S. Miller <[email protected]>
David S. Miller [Wed, 12 Jun 2019 18:08:15 +0000 (11:08 -0700)]
Merge branch 'mlxsw-Various-fixes'
Ido Schimmel says:
====================
mlxsw: Various fixes
This patchset contains various fixes for mlxsw.
Patch #1 fixes an hash polarization problem when a nexthop device is a
LAG device. This is caused by the fact that the same seed is used for
the LAG and ECMP hash functions.
Patch #2 fixes an issue in which the driver fails to refresh a nexthop
neighbour after it becomes dead. This prevents the nexthop from ever
being written to the adjacency table and used to forward traffic. Patch
Patch #4 fixes a wrong extraction of TOS value in flower offload code.
Patch #5 is a test case.
Patch #6 works around a buffer issue in Spectrum-2 by reducing the
default sizes of the shared buffer pools.
Patch #7 prevents prio-tagged packets from entering the switch when PVID
is removed from the bridge port.
Please consider patches #2, #4 and #6 for 5.1.y
====================
Petr Machata [Tue, 11 Jun 2019 07:19:45 +0000 (10:19 +0300)]
mlxsw: spectrum_buffers: Reduce pool size on Spectrum-2
Due to an issue on Spectrum-2, in front-panel ports split four ways, 2 out
of 32 port buffers cannot be used. To work around this, the next FW release
will mark them as unused, and will report correspondingly lower total
shared buffer size. mlxsw will pick up the new value through a query to
cap_total_buffer_size resource. However the initial size for shared buffer
pool 0 is hard-coded and therefore needs to be updated.
Thus reduce the pool size by 2.7 MiB (which corresponds to 2/32 of the
total size of 42 MiB), and round down to the whole number of cells.
Fixes: fe099bf682ab ("mlxsw: spectrum_buffers: Add Spectrum-2 shared buffer configuration") Signed-off-by: Petr Machata <[email protected]> Acked-by: Jiri Pirko <[email protected]> Signed-off-by: Ido Schimmel <[email protected]> Signed-off-by: David S. Miller <[email protected]>
Ido Schimmel [Tue, 11 Jun 2019 07:19:40 +0000 (10:19 +0300)]
mlxsw: spectrum: Use different seeds for ECMP and LAG hash
The same hash function and seed are used for both ECMP and LAG hash.
Therefore, when a LAG device is used as a nexthop device as part of an
ECMP group, hash polarization can occur and all the traffic will be
hashed to a single LAG slave.
Fix this by using a different seed for the LAG hash.
John Fastabend [Wed, 12 Jun 2019 17:23:57 +0000 (17:23 +0000)]
net: tls, correctly account for copied bytes with multiple sk_msgs
tls_sw_do_sendpage needs to return the total number of bytes sent
regardless of how many sk_msgs are allocated. Unfortunately, copied
(the value we return up the stack) is zero'd before each new sk_msg
is allocated so we only return the copied size of the last sk_msg used.
The caller (splice, etc.) of sendpage will then believe only part
of its data was sent and send the missing chunks again. However,
because the data actually was sent the receiver will get multiple
copies of the same data.
To reproduce this do multiple sendfile calls with a length close to
the max record size. This will in turn call splice/sendpage, sendpage
may use multiple sk_msg in this case and then returns the incorrect
number of bytes. This will cause splice to resend creating duplicate
data on the receiver. Andre created a C program that can easily
generate this case so we will push a similar selftest for this to
bpf-next shortly.
The fix is to _not_ zero the copied field so that the total sent
bytes is returned.
Get the ingress interface and increment ICMP counters based on that
instead of skb->dev when the the dev is a VRF device.
This is a follow up on the following message:
https://www.spinics.net/lists/netdev/msg560268.html
v2: Avoid changing skb->dev since it has unintended effect for local
delivery (David Ahern). Signed-off-by: Stephen Suryaputra <[email protected]> Reviewed-by: David Ahern <[email protected]> Signed-off-by: David S. Miller <[email protected]>
Joel Savitz [Wed, 12 Jun 2019 15:50:48 +0000 (11:50 -0400)]
cpuset: restore sanity to cpuset_cpus_allowed_fallback()
In the case that a process is constrained by taskset(1) (i.e.
sched_setaffinity(2)) to a subset of available cpus, and all of those are
subsequently offlined, the scheduler will set tsk->cpus_allowed to
the current value of task_cs(tsk)->effective_cpus.
