vfio/gvt: fix DRM_I915_GVT dependency on VFIO_MDEV
The Kconfig dependency is incomplete since DRM_I915_GVT is a 'bool'
symbol that depends on the 'tristate' VFIO_MDEV. This allows a
configuration with VFIO_MDEV=m, DRM_I915_GVT=y and DRM_I915=y that
causes a link failure:
x86_64-linux-ld: drivers/gpu/drm/i915/gvt/gvt.o: in function `available_instances_show':
gvt.c:(.text+0x67a): undefined reference to `mtype_get_parent_dev'
x86_64-linux-ld: gvt.c:(.text+0x6a5): undefined reference to `mtype_get_type_group_id'
x86_64-linux-ld: drivers/gpu/drm/i915/gvt/gvt.o: in function `description_show':
gvt.c:(.text+0x76e): undefined reference to `mtype_get_parent_dev'
x86_64-linux-ld: gvt.c:(.text+0x799): undefined reference to `mtype_get_type_group_id'
Clarify the dependency by specifically disallowing the broken
configuration. If VFIO_MDEV is built-in, it will work, but if
VFIO_MDEV=m, the i915 driver cannot be built-in here.
vfio/iommu_type1: Remove unused pinned_page_dirty_scope in vfio_iommu
pinned_page_dirty_scope is optimized out by commit 010321565a7d
("vfio/iommu_type1: Mantain a counter for non_pinned_groups"),
but appears again due to some issues during merging branches.
We can safely remove it here.
Jason Gunthorpe [Tue, 6 Apr 2021 19:40:41 +0000 (16:40 -0300)]
vfio/mdev: Correct the function signatures for the mdev_type_attributes
The driver core standard is to pass in the properly typed object, the
properly typed attribute and the buffer data. It stems from the root
kobject method:
Jason Gunthorpe [Tue, 6 Apr 2021 19:40:40 +0000 (16:40 -0300)]
vfio/mdev: Remove kobj from mdev_parent_ops->create()
The kobj here is a type-erased version of mdev_type, which is already
stored in the struct mdev_device being passed in. It was only ever used to
compute the type_group_id, which is now extracted directly from the mdev.
Jason Gunthorpe [Tue, 6 Apr 2021 19:40:39 +0000 (16:40 -0300)]
vfio/gvt: Use mdev_get_type_group_id()
intel_gvt_init_vgpu_type_groups() makes gvt->types 1:1 with the
supported_type_groups array, so the type_group_id is also the index into
gvt->types. Use it directly and remove the string matching.
Jason Gunthorpe [Tue, 6 Apr 2021 19:40:38 +0000 (16:40 -0300)]
vfio/gvt: Make DRM_I915_GVT depend on VFIO_MDEV
At some point there may have been some reason for this weird split in this
driver, but today only the VFIO side is actually implemented.
However, it got messed up at some point and mdev code was put in gvt.c and
is pretending to be "generic" by masquerading as some generic attribute list:
static MDEV_TYPE_ATTR_RO(description);
But MDEV_TYPE attributes are only usable with mdev_device, nothing else.
Ideally all of this would be moved to kvmgt.c, but it is entangled with
the rest of the "generic" code in an odd way. Thus put in a kconfig
dependency so we don't get randconfig failures when the next patch creates
a link time dependency related to the use of MDEV_TYPE.
Jason Gunthorpe [Tue, 6 Apr 2021 19:40:37 +0000 (16:40 -0300)]
vfio/mbochs: Use mdev_get_type_group_id()
The mbochs_types array is parallel to the supported_type_groups array, so
the type_group_id indexes both. Instead of doing string searching just
directly index with type_group_id in all places.
Jason Gunthorpe [Tue, 6 Apr 2021 19:40:36 +0000 (16:40 -0300)]
vfio/mdpy: Use mdev_get_type_group_id()
The mdpy_types array is parallel to the supported_type_groups array, so
the type_group_id indexes both. Instead of doing string searching just
directly index with type_group_id in all places.
Jason Gunthorpe [Tue, 6 Apr 2021 19:40:34 +0000 (16:40 -0300)]
vfio/mdev: Add mdev/mtype_get_type_group_id()
This returns the index in the supported_type_groups array that is
associated with the mdev_type attached to the struct mdev_device or its
containing struct kobject.
Each mdev_device can be spawned from exactly one mdev_type, which in turn
originates from exactly one supported_type_group.
Drivers are using weird string calculations to try and get back to this
index, providing a direct access to the index removes a bunch of wonky
driver code.
mdev_type->group can be deleted as the group is obtained using the
type_group_id.
Jason Gunthorpe [Tue, 6 Apr 2021 19:40:33 +0000 (16:40 -0300)]
vfio/mdev: Remove duplicate storage of parent in mdev_device
mdev_device->type->parent is the same thing.
The struct mdev_device was relying on the kref on the mdev_parent to also
indirectly hold a kref on the mdev_type pointer. Now that the type holds a
kref on the parent we can directly kref the mdev_type and remove this
implicit relationship.
Jason Gunthorpe [Tue, 6 Apr 2021 19:40:31 +0000 (16:40 -0300)]
vfio/mdev: Reorganize mdev_device_create()
Once the memory for the struct mdev_device is allocated it should
immediately be device_initialize()'d and filled in so that put_device()
can always be used to undo the allocation.
Place the mdev_get/put_parent() so that they are clearly protecting the
mdev->parent pointer. Move the final put to the release function so that
the lifetime rules are trivial to understand. Update the goto labels to
follow the normal convention.
Remove mdev_device_free() as the release function via device_put() is now
usable in all cases.
Jason Gunthorpe [Tue, 6 Apr 2021 19:40:30 +0000 (16:40 -0300)]
vfio/mdev: Add missing reference counting to mdev_type
struct mdev_type holds a pointer to the kref'd object struct mdev_parent,
but doesn't hold the kref. The lifetime of the parent becomes implicit
because parent_remove_sysfs_files() is supposed to remove all the access
before the parent can be freed, but this is very hard to reason about.
