linux-2.6-block.git
17 months agos390/cpum_cf: introduce static CPU counter facility information
Thomas Richter [Tue, 4 Apr 2023 12:47:55 +0000 (14:47 +0200)]
s390/cpum_cf: introduce static CPU counter facility information

The CPU measurement facility counter information instruction qctri()
retrieves information about the available counter sets.
The information varies between machine generations, but is constant
when running on a particular machine.
For example the CPU measurement facility counter first and second
version numbers determine the amount of counters in a counter
set. This information never changes.

The counter sets are identical for all CPUs in the system. It does
not matter which CPU performs the instruction.

Authorization control of the CPU Measurement facility can only
be changed in the activation profile while the LPAR is not running.

Retrieve the CPU measurement counter information at device driver
initialization time and use its constant values.

Function validate_ctr_version() verifies if a user provided
CPU Measurement counter facility counter is valid and defined.
It now uses the newly introduced static CPU counter facility
information.

To avoid repeated recalculation of the counter set sizes (numbers of
counters per set), which never changes on a running machine,
calculate the counter set size once at device driver initialization
and store the result in an array. Functions cpum_cf_make_setsize()
and cpum_cf_read_setsize() are introduced.

Finally remove cpu_cf_events::info member and use the static CPU
counter facility information instead.

Signed-off-by: Thomas Richter <tmricht@linux.ibm.com>
Acked-by: Heiko Carstens <hca@linux.ibm.com>
Acked-by: Sumanth Korikkar <sumanthk@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/zcrypt: rework arrays with length zero occurrences
Harald Freudenberger [Fri, 14 Apr 2023 15:33:19 +0000 (17:33 +0200)]
s390/zcrypt: rework arrays with length zero occurrences

Review and rework all the zero length array occurrences
within structs to flexible array fields or comment if
not used at all. However, some struct fields are there
for documentation purpose or to have correct sizeof()
evaluation of a struct and thus should not get deleted.

Signed-off-by: Harald Freudenberger <freude@linux.ibm.com>
Acked-by: Heiko Carstens <hca@linux.ibm.com>
Reviewed-by: Holger Dengler <dengler@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/cio: replace zero-length array with flexible-array member
Heiko Carstens [Tue, 11 Apr 2023 09:18:59 +0000 (11:18 +0200)]
s390/cio: replace zero-length array with flexible-array member

There are numerous patches which convert zero-length arrays with a
flexible-array member. Convert the remaining s390 occurrences.

Suggested-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Link: https://github.com/KSPP/linux/issues/78
Link: https://gcc.gnu.org/pipermail/gcc-patches/2022-October/602902.html
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/sclp: replace zero-length array with flexible-array member
Heiko Carstens [Tue, 11 Apr 2023 09:17:42 +0000 (11:17 +0200)]
s390/sclp: replace zero-length array with flexible-array member

There are numerous patches which convert zero-length arrays with a
flexible-array member. Convert the remaining s390 occurrences.

Suggested-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Link: https://github.com/KSPP/linux/issues/78
Link: https://gcc.gnu.org/pipermail/gcc-patches/2022-October/602902.html
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/debug: replace zero-length array with flexible-array member
Heiko Carstens [Tue, 11 Apr 2023 09:13:01 +0000 (11:13 +0200)]
s390/debug: replace zero-length array with flexible-array member

There are numerous patches which convert zero-length arrays with a
flexible-array member. Convert the remaining s390 occurrences.

Suggested-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Link: https://github.com/KSPP/linux/issues/78
Link: https://gcc.gnu.org/pipermail/gcc-patches/2022-October/602902.html
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/fcx: replace zero-length array with flexible-array member
Gustavo A. R. Silva [Thu, 6 Apr 2023 14:29:35 +0000 (08:29 -0600)]
s390/fcx: replace zero-length array with flexible-array member

Zero-length arrays are deprecated [1] and have to be replaced by C99
flexible-array members.

This helps with the ongoing efforts to tighten the FORTIFY_SOURCE routines
on memcpy() and help to make progress towards globally enabling
-fstrict-flex-arrays=3 [2]

Link: https://github.com/KSPP/linux/issues/78
Link: https://gcc.gnu.org/pipermail/gcc-patches/2022-October/602902.html
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Link: https://lore.kernel.org/r/ZC7XT5prvoE4Yunm@work
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/diag: replace zero-length array with flexible-array member
Gustavo A. R. Silva [Thu, 6 Apr 2023 14:28:42 +0000 (08:28 -0600)]
s390/diag: replace zero-length array with flexible-array member

Zero-length arrays are deprecated [1] and have to be replaced by C99
flexible-array members.

This helps with the ongoing efforts to tighten the FORTIFY_SOURCE routines
on memcpy() and help to make progress towards globally enabling
-fstrict-flex-arrays=3 [2]

Link: https://github.com/KSPP/linux/issues/78
Link: https://gcc.gnu.org/pipermail/gcc-patches/2022-October/602902.html
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Link: https://lore.kernel.org/r/ZC7XGpUtVhqlRLhH@work
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/mm: fix direct map accounting
Heiko Carstens [Thu, 6 Apr 2023 11:31:29 +0000 (13:31 +0200)]
s390/mm: fix direct map accounting

Commit bb1520d581a3 ("s390/mm: start kernel with DAT enabled") did not
implement direct map accounting in the early page table setup code. In
result the reported values are bogus now:

$cat /proc/meminfo
...
DirectMap4k:        5120 kB
DirectMap1M:    18446744073709546496 kB
DirectMap2G:           0 kB