This is done via a call to do_set_cpus_allowed() in the context of
cpuset_cpus_allowed_fallback() made by the scheduler when this case is
detected. This is the only call made to cpuset_cpus_allowed_fallback()
in the latest mainline kernel.
However, this is not sane behavior.
I will demonstrate this on a system running the latest upstream kernel
with the following initial configuration:
# grep -i cpu /proc/$$/status
Cpus_allowed: ffffffff,fffffff
Cpus_allowed_list: 0-63
(Where cpus 32-63 are provided via smt.)
If we limit our current shell process to cpu2 only and then offline it
and reonline it:
# taskset -p 4 $$
pid 2272's current affinity mask: ffffffffffffffff
pid 2272's new affinity mask: 4
# echo off > /sys/devices/system/cpu/cpu2/online
# dmesg | tail -3
[ 2195.866089] process 2272 (bash) no longer affine to cpu2
[ 2195.872700] IRQ 114: no longer affine to CPU2
[ 2195.879128] smpboot: CPU 2 is now offline
We see that our current process now has an affinity mask containing
every cpu available on the system _except_ the one we originally
constrained it to:
# grep -i cpu /proc/$$/status
Cpus_allowed: ffffffff,fffffffb
Cpus_allowed_list: 0-1,3-63
This is not sane behavior, as the scheduler can now not only place the
process on previously forbidden cpus, it can't even schedule it on
the cpu it was originally constrained to!
Other cases result in even more exotic affinity masks. Take for instance
a process with an affinity mask containing only cpus provided by smt at
the moment that smt is toggled, in a configuration such as the following:
A double toggle of smt results in the following behavior:
# echo off > /sys/devices/system/cpu/smt/control
# echo on > /sys/devices/system/cpu/smt/control
# grep -i cpus /proc/$$/status
Cpus_allowed: ffffff00,ffffffff
Cpus_allowed_list: 0-31,40-63
This is even less sane than the previous case, as the new affinity mask
excludes all smt-provided cpus with ids less than those that were
previously in the affinity mask, as well as those that were actually in
the mask.
With this patch applied, both of these cases end in the following state:
# grep -i cpu /proc/$$/status
Cpus_allowed: ffffffff,ffffffff
Cpus_allowed_list: 0-63
The original policy is discarded. Though not ideal, it is the simplest way
to restore sanity to this fallback case without reinventing the cpuset
wheel that rolls down the kernel just fine in cgroup v2. A user who wishes
for the previous affinity mask to be restored in this fallback case can use
that mechanism instead.
This patch modifies scheduler behavior by instead resetting the mask to
task_cs(tsk)->cpus_allowed by default, and cpu_possible mask in legacy
mode. I tested the cases above on both modes.
Note that the scheduler uses this fallback mechanism if and only if
_every_ other valid avenue has been traveled, and it is the last resort
before calling BUG().
Using ethtool, users can specify a classification action matching on the
full vlan tag, which includes the DEI bit (also previously called CFI).
However, when converting the ethool_flow_spec to a flow_rule, we use
dissector keys to represent the matching patterns.
Since the vlan dissector key doesn't include the DEI bit, this
information was silently discarded when translating the ethtool
flow spec in to a flow_rule.
This commit adds the DEI bit into the vlan dissector key, and allows
propagating the information to the driver when parsing the ethtool flow
spec.
Fixes: eca4205f9ec3 ("ethtool: add ethtool_rx_flow_spec to flow_rule structure translator") Reported-by: Michał Mirosław <[email protected]> Signed-off-by: Maxime Chevallier <[email protected]> Signed-off-by: David S. Miller <[email protected]>
Matteo Croce [Wed, 12 Jun 2019 09:50:37 +0000 (11:50 +0200)]
mpls: fix af_mpls dependencies for real
Randy reported that selecting MPLS_ROUTING without PROC_FS breaks
the build, because since commit c1a9d65954c6 ("mpls: fix af_mpls
dependencies"), MPLS_ROUTING selects PROC_SYSCTL, but Kconfig's select
doesn't recursively handle dependencies.
Change the select into a dependency.