Make it obviously correct by adding the missing get.
Jason Gunthorpe [Tue, 6 Apr 2021 19:40:28 +0000 (16:40 -0300)]
vfio/mdev: Use struct mdev_type in struct mdev_device
The kobj pointer in mdev_device is actually pointing at a struct
mdev_type. Use the proper type so things are understandable.
There are a number of places that are confused and passing both the mdev
and the mtype as function arguments, fix these to derive the mtype
directly from the mdev to remove the redundancy.
Jason Gunthorpe [Tue, 6 Apr 2021 19:40:27 +0000 (16:40 -0300)]
vfio/mdev: Simplify driver registration
This is only done once, we don't need to generate code to initialize a
structure stored in the ELF .data segment. Fill in the three required
.driver members directly instead of copying data into them during
mdev_register_driver().
Further the to_mdev_driver() function doesn't belong in a public header,
just inline it into the two places that need it. Finally, we can now
clearly see that 'drv' derived from dev->driver cannot be NULL, firstly
because the driver core forbids it, and secondly because NULL won't pass
through the container_of(). Remove the dead code.
Jason Gunthorpe [Tue, 6 Apr 2021 19:40:26 +0000 (16:40 -0300)]
vfio/mdev: Add missing typesafety around mdev_device
The mdev API should accept and pass a 'struct mdev_device *' in all
places, not pass a 'struct device *' and cast it internally with
to_mdev_device(). Particularly in its struct mdev_driver functions, the
whole point of a bus's struct device_driver wrapper is to provide type
safety compared to the default struct device_driver.
Further, the driver core standard is for bus drivers to expose their
device structure in their public headers that can be used with
container_of() inlines and '&foo->dev' to go between the class levels, and
'&foo->dev' to be used with dev_err/etc driver core helper functions. Move
'struct mdev_device' to mdev.h
Once done this allows moving some one instruction exported functions to
static inlines, which in turns allows removing one of the two grotesque
symbol_get()'s related to mdev in the core code.
Jason Gunthorpe [Tue, 30 Mar 2021 15:53:08 +0000 (09:53 -0600)]
vfio/pci: Replace uses of vfio_device_data() with container_of
This tidies a few confused places that think they can have a refcount on
the vfio_device but the device_data could be NULL, that isn't possible by
design.
Most of the change falls out when struct vfio_devices is updated to just
store the struct vfio_pci_device itself. This wasn't possible before
because there was no easy way to get from the 'struct vfio_pci_device' to
the 'struct vfio_device' to put back the refcount.
Jason Gunthorpe [Tue, 30 Mar 2021 15:53:08 +0000 (09:53 -0600)]
vfio: Make vfio_device_ops pass a 'struct vfio_device *' instead of 'void *'
This is the standard kernel pattern, the ops associated with a struct get
the struct pointer in for typesafety. The expected design is to use
container_of to cleanly go from the subsystem level type to the driver
level type without having any type erasure in a void *.
Jason Gunthorpe [Tue, 30 Mar 2021 15:53:07 +0000 (09:53 -0600)]
vfio/mdev: Make to_mdev_device() into a static inline
The macro wrongly uses 'dev' as both the macro argument and the member
name, which means it fails compilation if any caller uses a word other
than 'dev' as the single argument. Fix this defect by making it into
proper static inline, which is more clear and typesafe anyhow.
Jason Gunthorpe [Tue, 30 Mar 2021 15:53:07 +0000 (09:53 -0600)]
vfio/mdev: Use vfio_init/register/unregister_group_dev
mdev gets little benefit because it doesn't actually do anything, however
it is the last user, so move the vfio_init/register/unregister_group_dev()
code here for now.
Jason Gunthorpe [Tue, 30 Mar 2021 15:53:07 +0000 (09:53 -0600)]
vfio/pci: Use vfio_init/register/unregister_group_dev
pci already allocates a struct vfio_pci_device with exactly the same
lifetime as vfio_device, switch to the new API and embed vfio_device in
vfio_pci_device.
Jason Gunthorpe [Tue, 30 Mar 2021 15:53:06 +0000 (09:53 -0600)]
vfio/pci: Re-order vfio_pci_probe()
vfio_add_group_dev() must be called only after all of the private data in
vdev is fully setup and ready, otherwise there could be races with user
space instantiating a device file descriptor and starting to call ops.
For instance vfio_pci_reflck_attach() sets vdev->reflck and
vfio_pci_open(), called by fops open, unconditionally derefs it, which
will crash if things get out of order.
Fixes: cc20d7999000 ("vfio/pci: Introduce VF token") Fixes: e309df5b0c9e ("vfio/pci: Parallelize device open and release") Fixes: 6eb7018705de ("vfio-pci: Move idle devices to D3hot power state") Fixes: ecaa1f6a0154 ("vfio-pci: Add VGA arbiter client") Reviewed-by: Christoph Hellwig <[email protected]> Reviewed-by: Max Gurtovoy <[email protected]> Reviewed-by: Kevin Tian <[email protected]> Reviewed-by: Cornelia Huck <[email protected]> Reviewed-by: Eric Auger <[email protected]> Signed-off-by: Jason Gunthorpe <[email protected]>
Message-Id: <8-v3-225de1400dfc[email protected]> Signed-off-by: Alex Williamson <[email protected]>
Jason Gunthorpe [Tue, 30 Mar 2021 15:53:06 +0000 (09:53 -0600)]
vfio/pci: Move VGA and VF initialization to functions
vfio_pci_probe() is quite complicated, with optional VF and VGA sub
components. Move these into clear init/uninit functions and have a linear
flow in probe/remove.
This fixes a few little buglets:
- vfio_pci_remove() is in the wrong order, vga_client_register() removes
a notifier and is after kfree(vdev), but the notifier refers to vdev,
so it can use after free in a race.