Fix this by adding the missing accounting. The result looks sane again:

$cat /proc/meminfo
...
DirectMap4k:        6156 kB
DirectMap1M:     2091008 kB
DirectMap2G:     6291456 kB

Fixes: bb1520d581a3 ("s390/mm: start kernel with DAT enabled")
Reviewed-by: Alexander Gordeev <agordeev@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/mm: rename POPULATE_ONE2ONE to POPULATE_DIRECT
Heiko Carstens [Thu, 6 Apr 2023 11:31:28 +0000 (13:31 +0200)]
s390/mm: rename POPULATE_ONE2ONE to POPULATE_DIRECT

Architectures generally use the "direct map" wording for mapping the whole
physical memory. Use that wording as well in arch/s390/boot/vmem.c, instead
of "one to one" in order to avoid confusion.

This also matches what is already done in arch/s390/mm/vmem.c.

Reviewed-by: Alexander Gordeev <agordeev@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/boot: improve install.sh script
Marc Hartmayer [Mon, 31 Jan 2022 13:22:31 +0000 (13:22 +0000)]
s390/boot: improve install.sh script

Use proper quoting for the variables and explicitly distinguish between
command options and positional arguments.

Acked-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/cpum_cf: simplify pr_err() statement in cpumf_pmu_enable/disable
Thomas Richter [Fri, 24 Feb 2023 08:01:51 +0000 (09:01 +0100)]
s390/cpum_cf: simplify pr_err() statement in cpumf_pmu_enable/disable

Simplify pr_err() statement into one line and omit return statement.

No functional change.

Signed-off-by: Thomas Richter <tmricht@linux.ibm.com>
Acked-by: Sumanth Korikkar <sumanthk@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/kaslr: randomize amode31 base address
Vasily Gorbik [Wed, 15 Mar 2023 12:54:14 +0000 (13:54 +0100)]
s390/kaslr: randomize amode31 base address

When the KASLR is enabled, randomize the base address of the amode31 image
within the first 2 GB, similar to the approach taken for the vmlinux
image. This makes it harder to predict the location of amode31 data
and code.

Reviewed-by: Alexander Gordeev <agordeev@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/kaslr: generalize and improve random base distribution
Vasily Gorbik [Tue, 21 Feb 2023 22:08:42 +0000 (23:08 +0100)]
s390/kaslr: generalize and improve random base distribution

Improve the distribution algorithm of random base address to ensure
a uniformity among all suitable addresses. To generate a random value
once, and to build a continuous range in which every value is suitable,
count all the suitable addresses (referred to as positions) that can be
used as a base address. The positions are counted by iterating over the
usable memory ranges. For each range that is big enough to accommodate
the image, count all the suitable addresses where the image can be placed,
while taking reserved memory ranges into consideration.

A new function "iterate_valid_positions()" has dual purpose. Firstly, it
is called to count the positions in a given memory range, and secondly,
to convert a random position back to an address.

"get_random_base()" has been replaced with more generic
"randomize_within_range()" which now could be called for randomizing
base addresses not just for the kernel image.

Acked-by: Alexander Gordeev <agordeev@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/boot: pin amode31 default lma
Vasily Gorbik [Wed, 15 Mar 2023 10:00:19 +0000 (11:00 +0100)]
s390/boot: pin amode31 default lma

The special amode31 part of the kernel must always remain below 2Gb. Place
it just under vmlinux.default_lma by default, which makes it easier to
debug amode31 as its default lma is known 0x10000 - 0x3000 (currently,
amode31's size is 3 pages). This location is always available as it is
originally occupied by the vmlinux archive.

Reviewed-by: Alexander Gordeev <agordeev@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/boot: do not change default_lma
Vasily Gorbik [Wed, 15 Mar 2023 09:19:57 +0000 (10:19 +0100)]
s390/boot: do not change default_lma

The current modification of the default_lma is illogical and should be
avoided. It would be more appropriate to introduce and utilize a new
variable vmlinux_lma instead, so that default_lma remains unchanged and
at its original "default" value of 0x100000.

Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/cpum_cf: remove unnecessary copy_from_user call
Thomas Richter [Fri, 31 Mar 2023 09:42:20 +0000 (11:42 +0200)]
s390/cpum_cf: remove unnecessary copy_from_user call

Struct s390_ctrset_read userdata is filled by ioctl_read operation
using put_user/copy_to_user. However, the ctrset->data value access
is not performed anywhere during the ioctl_read operation.
Remove unnecessary copy_from_user() call.
No functional change.

Signed-off-by: Thomas Richter <tmricht@linux.ibm.com>
Suggested-by: Sumanth Korikkar <sumanthk@linux.ibm.com>
Acked-by: Sumanth Korikkar <sumanthk@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/cpum_cf: log bad return code of function cfset_all_copy
Thomas Richter [Fri, 24 Feb 2023 07:55:21 +0000 (08:55 +0100)]
s390/cpum_cf: log bad return code of function cfset_all_copy

When function cfset_all_copy() fails, also log the bad return code
in the debug statement (when turned on).
No functional change

Signed-off-by: Thomas Richter <tmricht@linux.ibm.com>
Acked-by: Sumanth Korikkar <sumanthk@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/module: create module allocations without exec permissions
Heiko Carstens [Sun, 2 Apr 2023 18:55:21 +0000 (20:55 +0200)]
s390/module: create module allocations without exec permissions

This is the s390 variant of commit 7dfac3c5f40e ("arm64: module: create
module allocations without exec permissions"):

"The core code manages the executable permissions of code regions of
modules explicitly. It is no longer necessary to create the module vmalloc
regions with RWX permissions. So create them with RW- permissions instead,
which is preferred from a security perspective."