Gen Zhang [Wed, 12 Jun 2019 13:55:38 +0000 (21:55 +0800)]
selinux: fix a missing-check bug in selinux_sb_eat_lsm_opts()
In selinux_sb_eat_lsm_opts(), 'arg' is allocated by kmemdup_nul(). It
returns NULL when fails. So 'arg' should be checked. And 'mnt_opts'
should be freed when error.
Linus Torvalds [Wed, 12 Jun 2019 15:57:05 +0000 (05:57 -1000)]
Merge tag 'media/v5.2-2' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media
Pull media fixes from Mauro Carvalho Chehab:
- a debug warning for satellite tuning at dvb core was producing too
much noise
- a regression at hfi_parser on Venus driver
* tag 'media/v5.2-2' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media:
media: venus: hfi_parser: fix a regression in parser
media: dvb: warning about dvb frequency limits produces too much noise
Gen Zhang [Wed, 12 Jun 2019 13:28:21 +0000 (21:28 +0800)]
selinux: fix a missing-check bug in selinux_add_mnt_opt( )
In selinux_add_mnt_opt(), 'val' is allocated by kmemdup_nul(). It returns
NULL when fails. So 'val' should be checked. And 'mnt_opts' should be
freed when error.
Signed-off-by: Gen Zhang <[email protected]> Fixes: 757cbe597fe8 ("LSM: new method: ->sb_add_mnt_opt()") Cc: <[email protected]>
[PM: fixed some indenting problems] Signed-off-by: Paul Moore <[email protected]>
Will Deacon [Tue, 11 Jun 2019 11:47:34 +0000 (12:47 +0100)]
arm64: tlbflush: Ensure start/end of address range are aligned to stride
Since commit 3d65b6bbc01e ("arm64: tlbi: Set MAX_TLBI_OPS to
PTRS_PER_PTE"), we resort to per-ASID invalidation when attempting to
perform more than PTRS_PER_PTE invalidation instructions in a single
call to __flush_tlb_range(). Whilst this is beneficial, the mmu_gather
code does not ensure that the end address of the range is rounded-up
to the stride when freeing intermediate page tables in pXX_free_tlb(),
which defeats our range checking.
Dave Martin [Thu, 6 Jun 2019 10:58:07 +0000 (11:58 +0100)]
KVM: arm/arm64: vgic: Fix kvm_device leak in vgic_its_destroy
kvm_device->destroy() seems to be supposed to free its kvm_device
struct, but vgic_its_destroy() is not currently doing this,
resulting in a memory leak, resulting in kmemleak reports such as
the following:
Cc: Andre Przywara <[email protected]> Fixes: 1085fdc68c60 ("KVM: arm64: vgic-its: Introduce new KVM ITS device") Signed-off-by: Dave Martin <[email protected]> Signed-off-by: Marc Zyngier <[email protected]>
Dave Martin [Wed, 12 Jun 2019 12:44:49 +0000 (13:44 +0100)]
KVM: arm64: Filter out invalid core register IDs in KVM_GET_REG_LIST
Since commit d26c25a9d19b ("arm64: KVM: Tighten guest core register
access from userspace"), KVM_{GET,SET}_ONE_REG rejects register IDs
that do not correspond to a single underlying architectural register.
KVM_GET_REG_LIST was not changed to match however: instead, it
simply yields a list of 32-bit register IDs that together cover the
whole kvm_regs struct. This means that if userspace tries to use
the resulting list of IDs directly to drive calls to KVM_*_ONE_REG,
some of those calls will now fail.
This was not the intention. Instead, iterating KVM_*_ONE_REG over
the list of IDs returned by KVM_GET_REG_LIST should be guaranteed
to work.
This patch fixes the problem by splitting validate_core_offset()
into a backend core_reg_size_from_offset() which does all of the
work except for checking that the size field in the register ID
matches, and kvm_arm_copy_reg_indices() and num_core_regs() are
converted to use this to enumerate the valid offsets.
kvm_arm_copy_reg_indices() now also sets the register ID size field
appropriately based on the value returned, so the register ID
supplied to userspace is fully qualified for use with the register
access ioctls.
Ilya Maximets [Fri, 7 Jun 2019 17:27:32 +0000 (20:27 +0300)]
xdp: check device pointer before clearing
We should not call 'ndo_bpf()' or 'dev_put()' with NULL argument.