- vga_client_register() can fail but was ignored
Organize things so destruction order is the reverse of creation order.
Jason Gunthorpe [Tue, 30 Mar 2021 15:53:06 +0000 (09:53 -0600)]
vfio/fsl-mc: Use vfio_init/register/unregister_group_dev
fsl-mc already allocates a struct vfio_fsl_mc_device with exactly the same
lifetime as vfio_device, switch to the new API and embed vfio_device in
vfio_fsl_mc_device. While here remove the devm usage for the vdev, this
code is clean and doesn't need devm.
Jason Gunthorpe [Tue, 30 Mar 2021 15:53:06 +0000 (09:53 -0600)]
vfio/fsl-mc: Re-order vfio_fsl_mc_probe()
vfio_add_group_dev() must be called only after all of the private data in
vdev is fully setup and ready, otherwise there could be races with user
space instantiating a device file descriptor and starting to call ops.
For instance vfio_fsl_mc_reflck_attach() sets vdev->reflck and
vfio_fsl_mc_open(), called by fops open, unconditionally derefs it, which
will crash if things get out of order.
This driver started life with the right sequence, but two commits added
stuff after vfio_add_group_dev().
Fixes: 2e0d29561f59 ("vfio/fsl-mc: Add irq infrastructure for fsl-mc devices") Fixes: f2ba7e8c947b ("vfio/fsl-mc: Added lock support in preparation for interrupt handling") Co-developed-by: Diana Craciun OSS <[email protected]> Signed-off-by: Jason Gunthorpe <[email protected]>
Message-Id: <5-v3-225de1400dfc[email protected]> Signed-off-by: Alex Williamson <[email protected]>
Jason Gunthorpe [Tue, 30 Mar 2021 15:53:06 +0000 (09:53 -0600)]
vfio/platform: Use vfio_init/register/unregister_group_dev
platform already allocates a struct vfio_platform_device with exactly
the same lifetime as vfio_device, switch to the new API and embed
vfio_device in vfio_platform_device.
Jason Gunthorpe [Tue, 30 Mar 2021 15:53:05 +0000 (09:53 -0600)]
vfio: Split creation of a vfio_device into init and register ops
This makes the struct vfio_device part of the public interface so it
can be used with container_of and so forth, as is typical for a Linux
subystem.
This is the first step to bring some type-safety to the vfio interface by
allowing the replacement of 'void *' and 'struct device *' inputs with a
simple and clear 'struct vfio_device *'
For now the self-allocating vfio_add_group_dev() interface is kept so each
user can be updated as a separate patch.
The expected usage pattern is
driver core probe() function:
my_device = kzalloc(sizeof(*mydevice));
vfio_init_group_dev(&my_device->vdev, dev, ops, mydevice);
/* other driver specific prep */
vfio_register_group_dev(&my_device->vdev);
dev_set_drvdata(dev, my_device);
driver core remove() function:
my_device = dev_get_drvdata(dev);
vfio_unregister_group_dev(&my_device->vdev);
/* other driver specific tear down */
kfree(my_device);
Allowing the driver to be able to use the drvdata and vfio_device to go
to/from its own data.
The pattern also makes it clear that vfio_register_group_dev() must be
last in the sequence, as once it is called the core code can immediately
start calling ops. The init/register gap is provided to allow for the
driver to do setup before ops can be called and thus avoid races.
Jason Gunthorpe [Tue, 30 Mar 2021 15:53:05 +0000 (09:53 -0600)]
vfio: Simplify the lifetime logic for vfio_device
The vfio_device is using a 'sleep until all refs go to zero' pattern for
its lifetime, but it is indirectly coded by repeatedly scanning the group
list waiting for the device to be removed on its own.
Switch this around to be a direct representation, use a refcount to count
the number of places that are blocking destruction and sleep directly on a
completion until that counter goes to zero. kfree the device after other
accesses have been excluded in vfio_del_group_dev(). This is a fairly
common Linux idiom.
Due to this we can now remove kref_put_mutex(), which is very rarely used
in the kernel. Here it is being used to prevent a zero ref device from
being seen in the group list. Instead allow the zero ref device to
continue to exist in the device_list and use refcount_inc_not_zero() to
exclude it once refs go to zero.
This patch is organized so the next patch will be able to alter the API to
allow drivers to provide the kfree.
Jason Gunthorpe [Tue, 30 Mar 2021 15:53:05 +0000 (09:53 -0600)]
vfio: Remove extra put/gets around vfio_device->group
The vfio_device->group value has a get obtained during
vfio_add_group_dev() which gets moved from the stack to vfio_device->group
in vfio_group_create_device().
The reference remains until we reach the end of vfio_del_group_dev() when
it is put back.
Thus anything that already has a kref on the vfio_device is guaranteed a
valid group pointer. Remove all the extra reference traffic.
It is tricky to see, but the get at the start of vfio_del_group_dev() is
actually pairing with the put hidden inside vfio_device_put() a few lines
below.
A later patch merges vfio_group_create_device() into vfio_add_group_dev()
which makes the ownership and error flow on the create side easier to
follow.
This driver never had any open userspace (which for VFIO would include
VM kernel drivers) that use it, and thus should never have been added
by our normal userspace ABI rules.
Shenming Lu [Tue, 6 Apr 2021 13:50:09 +0000 (21:50 +0800)]
vfio/type1: Remove the almost unused check in vfio_iommu_type1_unpin_pages
The check i > npage at the end of vfio_iommu_type1_unpin_pages is unused
unless npage < 0, but if npage < 0, this function will return npage, which
should return -EINVAL instead. So let's just check the parameter npage at
the start of the function. By the way, replace unpin_exit with break.