Reviewed-by: Alexander Gordeev <agordeev@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/ftrace: do not assume module_alloc() returns executable memory
Heiko Carstens [Sun, 2 Apr 2023 18:55:19 +0000 (20:55 +0200)]
s390/ftrace: do not assume module_alloc() returns executable memory

The ftrace code assumes at two places that module_alloc() returns
executable memory. While this is currently true, this will be changed
with a subsequent patch to follow other architectures which implement
ARCH_HAS_STRICT_MODULE_RWX.

Acked-by: Alexander Gordeev <agordeev@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/mm: use set_memory_*() helpers instead of open coding
Heiko Carstens [Sun, 2 Apr 2023 18:55:18 +0000 (20:55 +0200)]
s390/mm: use set_memory_*() helpers instead of open coding

Given that set_memory_rox() and set_memory_rwnx() exist, it is possible
to get rid of all open coded __set_memory() usages and replace them with
proper helper calls everywhere.

Reviewed-by: Alexander Gordeev <agordeev@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/mm: implement set_memory_rwnx()
Heiko Carstens [Sun, 2 Apr 2023 18:55:17 +0000 (20:55 +0200)]
s390/mm: implement set_memory_rwnx()

Given that set_memory_rox() is implemented, provide also set_memory_rwnx().
This allows to get rid of all open coded __set_memory() usages in s390
architecture code.

Reviewed-by: Alexander Gordeev <agordeev@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/mm: implement set_memory_rox()
Heiko Carstens [Sun, 2 Apr 2023 18:55:16 +0000 (20:55 +0200)]
s390/mm: implement set_memory_rox()

Provide the s390 specific native set_memory_rox() implementation to avoid
frequent set_memory_ro(); set_memory_x() call pairs.

This is the s390 variant of commit 60463628c9e0 ("x86/mm: Implement native
set_memory_rox()").

Reviewed-by: Alexander Gordeev <agordeev@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/ipl: fix physical-virtual confusion for diag308
Nico Boehr [Fri, 10 Mar 2023 11:52:37 +0000 (12:52 +0100)]
s390/ipl: fix physical-virtual confusion for diag308

Diag 308 subcodes expect a physical address as their parameter.

This currently is not a bug, but in the future physical and virtual
addresses might differ.

Fix the confusion by doing a virtual-to-physical conversion in the
exported diag308() and leave the assembly wrapper __diag308() alone.

Note that several callers pass NULL as addr, so check for the case when
NULL is passed and pass 0 to hardware since virt_to_phys(0) might be
nonzero.

Suggested-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Signed-off-by: Nico Boehr <nrb@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/kaslr: randomize module base load address
Heiko Carstens [Fri, 31 Mar 2023 13:03:23 +0000 (15:03 +0200)]
s390/kaslr: randomize module base load address

Randomize the load address of modules in the kernel to make KASLR effective
for modules.
This is the s390 variant of commit e2b32e678513 ("x86, kaslr: randomize
module base load address").

Reviewed-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/kaslr: provide kaslr_enabled() function
Heiko Carstens [Fri, 31 Mar 2023 13:03:22 +0000 (15:03 +0200)]
s390/kaslr: provide kaslr_enabled() function

Just like other architectures provide a kaslr_enabled() function, instead
of directly accessing a global variable.

Also pass the renamed __kaslr_enabled variable from the decompressor to the
kernel, so that kalsr_enabled() is available there too. This will be used
by a subsequent patch which randomizes the module base load address.

Reviewed-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/checksum: remove not needed uaccess.h include
Heiko Carstens [Thu, 30 Mar 2023 10:55:47 +0000 (12:55 +0200)]
s390/checksum: remove not needed uaccess.h include

Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/checksum: always use cksm instruction
Heiko Carstens [Thu, 30 Mar 2023 10:55:46 +0000 (12:55 +0200)]
s390/checksum: always use cksm instruction

Commit dfe843dce775 ("s390/checksum: support GENERIC_CSUM, enable it for
KASAN") switched s390 to use the generic checksum functions, so that KASAN
instrumentation also works checksum functions by avoiding architecture
specific inline assemblies.

There is however the problem that the generic csum_partial() function
returns a 32 bit value with a 16 bit folded checksum, while the original
s390 variant does not fold to 16 bit. This in turn causes that the
ipib_checksum in lowcore contains different values depending on kernel
config options.

The ipib_checksum is used by system dumpers to verify if pointers in
lowcore point to valid data. Verification is done by comparing checksum
values. The system dumpers still use 32 bit checksum values which are not
folded, and therefore the checksum verification fails (incorrectly).

Symptom is that reboot after dump does not work anymore when a KASAN
instrumented kernel is dumped.

Fix this by not using the generic checksum implementation. Instead add an
explicit kasan_check_read() so that KASAN knows about the read access from
within the inline assembly.

Reported-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Fixes: dfe843dce775 ("s390/checksum: support GENERIC_CSUM, enable it for KASAN")
Tested-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390: enable HAVE_ARCH_STACKLEAK
Heiko Carstens [Mon, 27 Mar 2023 09:37:28 +0000 (11:37 +0200)]
s390: enable HAVE_ARCH_STACKLEAK

Add support for the stackleak feature. Whenever the kernel returns to user
space the kernel stack is filled with a poison value.