Fixes: c9b47cc1fabc ("xsk: fix bug when trying to use both copy and zero-copy on one queue id") Signed-off-by: Ilya Maximets <[email protected]> Acked-by: Jonathan Lemon <[email protected]> Signed-off-by: Daniel Borkmann <[email protected]>
Martin KaFai Lau [Tue, 11 Jun 2019 21:45:57 +0000 (14:45 -0700)]
bpf: net: Set sk_bpf_storage back to NULL for cloned sk
The cloned sk should not carry its parent-listener's sk_bpf_storage.
This patch fixes it by setting it back to NULL.
Fixes: 6ac99e8f23d4 ("bpf: Introduce bpf sk local storage") Signed-off-by: Martin KaFai Lau <[email protected]> Acked-by: Andrii Nakryiko <[email protected]> Signed-off-by: Daniel Borkmann <[email protected]>
Filipe Manana [Wed, 12 Jun 2019 10:05:42 +0000 (11:05 +0100)]
Btrfs: fix race between block group removal and block group allocation
If a task is removing the block group that currently has the highest start
offset amongst all existing block groups, there is a short time window
where it races with a concurrent block group allocation, resulting in a
transaction abort with an error code of EEXIST.
The following diagram explains the race in detail:
Task A Task B
btrfs_remove_block_group(bg offset X)
remove_extent_mapping(em offset X)
-> removes extent map X from the
tree of extent maps
(fs_info->mapping_tree), so the
next call to find_next_chunk()
will return offset X
btrfs_alloc_chunk()
find_next_chunk()
--> returns offset X
__btrfs_alloc_chunk(offset X)
btrfs_make_block_group()
btrfs_create_block_group_cache()
--> creates btrfs_block_group_cache
object with a key corresponding
to the block group item in the
extent, the key is:
(offset X, BTRFS_BLOCK_GROUP_ITEM_KEY, 1G)
--> adds the btrfs_block_group_cache object
to the list new_bgs of the transaction
handle
btrfs_end_transaction(trans handle)
__btrfs_end_transaction()
btrfs_create_pending_block_groups()
--> sees the new btrfs_block_group_cache
in the new_bgs list of the transaction
handle
--> its call to btrfs_insert_item() fails
with -EEXIST when attempting to insert
the block group item key
(offset X, BTRFS_BLOCK_GROUP_ITEM_KEY, 1G)
because task A has not removed that key yet
--> aborts the running transaction with
error -EEXIST
btrfs_del_item()
-> removes the block group's key from
the extent tree, key is
(offset X, BTRFS_BLOCK_GROUP_ITEM_KEY, 1G)
Fix this by calling remove_extent_mapping(), at btrfs_remove_block_group(),
only at the very end, after removing the block group item key from the
extent tree (and removing the free space tree entry if we are using the
free space tree feature).
Fixes: 04216820fe83d5 ("Btrfs: fix race between fs trimming and block group remove/allocation") CC: [email protected] # 4.4+ Signed-off-by: Filipe Manana <[email protected]> Signed-off-by: David Sterba <[email protected]>
With v5.2-rc1, The ftrace functions_graph tracer locks up whenever it is
enabled on arm64.
Since commit 0ea415390cd3 ("clocksource/arm_arch_timer: Use
arch_timer_read_counter to access stable counters") a function pointer
is consistently used to read the counter instead of potentially
referencing an inlinable function.
The graph tracers relies on accessing the timer counters to compute the
time spent in functions which causes the lockup when attempting to trace
these code paths.
Annotate the arm arch timer counter accessors as notrace.
Michael Ellerman [Wed, 12 Jun 2019 13:35:07 +0000 (23:35 +1000)]
powerpc/mm/64s/hash: Reallocate context ids on fork
When using the Hash Page Table (HPT) MMU, userspace memory mappings
are managed at two levels. Firstly in the Linux page tables, much like
other architectures, and secondly in the SLB (Segment Lookaside
Buffer) and HPT. It's the SLB and HPT that are actually used by the
hardware to do translations.
As part of the series adding support for 4PB user virtual address
space using the hash MMU, we added support for allocating multiple
"context ids" per process, one for each 512TB chunk of address space.