Fred Gao [Tue, 30 Mar 2021 15:54:21 +0000 (09:54 -0600)]
vfio/pci: Add support for opregion v2.1+
Before opregion version 2.0 VBT data is stored in opregion mailbox #4,
but when VBT data exceeds 6KB size and cannot be within mailbox #4
then from opregion v2.0+, Extended VBT region, next to opregion is
used to hold the VBT data, so the total size will be opregion size plus
extended VBT region size.
Since opregion v2.0 with physical host VBT address would not be
practically available for end user and guest can not directly access
host physical address, so it is not supported.
Zheyu Ma [Sat, 3 Apr 2021 06:58:36 +0000 (06:58 +0000)]
firewire: nosy: Fix a use-after-free bug in nosy_ioctl()
For each device, the nosy driver allocates a pcilynx structure.
A use-after-free might happen in the following scenario:
1. Open nosy device for the first time and call ioctl with command
NOSY_IOC_START, then a new client A will be malloced and added to
doubly linked list.
2. Open nosy device for the second time and call ioctl with command
NOSY_IOC_START, then a new client B will be malloced and added to
doubly linked list.
3. Call ioctl with command NOSY_IOC_START for client A, then client A
will be readded to the doubly linked list. Now the doubly linked
list is messed up.
4. Close the first nosy device and nosy_release will be called. In
nosy_release, client A will be unlinked and freed.
5. Close the second nosy device, and client A will be referenced,
resulting in UAF.
The root cause of this bug is that the element in the doubly linked list
is reentered into the list.
Fix this bug by adding a check before inserting a client. If a client
is already in the linked list, don't insert it.
The following KASAN report reveals it:
BUG: KASAN: use-after-free in nosy_release+0x1ea/0x210
Write of size 8 at addr ffff888102ad7360 by task poc
CPU: 3 PID: 337 Comm: poc Not tainted 5.12.0-rc5+ #6
Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.12.0-59-gc9ba5276e321-prebuilt.qemu.org 04/01/2014
Call Trace:
nosy_release+0x1ea/0x210
__fput+0x1e2/0x840
task_work_run+0xe8/0x180
exit_to_user_mode_prepare+0x114/0x120
syscall_exit_to_user_mode+0x1d/0x40
entry_SYSCALL_64_after_hwframe+0x44/0xae
The buggy address belongs to the object at ffff888102ad7300 which belongs to the cache kmalloc-128 of size 128
The buggy address is located 96 bytes inside of 128-byte region [ffff888102ad7300, ffff888102ad7380)
[ Modified to use 'list_empty()' inside proper lock - Linus ]
Merge tag 'io_uring-5.12-2021-04-03' of git://git.kernel.dk/linux-block
POull io_uring fix from Jens Axboe:
"Just fixing a silly braino in a previous patch, where we'd end up
failing to compile if CONFIG_BLOCK isn't enabled.
Not that a lot of people do that, but kernel bot spotted it and it's
probably prudent to just flush this out now before -rc6.
Sorry about that, none of my test compile configs have !CONFIG_BLOCK"
* tag 'io_uring-5.12-2021-04-03' of git://git.kernel.dk/linux-block:
io_uring: fix !CONFIG_BLOCK compilation failure
Merge tag 'gfs2-v5.12-rc2-fixes2' of git://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2
Pull gfs2 fixes from Andreas Gruenbacher:
"Two more gfs2 fixes"
* tag 'gfs2-v5.12-rc2-fixes2' of git://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2:
gfs2: report "already frozen/thawed" errors
gfs2: Flag a withdraw if init_threads() fails
Merge tag 'riscv-for-linus-5.12-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux
Pull RISC-V fixes from Palmer Dabbelt:
"A handful of fixes for 5.12:
- fix a stack tracing regression related to "const register asm"
variables, which have unexpected behavior.
- ensure the value to be written by put_user() is evaluated before
enabling access to userspace memory..
- align the exception vector table correctly, so we don't rely on the
firmware's handling of unaligned accesses.
- build fix to make NUMA depend on MMU, which triggered on some
randconfigs"
* tag 'riscv-for-linus-5.12-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux:
riscv: Make NUMA depend on MMU
riscv: remove unneeded semicolon
riscv,entry: fix misaligned base for excp_vect_table
riscv: evaluate put_user() arg before enabling user access
riscv: Drop const annotation for sp
Merge tag 'powerpc-5.12-5' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux
Pull powerpc fixes from Michael Ellerman:
"Fix a bug on pseries where spurious wakeups from H_PROD would prevent
partition migration from succeeding.
Fix oopses seen in pcpu_alloc(), caused by parallel faults of the
percpu mapping causing us to corrupt the protection key used for the
mapping, and cause a fatal key fault.
Thanks to Aneesh Kumar K.V, Murilo Opsfelder Araujo, and Nathan Lynch"
* tag 'powerpc-5.12-5' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
powerpc/mm/book3s64: Use the correct storage key value when calling H_PROTECT
powerpc/pseries/mobility: handle premature return from H_JOIN
powerpc/pseries/mobility: use struct for shared state
Merge tag 'driver-core-5.12-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core
Pull driver core fix from Greg KH:
"Here is a single driver core fix for a reported problem with differed
probing. It has been in linux-next for a while with no reported
problems"
* tag 'driver-core-5.12-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core:
driver core: clear deferred probe reason on probe retry
Merge tag 'char-misc-5.12-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc
Pull char/misc driver fixes from Greg KH:
"Here are a few small driver char/misc changes for 5.12-rc6.
Nothing major here, a few fixes for reported issues:
- interconnect fixes for problems found
- fbcon syzbot-found fix
- extcon fixes
- firmware stratix10 bugfix
- MAINTAINERS file update.