Enabling this feature is quite expensive: e.g. after instrumenting the
getpid() system call function to have a 4kb stack the result is an
increased runtime of the system call by a factor of 3.

Reviewed-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390: move on_thread_stack() to processor.h
Heiko Carstens [Mon, 27 Mar 2023 09:37:27 +0000 (11:37 +0200)]
s390: move on_thread_stack() to processor.h

As preparation for the stackleak feature move on_thread_stack() to
processor.h like x86.

Also make it __always_inline, and slightly optimize it by reading
current task's kernel stack pointer from lowcore.

Reviewed-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390: remove arch_early_irq_init()
Heiko Carstens [Mon, 27 Mar 2023 09:37:26 +0000 (11:37 +0200)]
s390: remove arch_early_irq_init()

Allocate early async stack like other early stacks and get rid of
arch_early_irq_init(). This way the async stack is allocated earlier,
and handled like all other stacks.

Reviewed-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/stacktrace: remove call_on_stack_noreturn()
Heiko Carstens [Mon, 27 Mar 2023 09:37:25 +0000 (11:37 +0200)]
s390/stacktrace: remove call_on_stack_noreturn()

There is no user left of call_on_stack_noreturn() - remove it.

Reviewed-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390: use init_thread_union aka initial stack for the first process
Heiko Carstens [Mon, 27 Mar 2023 09:37:24 +0000 (11:37 +0200)]
s390: use init_thread_union aka initial stack for the first process

s390 is the only architecture which switches from the initial stack to a
later on allocated different stack for the first process.
This is (at least) problematic for the stackleak feature, which instruments
functions to save the current stackpointer within the task structure of the
running process.

The stackleak code compares stack pointers of the current process - and
doesn't expect that the kernel stack of a task can change. Even though the
stackleak feature itself will not cause any harm, the assumption about
kernel stacks being consistent is there, and only s390 doesn't follow that.

Therefore switch back to use init_thread_union, just like all other
architectures.

Reviewed-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/stack: set lowcore kernel stack pointer early
Heiko Carstens [Mon, 27 Mar 2023 09:37:23 +0000 (11:37 +0200)]
s390/stack: set lowcore kernel stack pointer early

Make sure the lowcore kernel stack pointer reflects the kernel stack of the
current task as early as possible, instead of having a NULL pointer there.

Reviewed-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/stack: use STACK_INIT_OFFSET where possible
Heiko Carstens [Mon, 27 Mar 2023 09:37:22 +0000 (11:37 +0200)]
s390/stack: use STACK_INIT_OFFSET where possible

Make STACK_INIT_OFFSET also available for assembler code, and
use it everywhere instead of open-coding it at several places.

Reviewed-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/dumpstack: simplify in stack logic code
Heiko Carstens [Mon, 27 Mar 2023 09:37:21 +0000 (11:37 +0200)]
s390/dumpstack: simplify in stack logic code

The pattern for all in_<type>_stack() functions is the same; especially
also the size of all stacks is the same. Simplify the code by passing only
the stack address to the generic in_stack() helper, which then can assume a
THREAD_SIZE sized stack.

Reviewed-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/zcrypt: simplify prep of CCA key token
Harald Freudenberger [Wed, 22 Mar 2023 14:22:11 +0000 (15:22 +0100)]
s390/zcrypt: simplify prep of CCA key token

The preparation of the key data struct for a CCA RSA ME
operation had some improvement to skip leading zeros
in the key's exponent. However, all supported CCA cards
nowadays support leading zeros in key tokens.

So for simplifying the CCA key preparing code, this
patch simply removes this optimization code.

Signed-off-by: Harald Freudenberger <freude@linux.ibm.com>
Reviewed-by: Juergen Christ <jchrist@linux.ibm.com>
Reviewed-by: Holger Dengler <dengler@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/zcrypt: remove unused ancient padding code
Harald Freudenberger [Wed, 22 Mar 2023 14:15:08 +0000 (15:15 +0100)]
s390/zcrypt: remove unused ancient padding code

There was some ancient code which padded the results of
a clear key ME or CRT operation with some PKCS 1.2 header.
According to the comment this was only needed by crypto
cards older than the CEX2. These cards are not supported
any more and so this patch removes this obscure result
padding code.

Signed-off-by: Harald Freudenberger <freude@linux.ibm.com>
Reviewed-by: Juergen Christ <jchrist@linux.ibm.com>
Reviewed-by: Holger Dengler <dengler@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390: make extables read-only
Vasily Gorbik [Tue, 28 Mar 2023 08:00:33 +0000 (10:00 +0200)]
s390: make extables read-only

Currently, exception tables are marked as ro_after_init. However,
since they are sorted during compile time using scripts/sorttable,
they can be moved to RO_DATA using the RO_EXCEPTION_TABLE_ALIGN macro,
which is specifically designed for this purpose.

Suggested-by: Heiko Carstens <hca@linux.ibm.com>
Reviewed-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/entry: rely on long-displacement facility
Vasily Gorbik [Tue, 28 Mar 2023 09:09:09 +0000 (11:09 +0200)]
s390/entry: rely on long-displacement facility

Since commit 4efd417f298b ("s390: raise minimum supported machine
generation to z10"), the long-displacement facility is assumed and
required for the kernel. Clean up a couple of places in the entry code,
where long-displacement could be used directly instead of using a base
register.