These are tracked in an array called extended_id in the mm_context_t
of a process that has done a mapping above 512TB.
If such a process forks (ie. clone(2) without CLONE_VM set) it's mm is
copied, including the mm_context_t, and then init_new_context() is
called to reinitialise parts of the mm_context_t as appropriate to
separate the address spaces of the two processes.
The key step in ensuring the two processes have separate address
spaces is to allocate a new context id for the process, this is done
at the beginning of hash__init_new_context(). If we didn't allocate a
new context id then the two processes would share mappings as far as
the SLB and HPT are concerned, even though their Linux page tables
would be separate.
For mappings above 512TB, which use the extended_id array, we
neglected to allocate new context ids on fork, meaning the parent and
child use the same ids and therefore share those mappings even though
they're supposed to be separate. This can lead to the parent seeing
writes done by the child, which is essentially memory corruption.
There is an additional exposure which is that if the child process
exits, all its context ids are freed, including the context ids that
are still in use by the parent for mappings above 512TB. One or more
of those ids can then be reallocated to a third process, that process
can then read/write to the parent's mappings above 512TB. Additionally
if the freed id is used for the third process's primary context id,
then the parent is able to read/write to the third process's mappings
*below* 512TB.
All of these are fundamental failures to enforce separation between
processes. The only mitigating factor is that the bug only occurs if a
process creates mappings above 512TB, and most applications still do
not create such mappings.
Only machines using the hash page table MMU are affected, eg. PowerPC
970 (G5), PA6T, Power5/6/7/8/9. By default Power9 bare metal machines
(powernv) use the Radix MMU and are not affected, unless the machine
has been explicitly booted in HPT mode (using disable_radix on the
kernel command line). KVM guests on Power9 may be affected if the host
or guest is configured to use the HPT MMU. LPARs under PowerVM on
Power9 are affected as they always use the HPT MMU. Kernels built with
PAGE_SIZE=4K are not affected.
The fix is relatively simple, we need to reallocate context ids for
all extended mappings on fork.
Fixes: f384796c40dc ("powerpc/mm: Add support for handling > 512TB address in SLB miss") Cc: [email protected] # v4.17+ Signed-off-by: Michael Ellerman <[email protected]>
Linus Walleij [Thu, 30 May 2019 20:24:24 +0000 (22:24 +0200)]
i2c: pca-platform: Fix GPIO lookup code
The devm_gpiod_request_gpiod() call will add "-gpios" to
any passed connection ID before looking it up.
I do not think the reset GPIO on this platform is named
"reset-gpios-gpios" but rather "reset-gpios" in the device
tree, so fix this up so that we get a proper reset GPIO
handle.
Also drop the inclusion of the legacy GPIO header.
Mika Westerberg [Tue, 28 May 2019 15:56:20 +0000 (18:56 +0300)]
thunderbolt: Make sure device runtime resume completes before taking domain lock
When a device is authorized from userspace by writing to authorized
attribute we first take the domain lock and then runtime resume the
device in question. There are two issues with this.
First is that the device connected notifications are blocked during this
time which means we get them only after the authorization operation is
complete. Because of this the authorization needed flag from the
firmware notification is not reflecting the real authorization status
anymore. So what happens is that the "authorized" keeps returning 0 even
if the device was already authorized properly.
Second issue is that each time the controller is runtime resumed the
connection_id field of device connected notification may be different
than in the previous resume. We need to use the latest connection_id
otherwise the firmware rejects the authorization command.
Fix these by moving runtime resume operations to happen before the
domain lock is taken, and waiting for the updated device connected
notification from the firmware before we allow runtime resume of a
device to complete.
While there add missing locking to tb_switch_nvm_read().
Fixes: 09f11b6c99fe ("thunderbolt: Take domain lock in switch sysfs attribute callbacks") Reported-by: Pengfei Xu <[email protected]> Signed-off-by: Mika Westerberg <[email protected]>
We've moved the override and firmware EDID (simply "override EDID" from
now on) handling to the low level drm_do_get_edid() function in order to
transparently use the override throughout the stack. The idea is that
you get the override EDID via the ->get_modes() hook.
Unfortunately, there are scenarios where the DDC probe in drm_get_edid()
called via ->get_modes() fails, although the preceding ->detect()
succeeds.