All of these have been in linux-next for a while with no reported
issues"
* tag 'char-misc-5.12-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc:
drivers: video: fbcon: fix NULL dereference in fbcon_cursor()
mei: allow map and unmap of client dma buffer only for disconnected client
MAINTAINERS: Add linux-phy list and patchwork
interconnect: Fix kerneldoc warning
firmware: stratix10-svc: reset COMMAND_RECONFIG_FLAG_PARTIAL to 0
extcon: Fix error handling in extcon_dev_register
extcon: Add stubs for extcon_register_notifier_all() functions
interconnect: core: fix error return code of icc_link_destroy()
interconnect: qcom: msm8939: remove rpm-ids from non-RPM nodes
Merge tag 'staging-5.12-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging
Pull staging driver fixes from Greg KH:
"Here are two rtl8192e staging driver fixes for reported problems.
Both of these have been in linux-next for a while with no reported
issues"
* tag 'staging-5.12-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging:
staging: rtl8192e: Change state information from u16 to u8
staging: rtl8192e: Fix incorrect source in memcpy()
Merge tag 'tty-5.12-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty
Pull serial driver fix from Greg KH:
"Here is a single serial driver fix for 5.12-rc6. Is is a revert of a
change that showed up in 5.9 that has been reported to cause problems.
It has been in linux-next for a while with no reported issues"
* tag 'tty-5.12-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty:
soc: qcom-geni-se: Cleanup the code to remove proxy votes
kernel test robot correctly pinpoints a compilation failure if
CONFIG_BLOCK isn't set:
fs/io_uring.c: In function '__io_complete_rw':
>> fs/io_uring.c:2509:48: error: implicit declaration of function 'io_rw_should_reissue'; did you mean 'io_rw_reissue'? [-Werror=implicit-function-declaration]
2509 | if ((res == -EAGAIN || res == -EOPNOTSUPP) && io_rw_should_reissue(req)) {
| ^~~~~~~~~~~~~~~~~~~~
| io_rw_reissue
cc1: some warnings being treated as errors
Ensure that we have a stub declaration of io_rw_should_reissue() for
!CONFIG_BLOCK.
Fixes: 230d50d448ac ("io_uring: move reissue into regular IO path") Reported-by: kernel test robot <[email protected]> Signed-off-by: Jens Axboe <[email protected]>
Merge tag 'block-5.12-2021-04-02' of git://git.kernel.dk/linux-block
Pull block fixes from Jens Axboe:
- Remove comment that never came to fruition in 22 years of development
(Christoph)
- Remove unused request flag (Christoph)
- Fix for null_blk fake timeout handling (Damien)
- Fix for IOCB_NOWAIT being ignored for O_DIRECT on raw bdevs (Pavel)
- Error propagation fix for multiple split bios (Yufen)
* tag 'block-5.12-2021-04-02' of git://git.kernel.dk/linux-block:
block: remove the unused RQF_ALLOCED flag
block: update a few comments in uapi/linux/blkpg.h
block: don't ignore REQ_NOWAIT for direct IO
null_blk: fix command timeout completion handling
block: only update parent bi_status when bio fail
Merge tag 'io_uring-5.12-2021-04-02' of git://git.kernel.dk/linux-block
Pull io_uring fixes from Jens Axboe:
"Nothing really major in here, and finally nothing really related to
signals. A few minor fixups related to the threading changes, and some
general fixes, that's it.
There's the pending gdb-get-confused-about-arch, but that's more of a
cosmetic issue, nothing that hinder use of it. And given that other
archs will likely be affected by that oddity too, better to postpone
any changes there until 5.13 imho"
* tag 'io_uring-5.12-2021-04-02' of git://git.kernel.dk/linux-block:
io_uring: move reissue into regular IO path
io_uring: fix EIOCBQUEUED iter revert
io_uring/io-wq: protect against sprintf overflow
io_uring: don't mark S_ISBLK async work as unbounded
io_uring: drop sqd lock before handling signals for SQPOLL
io_uring: handle setup-failed ctx in kill_timeouts
io_uring: always go for cancellation spin on exec
Merge tag 'acpi-5.12-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
Pull ACPI fixes from Rafael Wysocki:
"These fix an ACPI tables management issue, an issue related to the
ACPI enumeration of devices and CPU wakeup in the ACPI processor
driver.
Specifics:
- Ensure that the memory occupied by ACPI tables on x86 will always
be reserved to prevent it from being allocated for other purposes
which was possible in some cases (Rafael Wysocki).
- Fix the ACPI device enumeration code to prevent it from attempting
to evaluate the _STA control method for devices with unmet
dependencies which is likely to fail (Hans de Goede).
- Fix the handling of CPU0 wakeup in the ACPI processor driver to
prevent CPU0 online failures from occurring (Vitaly Kuznetsov)"
* tag 'acpi-5.12-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
ACPI: processor: Fix CPU0 wakeup in acpi_idle_play_dead()
ACPI: scan: Fix _STA getting called on devices with unmet dependencies
ACPI: tables: x86: Reserve memory occupied by ACPI tables
Merge tag 'pm-5.12-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
Pull power management fixes from Rafael Wysocki:
"These fix a race condition and an ordering issue related to using
device links in the runtime PM framework and two kerneldoc comments in
cpufreq.
Specifics:
- Fix race condition related to the handling of supplier devices
during consumer device probe and fix the order of decrementation of
two related reference counters in the runtime PM core code handling
supplier devices (Adrian Hunter).
- Fix kerneldoc comments in cpufreq that have not been updated along
with the functions documented by them (Geert Uytterhoeven)"
* tag 'pm-5.12-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
PM: runtime: Fix race getting/putting suppliers at probe
PM: runtime: Fix ordering in pm_runtime_get_suppliers()
cpufreq: Fix scaling_{available,boost}_frequencies_show() comments
block: update a few comments in uapi/linux/blkpg.h
The big top of the file comment talk about grand plans that never
happened, so remove them to not confuse the readers. Also mark the
devname and volname fields as ignored as they were never used by the
kernel.