However, there are still a few other places where a base register has
to be used to extend short-displacement for the second lowcore page
access. Notably, boot/head.S still has to be built for z900, and in
mcck_int_handler, spt and lbear, which don't have long-displacements,
but need to access save areas at the second lowcore page.

Reviewed-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agoMerge branch 'uaccess-inline-asm-cleanup' into features
Vasily Gorbik [Tue, 4 Apr 2023 16:29:28 +0000 (18:29 +0200)]
Merge branch 'uaccess-inline-asm-cleanup' into features

Heiko Carstens says:

===================
There are a couple of oddities within the s390 uaccess library
functions. Therefore cleanup the whole uaccess.c file.

There is no functional change, only improved readability. The output
of "objdump -Dr" was always compared before/after each patch to make
sure that the generated object file is identical, if that could be
expected. Therefore the series also includes more patches than really
required to cleanup the code.

Furthermore the kunit usercopy tests also still pass.
===================

Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/uaccess: remove extra blank line
Heiko Carstens [Fri, 24 Mar 2023 14:00:24 +0000 (15:00 +0100)]
s390/uaccess: remove extra blank line

In order to get uaccess.c (nearly) checkpatch warning free remove an
extra blank line:

CHECK: Blank lines aren't necessary before a close brace '}'
+
+}

Reviewed-by: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/uaccess: get rid of not needed local variable
Heiko Carstens [Fri, 24 Mar 2023 14:00:23 +0000 (15:00 +0100)]
s390/uaccess: get rid of not needed local variable

Get rid of the not needed val local variable and pass the constant
value directly as operand value. In addition this turns the val
operand into an input operand, since it is not changed within the
inline assemblies.

This in turn requires also to add the earlyclobber contraint modifier
to all output operands, since the (former) val operand is used after
all output variants have been modified.

The usercopy kunit tests still pass after this change.

Reviewed-by: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/uaccess: rename tmp1 and tmp2 variables
Heiko Carstens [Fri, 24 Mar 2023 14:00:22 +0000 (15:00 +0100)]
s390/uaccess: rename tmp1 and tmp2 variables

Rename tmp1 and tmp2 variables to more meaningful val (for value) and rem
(for remainder).

Except for debug sections the output of "objdump -Dr" of the uaccess object
file is identical before/after this change.

Reviewed-by: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/uaccess: sort EX_TABLE list for inline assemblies
Heiko Carstens [Fri, 24 Mar 2023 14:00:21 +0000 (15:00 +0100)]
s390/uaccess: sort EX_TABLE list for inline assemblies

Reviewed-by: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/uaccess: rename/sort labels in inline assemblies
Heiko Carstens [Fri, 24 Mar 2023 14:00:20 +0000 (15:00 +0100)]
s390/uaccess: rename/sort labels in inline assemblies

Rename and sort labels in uaccess inline assemblies to increase
readability. In addition have only one EX_TABLE entry per line - also to
increase readability.

Except for debug sections the output of "objdump -Dr" of the uaccess object
file is identical before/after this change.

Reviewed-by: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/uaccess: remove unused label in inline assemblies
Heiko Carstens [Fri, 24 Mar 2023 14:00:19 +0000 (15:00 +0100)]
s390/uaccess: remove unused label in inline assemblies

Remove an unused label in all three uaccess inline assemblies.

Reviewed-by: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/uaccess: use symbolic names for inline assembly operands
Heiko Carstens [Fri, 24 Mar 2023 14:00:18 +0000 (15:00 +0100)]
s390/uaccess: use symbolic names for inline assembly operands

Improve readability of the uaccess inline assemblies by using symbolic
names for all input and output operands.

Except for debug sections the output of "objdump -Dr" of the uaccess object
file is identical before/after this change.

Reviewed-by: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/uaccess: add missing earlyclobber annotations to __clear_user()
Heiko Carstens [Thu, 23 Mar 2023 12:09:16 +0000 (13:09 +0100)]
s390/uaccess: add missing earlyclobber annotations to __clear_user()

Add missing earlyclobber annotation to size, to, and tmp2 operands of the
__clear_user() inline assembly since they are modified or written to before
the last usage of all input operands. This can lead to incorrect register
allocation for the inline assembly.

Fixes: 6c2a9e6df604 ("[S390] Use alternative user-copy operations for new hardware.")
Reported-by: Mark Rutland <mark.rutland@arm.com>
Link: https://lore.kernel.org/all/20230321122514.1743889-3-mark.rutland@arm.com/
Cc: stable@vger.kernel.org
Reviewed-by: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/vfio-ap: remove redundant driver match function
Lizhe [Sun, 19 Mar 2023 04:19:41 +0000 (12:19 +0800)]
s390/vfio-ap: remove redundant driver match function

If there is no driver match function, the driver core assumes that each
candidate pair (driver, device) matches, see driver_match_device().

Drop the matrix bus's match function that always returned 1 and so
implements the same behaviour as when there is no match function

Signed-off-by: Lizhe <sensor1010@163.com>
Reviewed-by: Tony Krowiak <akrowiak@linux.ibm.com>
Link: https://lore.kernel.org/r/20230319041941.259830-1-sensor1010@163.com
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390: enable ARCH_HAS_MEMBARRIER_SYNC_CORE
Heiko Carstens [Fri, 17 Mar 2023 09:31:40 +0000 (10:31 +0100)]
s390: enable ARCH_HAS_MEMBARRIER_SYNC_CORE

s390 trivially supports the ARCH_HAS_MEMBARRIER_SYNC_CORE requirements
since the used lpswe(y) instruction to return from any kernel context to
user space performs CPU serialization. This is very similar to arm, arm64
and powerpc.