In the case reported by Paul Wise, the ->detect() hook,
intel_crt_detect(), relies on hotplug detect, bypassing the DDC. In the
case reported by Ilpo Järvinen, there is no ->detect() hook, which is
interpreted as connected. The subsequent DDC probe reached via
->get_modes() fails, and we don't even look at the override EDID,
resulting in no modes being added.
Because drm_get_edid() is used via ->detect() all over the place, we
can't trivially remove the DDC probe, as it leads to override EDID
effectively meaning connector forcing. The goal is that connector
forcing and override EDID remain orthogonal.
Generally, the underlying problem here is the conflation of ->detect()
and ->get_modes() via drm_get_edid(). The former should just detect, and
the latter should just get the modes, typically via reading the EDID. As
long as drm_get_edid() is used in ->detect(), it needs to retain the DDC
probe. Or such users need to have a separate DDC probe step first.
The EDID caching between ->detect() and ->get_modes() done by some
drivers is a further complication that prevents us from making
drm_do_get_edid() adapt to the two cases.
Work around the regression by falling back to a separate attempt at
getting the override EDID at drm_helper_probe_single_connector_modes()
level. With a working DDC and override EDID, it'll never be called; the
override EDID will come via ->get_modes(). There will still be a failing
DDC probe attempt in the cases that require the fallback.
Russell King [Tue, 11 Jun 2019 16:48:18 +0000 (17:48 +0100)]
i2c: acorn: fix i2c warning
The Acorn i2c driver (for RiscPC) triggers the "i2c adapter has no name"
warning in the I2C core driver, resulting in the RTC being inaccessible.
Fix this.
ld.lld: error: undefined symbol: __efistub___stack_chk_guard
>>> referenced by arm-stub.c:73
(/home/nathan/cbl/linux/drivers/firmware/efi/libstub/arm-stub.c:73)
>>> arm-stub.stub.o:(__efistub_install_memreserve_table)
in archive ./drivers/firmware/efi/libstub/lib.a
These failures come from the lack of -fno-stack-protector, which is
added via cc-option in drivers/firmware/efi/libstub/Makefile. When an
unknown flag is added to KBUILD_CFLAGS, clang will noisily warn that it
is ignoring the option like above, unlike gcc, who will just error.
$ gcc -Wsometimes-uninitialized tmp.c; echo $?
gcc: error: unrecognized command line option
‘-Wsometimes-uninitialized’; did you mean ‘-Wmaybe-uninitialized’?
1
For cc-option to work properly with clang and behave like gcc, -Werror
is needed, which was done in commit c3f0d0bc5b01 ("kbuild, LLVMLinux:
Add -Werror to cc-option to support clang").
As a consequence of this, when an unknown flag is unconditionally added
to KBUILD_CFLAGS, it will cause cc-option to always fail and those flags
will never get added:
This can be seen when compiling the whole kernel as some warnings that
are normally disabled (see below) show up. The full list of flags
missing from drivers/firmware/efi/libstub are the following (gathered
from diffing .arm64-stub.o.cmd):
Vadim Pasternak [Thu, 23 May 2019 16:41:52 +0000 (16:41 +0000)]
platform/mellanox: mlxreg-hotplug: Add devm_free_irq call to remove flow
Add devm_free_irq() call to mlxreg-hotplug remove() for clean release
of devices irq resource. Fix debugobjects warning triggered by rmmod
It prevents of use-after-free memory, related to
mlxreg_hotplug_work_handler.
Issue has been reported as debugobjects warning triggered by
'rmmod mlxtreg-hotplug' flow, while running kernel with
CONFIG_DEBUG_OBJECTS* options.
Vadim Pasternak [Wed, 5 Jun 2019 07:51:03 +0000 (07:51 +0000)]
platform/x86: mlx-platform: Fix parent device in i2c-mux-reg device registration
Fix the issue found while running kernel with the option
CONFIG_DEBUG_TEST_DRIVER_REMOVE.