Merge tag 'trace-v5.12-rc5-2' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace
Pull tracing fix from Steven Rostedt:
"Fix stack trace entry size to stop showing garbage
The macro that creates both the structure and the format displayed to
user space for the stack trace event was changed a while ago to fix
the parsing by user space tooling. But this change also modified the
structure used to store the stack trace event. It changed the caller
array field from [0] to [8].
Even though the size in the ring buffer is dynamic and can be
something other than 8 (user space knows how to handle this), the 8
extra words was not accounted for when reserving the event on the ring
buffer, and added 8 more entries, due to the calculation of
"sizeof(*entry) + nr_entries * sizeof(long)", as the sizeof(*entry)
now contains 8 entries.
The size of the caller field needs to be subtracted from the size of
the entry to create the correct allocation size"
* tag 'trace-v5.12-rc5-2' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
tracing: Fix stack trace event size
It's non-obvious how retry is done for block backed files, when it happens
off the kiocb done path. It also makes it tricky to deal with the iov_iter
handling.
Just mark the req as needing a reissue, and handling it from the
submission path instead. This makes it directly obvious that we're not
re-importing the iovec from userspace past the submit point, and it means
that we can just reuse our usual -EAGAIN retry path from the read/write
handling.
At some point in the future, we'll gain the ability to always reliably
return -EAGAIN through the stack. A previous attempt on the block side
didn't pan out and got reverted, hence the need to check for this
information out-of-band right now.
Pavel Begunkov [Fri, 20 Nov 2020 17:10:28 +0000 (17:10 +0000)]
block: don't ignore REQ_NOWAIT for direct IO
If IOCB_NOWAIT is set on submission, then that needs to get propagated to
REQ_NOWAIT on the block side. Otherwise we completely lose this
information, and any issuer of IOCB_NOWAIT IO will potentially end up
blocking on eg request allocation on the storage side.
Zihao Yu [Wed, 17 Mar 2021 08:17:25 +0000 (16:17 +0800)]
riscv,entry: fix misaligned base for excp_vect_table
In RV64, the size of each entry in excp_vect_table is 8 bytes. If the
base of the table is not 8-byte aligned, loading an entry in the table
will raise a misaligned exception. Although such exception will be
handled by opensbi/bbl, this still causes performance degradation.
Ben Dooks [Mon, 29 Mar 2021 09:57:49 +0000 (10:57 +0100)]
riscv: evaluate put_user() arg before enabling user access
The <asm/uaccess.h> header has a problem with put_user(a, ptr) if
the 'a' is not a simple variable, such as a function. This can lead
to the compiler producing code as so:
1: enable_user_access()
2: evaluate 'a' into register 'r'
3: put 'r' to 'ptr'
4: disable_user_acess()
The issue is that 'a' is now being evaluated with the user memory
protections disabled. So we try and force the evaulation by assigning
'x' to __val at the start, and hoping the compiler barriers in
enable_user_access() do the job of ordering step 2 before step 1.
This has shown up in a bug where 'a' sleeps and thus schedules out
and loses the SR_SUM flag. This isn't sufficient to fully fix, but
should reduce the window of opportunity. The first instance of this
we found is in scheudle_tail() where the code does:
$ less -N kernel/sched/core.c
4263 if (current->set_child_tid)
4264 put_user(task_pid_vnr(current), current->set_child_tid);
Here, the task_pid_vnr(current) is called within the block that has
enabled the user memory access. This can be made worse with KASAN
which makes task_pid_vnr() a rather large call with plenty of
opportunity to sleep.
Signed-off-by: Ben Dooks <[email protected]> Reported-by: [email protected] Suggested-by: Arnd Bergman <[email protected]>
--
Changes since v1:
- fixed formatting and updated the patch description with more info
Changes since v2:
- fixed commenting on __put_user() ([email protected])
Change since v3:
- fixed RFC in patch title. Should be ready to merge.
Merge tag 'lto-v5.12-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/kees/linux
Pull LTO fix from Kees Cook:
"It seems that there is a bug in ld.bfd when doing module section
merging.
As explicit merging is only needed for LTO, the work-around is to only
do it under LTO, leaving the original section layout choices alone
under normal builds:
- Only perform explicit module section merges under LTO (Sean
Christopherson)"
* tag 'lto-v5.12-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/kees/linux:
kbuild: lto: Merge module sections if and only if CONFIG_LTO_CLANG is enabled
kbuild: lto: Merge module sections if and only if CONFIG_LTO_CLANG is enabled
Merge module sections only when using Clang LTO. With ld.bfd, merging
sections does not appear to update the symbol tables for the module,
e.g. 'readelf -s' shows the value that a symbol would have had, if
sections were not merged. ld.lld does not show this problem.
The stale symbol table breaks gdb's function disassembler, and presumably
other things, e.g.
Merge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Pull kvm fixes from Paolo Bonzini:
"It's a bit larger than I (and probably you) would like by the time we
get to -rc6, but perhaps not entirely unexpected since the changes in
the last merge window were larger than usual.
x86:
- Fixes for missing TLB flushes with TDP MMU
- Fixes for race conditions in nested SVM
- Fixes for lockdep splat with Xen emulation
- Fix for kvmclock underflow
- Fix srcdir != builddir builds
- Other small cleanups
ARM:
- Fix GICv3 MMIO compatibility probing
- Prevent guests from using the ARMv8.4 self-hosted tracing
extension"
* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
selftests: kvm: Check that TSC page value is small after KVM_SET_CLOCK(0)
KVM: x86: Prevent 'hv_clock->system_time' from going negative in kvm_guest_time_update()
KVM: x86: disable interrupts while pvclock_gtod_sync_lock is taken
KVM: x86: reduce pvclock_gtod_sync_lock critical sections
KVM: SVM: ensure that EFER.SVME is set when running nested guest or on nested vmexit
KVM: SVM: load control fields from VMCB12 before checking them
KVM: x86/mmu: Don't allow TDP MMU to yield when recovering NX pages
KVM: x86/mmu: Ensure TLBs are flushed for TDP MMU during NX zapping
KVM: x86/mmu: Ensure TLBs are flushed when yielding during GFN range zap
KVM: make: Fix out-of-source module builds
selftests: kvm: make hardware_disable_test less verbose
KVM: x86/vPMU: Forbid writing to MSR_F15H_PERF MSRs when guest doesn't have X86_FEATURE_PERFCTR_CORE
KVM: x86: remove unused declaration of kvm_write_tsc()
KVM: clean up the unused argument
tools/kvm_stat: Add restart delay
KVM: arm64: Fix CPU interface MMIO compatibility detection
KVM: arm64: Disable guest access to trace filter controls
KVM: arm64: Hide system instruction access to Trace registers
Merge tag 'drm-fixes-2021-04-02' of git://anongit.freedesktop.org/drm/drm
Pull drm fixes from Dave Airlie:
"Things have settled down in time for Easter, a random smattering of
small fixes across a few drivers.