See commit 70216e18e519 ("membarrier: Provide core serializing command,
*_SYNC_CORE") for further details.

Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
17 months agos390/cpum_sf: remove flag PERF_CPUM_SF_FULL_BLOCKS
Thomas Richter [Thu, 16 Mar 2023 15:34:46 +0000 (16:34 +0100)]
s390/cpum_sf: remove flag PERF_CPUM_SF_FULL_BLOCKS

This flag is used to process only fully populated sampling buffers
when an sampling event is stopped on a CPU. By default the last sampling
buffer is also scanned for samples even if the sampling block full
indicator is not set in the trailer entry of a sampling buffer page.

This flag can be set via perf_event_attr::config1 field. It was never
used and never documented. It is useless now.

With PERF_CPUM_SF_FULL_BLOCKS:
When a process is scheduled off the CPU, the sampling is stopped and
the samples are copied to the perf ring buffer and marked invalid.
When stopped at the last full sample buffer page (which is
achieved with the PERF_CPUM_SF_FULL_BLOCKS options), the hardware
sampling will resume at the first free sample entry in the current,
partially filled sample buffer.

Without PERF_CPUM_SF_FULL_BLOCKS (default behavior):
The partially filled last sample buffer is scanned and valid samples
are saved to the perf ring buffer. The valid samples are marked invalid.
The sampling is resumed when the process is scheduled on this CPU.
Again the hardware sampling will resume at the first free sample entry in
the current, partially filled sample buffer.

Now the next interrupt handler invocation scans the
full sample block and saves the valid samples to the ring buffer.
It omits the invalid samples at the top of the buffer.
The default behavior is fully sufficient, therefore remove this feature.

Signed-off-by: Thomas Richter <tmricht@linux.ibm.com>
Acked-by: Hendrik Brueckner <brueckner@linux.ibm.com>
Acked-by: Sumanth Korikkar <sumanthk@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
18 months agos390/mm: make use of atomic_fetch_xor()
Heiko Carstens [Fri, 17 Mar 2023 13:51:41 +0000 (14:51 +0100)]
s390/mm: make use of atomic_fetch_xor()

Make use of atomic_fetch_xor() instead of an atomic_cmpxchg() loop to
implement atomic_xor_bits() (aka atomic_xor_return()). This makes the C
code more readable and in addition generates better code, since for z196
and newer a single lax instruction is generated instead of a cmpxchg()
loop.

Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
18 months agos390/ap: add ap status asynch error support
Harald Freudenberger [Wed, 15 Feb 2023 14:08:15 +0000 (15:08 +0100)]
s390/ap: add ap status asynch error support

Review and extend the low level AP code to be able to
deal with asynchronous reported errors on APQNs.

The hypervisor and the SE guest may be confronted with
an asynchronously reported error at return of an AP
instruction. So all places where AP instructions are
called need review and may eventually need extensions.
However, not all places need rework. As together with
the AP status and the enabled asynch bit there is always
a response code set. The asynch error reporting comes
with new response codes which may be simple handled in
the default case of a switch statement.

The idea behind this patch is to report asynch errors
as -EPERM (read this as "Operation not permitted") which
reflects the fact that only a rapq (with F bit enabled)
is a valid AP instruction when an asynch error is flagged.

The AP queue state machine functions return
AP_SM_WAIT_NONE when a asynch error is detected to reflect
the fact, that the state machine can't do anything with
such an error as long as the queue is reset.

Unfortunately the ap bus scan function needed some
update as the ap_queue_info() now needs to return
3 states: 1 if an APQN exists and info is available,
-1 if it is assumed an APQN does not exist and the new
return value 0 without any info values filled. This 0
returncode is handled as "there is an APQN but we currently
don't know any more hw info about this, so please use
your previous info and try again later".

Signed-off-by: Harald Freudenberger <freude@linux.ibm.com>
Reviewed-by: Holger Dengler <dengler@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
18 months agos390/ap: implement SE AP bind, unbind and associate
Harald Freudenberger [Fri, 10 Mar 2023 16:46:49 +0000 (17:46 +0100)]
s390/ap: implement SE AP bind, unbind and associate

Implementation of the new functions for SE AP support:
bind, unbind and associate. There are two new sysfs
attributes for this:

/sys/devices/ap/cardxx/xx.yyyy/se_bind
/sys/devices/ap/cardxx/xx.yyyy/se_associate

Writing a 1 into the se_bind attribute triggers the
SE AP bind for this AP queue, writing a 0 into does
an unbind - that's a reset (RAPQ) with the F bit enabled.

The se_associate attribute needs an integer value in
range 0...2^16-1 written in. This is the index into a
secrets table feed into the ultravisor. For more details
please see the Architecture documents.

These both new ap queue attributes are only visible
inside a SE guest with SB (Secure Binding) available.

Signed-off-by: Harald Freudenberger <freude@linux.ibm.com>
Reviewed-by: Holger Dengler <dengler@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
18 months agos390/ap: introduce low frequency polling possibility
Harald Freudenberger [Fri, 21 Oct 2022 13:41:00 +0000 (15:41 +0200)]
s390/ap: introduce low frequency polling possibility

For some events the ap bus needs to poll. For example
when an AP queue is reset until the reset is through.
Also when no interrupt support is available (e.g. zVM)
there is a need to poll until all requests have been
processed and all replies have been delivered.

Polling is done with a high resolution timer by default
run with a rate of 4kHz (LPAR) or 666Hz (zVM guest).