Driver 'mlx-platform' registers 'i2c_mlxcpld' device and then registers
few underlying 'i2c-mux-reg' devices:
priv->pdev_i2c = platform_device_register_simple("i2c_mlxcpld", nr,
NULL, 0);
...
for (i = 0; i < ARRAY_SIZE(mlxplat_mux_data); i++) {
priv->pdev_mux[i] = platform_device_register_resndata(
&mlxplat_dev->dev,
"i2c-mux-reg", i, NULL,
0, &mlxplat_mux_data[i],
sizeof(mlxplat_mux_data[i]));
But actual parent of "i2c-mux-reg" device is priv->pdev_i2c->dev and
not mlxplat_dev->dev.
Patch fixes parent device parameter in a call to
platform_device_register_resndata() for "i2c-mux-reg".
It solves the race during initialization flow while 'i2c_mlxcpld.1' is
removing after probe, while 'i2c-mux-reg.0' is still in probing flow:
'i2c_mlxcpld.1' flow: probe -> remove -> probe.
'i2c-mux-reg.0' flow: probe -> ...
Mathew King [Mon, 20 May 2019 22:41:24 +0000 (16:41 -0600)]
platform/x86: intel-vbtn: Report switch events when event wakes device
When a switch event, such as tablet mode/laptop mode or docked/undocked,
wakes a device make sure that the value of the swich is reported.
Without when a device is put in tablet mode from laptop mode when it is
suspended or vice versa the device will wake up but mode will be
incorrect.
Tested by suspending a device in laptop mode and putting it in tablet
mode, the device resumes and is in tablet mode. When suspending the
device in tablet mode and putting it in laptop mode the device resumes
and is in laptop mode.
Hans de Goede [Wed, 12 Jun 2019 07:02:02 +0000 (09:02 +0200)]
platform/x86: asus-wmi: Only Tell EC the OS will handle display hotkeys from asus_nb_wmi
Commit 78f3ac76d9e5 ("platform/x86: asus-wmi: Tell the EC the OS will
handle the display off hotkey") causes the backlight to be permanently off
on various EeePC laptop models using the eeepc-wmi driver (Asus EeePC
1015BX, Asus EeePC 1025C).
The asus_wmi_set_devstate(ASUS_WMI_DEVID_BACKLIGHT, 2, NULL) call added
by that commit is made conditional in this commit and only enabled in
the quirk_entry structs in the asus-nb-wmi driver fixing the broken
display / backlight on various EeePC laptop models.
Cc: João Paulo Rechi Vita <[email protected]> Fixes: 78f3ac76d9e5 ("platform/x86: asus-wmi: Tell the EC the OS will handle the display off hotkey") Signed-off-by: Hans de Goede <[email protected]> Signed-off-by: Andy Shevchenko <[email protected]>
Jan Kundrát [Fri, 17 May 2019 15:01:42 +0000 (17:01 +0200)]
ARM: mvebu_v7_defconfig: fix Ethernet on Clearfog
Compared to kernel 5.0, patches merged for 5.1 added support for A38x'
PHY guarded by a config option which was not enabled by default. As a
result, there was no eth1 and eth2 on a Solid Run Clearfog Base.
Ensure that A38x PHY is enabled on mvebu.
[gregory: issue appeared in 5.1 not in 5.2 and added Fixes tag]
Prarit Bhargava [Mon, 10 Jun 2019 17:15:44 +0000 (13:15 -0400)]
x86/resctrl: Prevent NULL pointer dereference when local MBM is disabled
Booting with kernel parameter "rdt=cmt,mbmtotal,memlocal,l3cat,mba" and
executing "mount -t resctrl resctrl -o mba_MBps /sys/fs/resctrl" results in
a NULL pointer dereference on systems which do not have local MBM support
enabled..
James Morse [Mon, 3 Jun 2019 17:25:31 +0000 (18:25 +0100)]
x86/resctrl: Don't stop walking closids when a locksetup group is found
When a new control group is created __init_one_rdt_domain() walks all
the other closids to calculate the sets of used and unused bits.
If it discovers a pseudo_locksetup group, it breaks out of the loop. This
means any later closid doesn't get its used bits added to used_b. These
bits will then get set in unused_b, and added to the new control group's
configuration, even if they were marked as exclusive for a later closid.
When encountering a pseudo_locksetup group, we should continue. This is
because "a resource group enters 'pseudo-locked' mode after the schemata is
written while the resource group is in 'pseudo-locksetup' mode." When we
find a pseudo_locksetup group, its configuration is expected to be
overwritten, we can skip it.