I'm guessing though there might be some i915 and misc fixes out there
I haven't gotten yet, but since today is a public holiday here, I'm
sending this early so I can have the day off, I'll see if more
requests come in and decide what to do with them later.
amdgpu:
- Polaris idle power fix
- VM fix
- Vangogh S3 fix
- Fixes for non-4K page sizes
* tag 'drm-fixes-2021-04-02' of git://anongit.freedesktop.org/drm/drm:
drm/amdgpu: check alignment on CPU page for bo map
drm/amdgpu: Set a suitable dev_info.gart_page_size
drm/amdgpu/vangogh: don't check for dpm in is_dpm_running when in suspend
drm/amdkfd: dqm fence memory corruption
drm/tegra: sor: Grab runtime PM reference across reset
drm/tegra: dc: Restore coupling of display controllers
gpu: host1x: Use different lock classes for each client
drm/tegra: dc: Don't set PLL clock to 0Hz
drm/amdgpu: fix offset calculation in amdgpu_vm_bo_clear_mappings()
drm/amd/pm: no need to force MCLK to highest when no display connected
drm/exynos/decon5433: Remove the unused include statements
drm/imx: imx-ldb: fix out of bounds array access warning
drm/imx: imx-ldb: Register LDB channel1 when it is the only channel to be used
drm/imx: fix memory leak when fails to init
Dave Airlie [Thu, 1 Apr 2021 18:52:45 +0000 (04:52 +1000)]
Merge tag 'imx-drm-fixes-2021-04-01' of git://git.pengutronix.de/git/pza/linux into drm-fixes
drm/imx: imx-drm-core and imx-ldb fixes
Fix a memory leak in an error path during DRM device initialization,
fix the LDB driver to register channel 1 even if channel 0 is unused,
and fix an out of bounds array access warning in the LDB driver.
Dave Airlie [Thu, 1 Apr 2021 18:44:28 +0000 (04:44 +1000)]
Merge tag 'drm/tegra/for-5.12-rc6' of ssh://git.freedesktop.org/git/tegra/linux into drm-fixes
drm/tegra: Fixes for v5.12-rc6
This contains a couple of fixes for various issues such as lockdep
warnings, runtime PM references, coupled display controllers and
misconfigured PLLs.
Commit cbc3b92ce037 fixed an issue to modify the macros of the stack trace
event so that user space could parse it properly. Originally the stack
trace format to user space showed that the called stack was a dynamic
array. But it is not actually a dynamic array, in the way that other
dynamic event arrays worked, and this broke user space parsing for it. The
update was to make the array look to have 8 entries in it. Helper
functions were added to make it parse it correctly, as the stack was
dynamic, but was determined by the size of the event stored.
Although this fixed user space on how it read the event, it changed the
internal structure used for the stack trace event. It changed the array
size from [0] to [8] (added 8 entries). This increased the size of the
stack trace event by 8 words. The size reserved on the ring buffer was the
size of the stack trace event plus the number of stack entries found in
the stack trace. That commit caused the amount to be 8 more than what was
needed because it did not expect the caller field to have any size. This
produced 8 entries of garbage (and reading random data) from the stack
trace event:
Instead, subtract the size of the caller field from the size of the event
to make sure that only the amount needed to store the stack trace is
reserved.
Merge tag 'tomoyo-pr-20210401' of git://git.osdn.net/gitroot/tomoyo/tomoyo-test1
Pull tomory fix from Tetsuo Handa:
"An update on 'tomoyo: recognize kernel threads correctly' from Jens
Axboe to not special case PF_IO_WORKER for PF_KTHREAD"
* tag 'tomoyo-pr-20210401' of git://git.osdn.net/gitroot/tomoyo/tomoyo-test1:
tomoyo: don't special case PF_IO_WORKER for PF_KTHREAD
Merge tag 'xarray-5.12' of git://git.infradead.org/users/willy/xarray
Pull XArray fixes from Matthew Wilcox:
"My apologies for the lateness of this. I had a bug reported in the
test suite, and when I started working on it, I realised I had two
fixes sitting in the xarray tree since last November. Anyway,
everything here is fixes, apart from adding xa_limit_16b. The test
suite passes.
Summary:
- Fix a bug when splitting to a non-zero order
- Documentation fix
- Add a predefined 16-bit allocation limit
- Various test suite fixes"
* tag 'xarray-5.12' of git://git.infradead.org/users/willy/xarray:
idr test suite: Improve reporting from idr_find_test_1
idr test suite: Create anchor before launching throbber
idr test suite: Take RCU read lock in idr_find_test_1
radix tree test suite: Register the main thread with the RCU library
radix tree test suite: Fix compilation
XArray: Add xa_limit_16b
XArray: Fix splitting to non-zero orders
XArray: Fix split documentation
Pavel Begunkov [Thu, 1 Apr 2021 11:18:48 +0000 (12:18 +0100)]
io_uring: fix EIOCBQUEUED iter revert
iov_iter_revert() is done in completion handlers that happensf before
read/write returns -EIOCBQUEUED, no need to repeat reverting afterwards.