For some events (wait for reset complete, wait for irq
enabled complete) this is a much too high poll rate
which triggers a lot of TAPQ invocations.

This patch introduces the possibility for the state
machine functions to return a new wait enum
AP_SM_WAIT_LOW_TIMEOUT which gives a hint to the
ap_wait() function to eventually set up the timer
with a more relaxed timeout value of 25Hz.

This patch also includes a slight rework of the sysfs
functions parsing the timer related stuff: Use of
kstrtobool and kstrtoul instead of sscanf.

Signed-off-by: Harald Freudenberger <freude@linux.ibm.com>
Reviewed-by: Holger Dengler <dengler@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
18 months agos390/ap: new low level inline functions ap_bapq() and ap_aapq()
Harald Freudenberger [Wed, 7 Sep 2022 17:05:18 +0000 (19:05 +0200)]
s390/ap: new low level inline functions ap_bapq() and ap_aapq()

Introduce two new low level functions ap_bapq() (calls
PQAP(BAPQ)) and ap_aapq (calls PQAP(AAPQ)). Both functions
are only meant to be used in SE environment with the SE
AP binding facility available.

Signed-off-by: Harald Freudenberger <freude@linux.ibm.com>
Reviewed-by: Holger Dengler <dengler@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
18 months agos390/ap: provide F bit parameter for ap_rapq() and ap_zapq()
Harald Freudenberger [Wed, 7 Sep 2022 16:04:03 +0000 (18:04 +0200)]
s390/ap: provide F bit parameter for ap_rapq() and ap_zapq()

Extent the ap inline functions ap_rapq() (calls PQAP(RAPQ))
and ap_zapq() (calls PQAP(ZAPQ)) with a new parameter to
enable the new architectured F bit which forces an
unassociate and/or unbind on a secure execution associated
and/or bound queue.

Signed-off-by: Harald Freudenberger <freude@linux.ibm.com>
Reviewed-by: Tony Krowiak <akrowiak@linux.ibm.com>
Reviewed-by: Holger Dengler <dengler@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
18 months agos390/ap: filter ap card functions, new queue functions attribute
Harald Freudenberger [Wed, 7 Sep 2022 15:25:45 +0000 (17:25 +0200)]
s390/ap: filter ap card functions, new queue functions attribute

With SE SB (Secure Binding) some currently unused and thus always
zero bits in the TAPQ GR2 result are now used to show the binding
state of a queue. So to check if a card has changed the comparing
base is exactly this GR2 value shown as 'ap_function' in sysfs
(/sys/devices/ap/cardxx/ap_functions). Now there is some queue
specific info in this info and so a new mask TAPQ_CARD_FUNC_CMP_MASK
is used to filter out only the relevant bits for card compare.

For the same reason now the function bits (including exactly this
bind/associate information) need to be exposed to user space now.
So tools like lszcrypt can evaluate binding/association state on a
queue base. So here comes a new sysfs attribute

  /sys/devices/ap/cardxx/xx.yyyy/ap_functions

This sysfs attribute is similar to the already existing
ap_functions attribute at ap card level. It shows the
upper 32 bits of GR2 from an invocation of TAPQ for this
AP queue.

Signed-off-by: Harald Freudenberger <freude@linux.ibm.com>
Reviewed-by: Holger Dengler <dengler@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
18 months agos390/ap: make tapq gr2 response a struct
Harald Freudenberger [Mon, 12 Sep 2022 16:02:44 +0000 (18:02 +0200)]
s390/ap: make tapq gr2 response a struct

This patch introduces a new struct ap_tapq_gr2 which covers
the response in GR2 on TAPQ invocation. This makes it much
easier and less error-prone for the calling functions to
access the right field without shifting and masking.

Signed-off-by: Harald Freudenberger <freude@linux.ibm.com>
Reviewed-by: Tony Krowiak <akrowiak@linux.ibm.com>
Reviewed-by: Holger Dengler <dengler@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
18 months agos390/ap: introduce new AP bus sysfs attribute features
Harald Freudenberger [Tue, 20 Sep 2022 13:18:39 +0000 (15:18 +0200)]
s390/ap: introduce new AP bus sysfs attribute features

Introduce a new AP bus sysfs attribute /sys/bus/ap/features
which shows the features from the QCI information.
Currently these feature bits are evaluated:

- QCI S bit is shown as 'APSC'
- QCI N bit is shown as 'APXA'
- QCI C bit is shown as 'QACT'
- QCI R bit is shown as 'RC8A'
- QCI B bit is shown as 'APSB'

Signed-off-by: Harald Freudenberger <freude@linux.ibm.com>
Reviewed-by: Tony Krowiak <akrowiak@linux.ibm.com>
Reviewed-by: Holger Dengler <dengler@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
18 months agos390/ap: exploit new B bit from QCI config info
Harald Freudenberger [Tue, 20 Sep 2022 13:12:01 +0000 (15:12 +0200)]
s390/ap: exploit new B bit from QCI config info

This patch introduces an update to the ap_config_info
struct which is filled with the QCI subfunction. There
is a new bit apsb (short 'B') showing if the AP secure
bind facility is available. The patch also includes a
simple function ap_sb_available() wrapping this bit test.