Robin Murphy [Mon, 3 Jun 2019 12:15:37 +0000 (14:15 +0200)]
iommu/arm-smmu: Avoid constant zero in TLBI writes
Apparently, some Qualcomm arm64 platforms which appear to expose their
SMMU global register space are still, in fact, using a hypervisor to
mediate it by trapping and emulating register accesses. Sadly, some
deployed versions of said trapping code have bugs wherein they go
horribly wrong for stores using r31 (i.e. XZR/WZR) as the source
register.
While this can be mitigated for GCC today by tweaking the constraints
for the implementation of writel_relaxed(), to avoid any potential
arms race with future compilers more aggressively optimising register
allocation, the simple way is to just remove all the problematic
constant zeros. For the write-only TLB operations, the actual value is
irrelevant anyway and any old nearby variable will provide a suitable
GPR to encode. The one point at which we really do need a zero to clear
a context bank happens before any of the TLB maintenance where crashes
have been reported, so is apparently not a problem... :/
Ville Syrjälä [Tue, 9 Apr 2019 14:40:49 +0000 (17:40 +0300)]
drm/i915/sdvo: Implement proper HDMI audio support for SDVO
Our SDVO audio support is pretty bogus. We can't push audio over the
SDVO bus, so trying to enable audio in the SDVO control register doesn't
do anything. In fact it looks like the SDVO encoder will always mix in
the audio coming over HDA, and there's no (at least documented) way to
disable that from our side. So HDMI audio does work currently on gen4
but only by luck really. On gen3 it got broken by the referenced commit.
And what has always been missing on every platform is the ELD.
To pass the ELD to the audio driver we need to write it to magic buffer
in the SDVO encoder hardware which then gets pulled out via HDA in the
other end. Ie. pretty much the same thing we had for native HDMI before
we started to just pass the ELD between the drivers. This sort of
explains why we even have that silly hardware buffer with native HDMI.
$ cat /proc/asound/card0/eld#1.0
-monitor_present 0
-eld_valid 0
+monitor_present 1
+eld_valid 1
+monitor_name LG TV
+connection_type HDMI
+...
This also fixes our state readout since we can now query the SDVO
encoder about the state of the "ELD valid" and "presence detect"
bits. As mentioned those don't actually control whether audio
gets sent over the HDMI cable, but it's the best we can do. And with
the state checker appeased we can re-enable HDMI audio for gen3.
Ville Syrjälä [Mon, 3 Jun 2019 14:25:00 +0000 (17:25 +0300)]
drm/i915: Fix per-pixel alpha with CCS
We forgot to set .has_alpha=true for the A+CCS formats when the code
started to consult .has_alpha. This manifests as A+CCS being treated
as X+CCS which means no per-pixel alpha blending. Fix the format
list appropriately.
Lucas De Marchi [Wed, 5 Jun 2019 23:55:35 +0000 (16:55 -0700)]
drm/i915/dmc: protect against reading random memory
While loading the DMC firmware we were double checking the headers made
sense, but in no place we checked that we were actually reading memory
we were supposed to. This could be wrong in case the firmware file is
truncated or malformed.
Before this patch:
# ls -l /lib/firmware/i915/icl_dmc_ver1_07.bin
-rw-r--r-- 1 root root 25716 Feb 1 12:26 icl_dmc_ver1_07.bin
# truncate -s 25700 /lib/firmware/i915/icl_dmc_ver1_07.bin
# modprobe i915
# dmesg| grep -i dmc
[drm:intel_csr_ucode_init [i915]] Loading i915/icl_dmc_ver1_07.bin
[drm] Finished loading DMC firmware i915/icl_dmc_ver1_07.bin (v1.7)
i.e. it loads random data. Now it fails like below:
[drm:intel_csr_ucode_init [i915]] Loading i915/icl_dmc_ver1_07.bin
[drm:csr_load_work_fn [i915]] *ERROR* Truncated DMC firmware, rejecting.
i915 0000:00:02.0: Failed to load DMC firmware i915/icl_dmc_ver1_07.bin. Disabling runtime power management.
i915 0000:00:02.0: DMC firmware homepage: https://git.kernel.org/pub/scm/linux/kernel/git/firmware/linux-firmware.git/tree/i915
Before reading any part of the firmware file, validate the input first.
projects / linux.git / log