Moreover, even though it may appear being just a no-op, it's actually
races with 1) user forging a new iovec of a different size 2) reissue,
that is done via io-wq continues completely asynchronously.
Fixes: 3e6a0d3c7571c ("io_uring: fix -EAGAIN retry with IOPOLL") Signed-off-by: Pavel Begunkov <[email protected]> Signed-off-by: Jens Axboe <[email protected]>
Pavel Begunkov [Thu, 1 Apr 2021 08:55:04 +0000 (09:55 +0100)]
io_uring/io-wq: protect against sprintf overflow
task_pid may be large enough to not fit into the left space of
TASK_COMM_LEN-sized buffers and overflow in sprintf. We not so care
about uniqueness, so replace it with safer snprintf().
io_uring: don't mark S_ISBLK async work as unbounded
S_ISBLK is marked as unbounded work for async preparation, because it
doesn't match S_ISREG. That is incorrect, as any read/write to a block
device is also a bounded operation. Fix it up and ensure that S_ISBLK
isn't marked unbounded.
Damien Le Moal [Wed, 31 Mar 2021 22:52:44 +0000 (07:52 +0900)]
null_blk: fix command timeout completion handling
Memory backed or zoned null block devices may generate actual request
timeout errors due to the submission path being blocked on memory
allocation or zone locking. Unlike fake timeouts or injected timeouts,
the request submission path will call blk_mq_complete_request() or
blk_mq_end_request() for these real timeout errors, causing a double
completion and use after free situation as the block layer timeout
handler executes blk_mq_rq_timed_out() and __blk_mq_free_request() in
blk_mq_check_expired(). This problem often triggers a NULL pointer
dereference such as:
This problem very often triggers when running the full btrfs xfstests
on a memory-backed zoned null block device in a VM with limited amount
of memory.
Avoid this by executing blk_mq_complete_request() in null_timeout_rq()
only for commands that are marked for a fake timeout completion using
the fake_timeout boolean in struct null_cmd. For timeout errors injected
through debugfs, the timeout handler will execute
blk_mq_complete_request()i as before. This is safe as the submission
path does not execute complete requests in this case.
In null_timeout_rq(), also make sure to set the command error field to
BLK_STS_TIMEOUT and to propagate this error through to the request
completion.
radix tree test suite: Register the main thread with the RCU library
Several test runners register individual worker threads with the
RCU library, but neglect to register the main thread, which can lead
to objects being freed while the main thread is in what appears to be
an RCU critical section.
Vitaly Kuznetsov [Wed, 24 Mar 2021 15:22:19 +0000 (16:22 +0100)]
ACPI: processor: Fix CPU0 wakeup in acpi_idle_play_dead()
Commit 496121c02127 ("ACPI: processor: idle: Allow probing on platforms
with one ACPI C-state") broke CPU0 hotplug on certain systems, e.g.
I'm observing the following on AWS Nitro (e.g r5b.xlarge but other
instance types are affected as well):
In fact, the above mentioned commit only revealed the problem and did
not introduce it. On x86, to wakeup CPU an NMI is being used and
hlt_play_dead()/mwait_play_dead() loops are prepared to handle it:
/*
* If NMI wants to wake up CPU0, start CPU0.
*/
if (wakeup_cpu0())
start_cpu0();
cpuidle_play_dead() -> acpi_idle_play_dead() (which is now being called on
systems where it wasn't called before the above mentioned commit) serves
the same purpose but it doesn't have a path for CPU0. What happens now on
wakeup is:
- NMI is sent to CPU0
- wakeup_cpu0_nmi() works as expected
- we get back to while (1) loop in acpi_idle_play_dead()
- safe_halt() puts CPU0 to sleep again.
The straightforward/minimal fix is add the special handling for CPU0 on x86
and that's what the patch is doing.
Fixes: 496121c02127 ("ACPI: processor: idle: Allow probing on platforms with one ACPI C-state") Signed-off-by: Vitaly Kuznetsov <[email protected]> Cc: 5.10+ <[email protected]> # 5.10+ Signed-off-by: Rafael J. Wysocki <[email protected]>
Vitaly Kuznetsov [Fri, 26 Mar 2021 15:55:51 +0000 (16:55 +0100)]
selftests: kvm: Check that TSC page value is small after KVM_SET_CLOCK(0)
Add a test for the issue when KVM_SET_CLOCK(0) call could cause
TSC page value to go very big because of a signedness issue around
hv_clock->system_time.
Vitaly Kuznetsov [Wed, 31 Mar 2021 12:41:29 +0000 (14:41 +0200)]
KVM: x86: Prevent 'hv_clock->system_time' from going negative in kvm_guest_time_update()
When guest time is reset with KVM_SET_CLOCK(0), it is possible for
'hv_clock->system_time' to become a small negative number. This happens
because in KVM_SET_CLOCK handling we set 'kvm->arch.kvmclock_offset' based
on get_kvmclock_ns(kvm) but when KVM_REQ_CLOCK_UPDATE is handled,
kvm_guest_time_update() does (masterclock in use case):
And 'master_kernel_ns' represents the last time when masterclock
got updated, it can precede KVM_SET_CLOCK() call. Normally, this is not a
problem, the difference is very small, e.g. I'm observing
hv_clock.system_time = -70 ns. The issue comes from the fact that
'hv_clock.system_time' is stored as unsigned and 'system_time / 100' in
compute_tsc_page_parameters() becomes a very big number.
Use 'master_kernel_ns' instead of get_kvmclock_ns() when masterclock is in
use and get_kvmclock_base_ns() when it's not to prevent 'system_time' from
going negative.
There is no need to include changes to vcpu->requests into
the pvclock_gtod_sync_lock critical section. The changes to
the shared data structures (in pvclock_update_vm_gtod_copy)
already occur under the lock.
projects / linux.git / log