Signed-off-by: Harald Freudenberger <freude@linux.ibm.com>
Reviewed-by: Tony Krowiak <akrowiak@linux.ibm.com>
Reviewed-by: Holger Dengler <dengler@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
18 months agos390/zcrypt: replace scnprintf with sysfs_emit
Harald Freudenberger [Mon, 6 Feb 2023 09:53:08 +0000 (10:53 +0100)]
s390/zcrypt: replace scnprintf with sysfs_emit

Replace scnprintf() with sysfs_emit() and friends
where possible.

Signed-off-by: Harald Freudenberger <freude@linux.ibm.com>
Reviewed-by: Holger Dengler <dengler@linux.ibm.com>
Reviewed-by: Tony Krowiak <akrowiak@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
18 months agos390/zcrypt: rework length information for dqap
Harald Freudenberger [Tue, 14 Feb 2023 16:13:18 +0000 (17:13 +0100)]
s390/zcrypt: rework length information for dqap

The inline ap_dqap function does not return the number of
bytes actually written into the message buffer. The calling
code inspects the AP message header to figure out what kind
of AP message has been received and pulls the length
information from this header. This processing may not work
correctly in cases where only a fragment of the reply is
received.

With this patch the ap_dqap inline function now returns
the number of actually written bytes in the *length parameter.
So the calling function has a chance to compare the number of
received bytes against what the AP message header length
field states. This is especially useful in cases where a
message could only get partially received.

The low level reply processing functions needed some rework
to be able to catch this new length information and compare
it the right way. The rework also deals with some situations
where until now the reply length was not correctly calculated
and/or set.

All this has been heavily tested as the modifications on
the reply length information may affect crypto load.

Signed-off-by: Harald Freudenberger <freude@linux.ibm.com>
Reviewed-by: Holger Dengler <dengler@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
18 months agos390/zcrypt: make psmid unsigned long instead of long long
Harald Freudenberger [Sun, 29 Jan 2023 18:45:25 +0000 (19:45 +0100)]
s390/zcrypt: make psmid unsigned long instead of long long

Since s390 kernel build does not support 32 bit build any
more there is no difference between long and long long.
So this patch reworks all occurrences of psmid (a 64 bit
value) to use unsigned long now.

Signed-off-by: Harald Freudenberger <freude@linux.ibm.com>
Acked-by: Heiko Carstens <hca@linux.ibm.com>
Reviewed-by: Holger Dengler <dengler@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
18 months agos390: enable DEBUG_FORCE_FUNCTION_ALIGN_64B
Heiko Carstens [Mon, 13 Mar 2023 12:50:39 +0000 (13:50 +0100)]
s390: enable DEBUG_FORCE_FUNCTION_ALIGN_64B

Allow to enforce 64 byte function alignment like it is possible for a
couple of other architectures. This may or may not be helpful for
debugging performance problems, as described with the Kconfig option.

Since the kernel works also with 64 byte function alignment there is
no reason for not allowing to enforce this function alignment.

Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
18 months agos390/vdso: use __ALIGN instead of open coded .align
Heiko Carstens [Mon, 13 Mar 2023 12:50:38 +0000 (13:50 +0100)]
s390/vdso: use __ALIGN instead of open coded .align

Use __ALIGN instead of open coded .align statement to make sure that
vdso code follows global kernel function alignment rules.

Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
18 months agos390/expoline: use __ALIGN instead of open coded .align
Heiko Carstens [Mon, 13 Mar 2023 12:50:37 +0000 (13:50 +0100)]
s390/expoline: use __ALIGN instead of open coded .align

Use __ALIGN instead of open coded .align statement to make sure that
external expoline thunks follow global function alignment rules.

Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
18 months agos390/ftrace: move hotpatch trampolines to mcount.S
Heiko Carstens [Mon, 13 Mar 2023 12:50:36 +0000 (13:50 +0100)]
s390/ftrace: move hotpatch trampolines to mcount.S

Move the ftrace hotpatch trampolines to mcount.S. This allows to make
use of the standard SYM_CODE macros which again makes sure that the
hotpatch trampolines follow the function alignment rules of the rest
of the kernel.

Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Acked-by: Ilya Leoshkevich <iii@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
18 months agos390: make use of CONFIG_FUNCTION_ALIGNMENT
Heiko Carstens [Mon, 13 Mar 2023 12:50:35 +0000 (13:50 +0100)]
s390: make use of CONFIG_FUNCTION_ALIGNMENT

Make use of CONFIG_FUNCTION_ALIGNMENT which was introduced with commit
d49a0626216b ("arch: Introduce CONFIG_FUNCTION_ALIGNMENT").

Select FUNCTION_ALIGNMENT_8B for gcc in order to reflect gcc's default
function alignment. For all other compilers, which is only clang, select
a function alignment of 16 bytes which reflects the default function
alignment for clang.

Also change the __ALIGN define to follow whatever the value of
CONFIG_FUNCTION_ALIGNMENT is. This makes sure that the alignment of C and
assembler functions is the same.

In result everything still uses the default function alignment for both
compilers. However in addition this is now also true for all assembly
functions, so that all functions have a consistent alignment.

Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
18 months agoMerge branch 'decompressor-memory-tracking' into features
Heiko Carstens [Mon, 20 Mar 2023 10:04:10 +0000 (11:04 +0100)]
Merge branch 'decompressor-memory-tracking' into features

Vasily Gorbik says:

===================
Combine and generalize all methods for finding unused memory in
decompressor, while decreasing complexity, add memory holes support,
while improving error handling (especially in low-memory conditions)
and debug-ability.
===================

Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
18 months agos390/kasan: move shadow mapping to decompressor
Vasily Gorbik [Thu, 9 Feb 2023 21:05:11 +0000 (22:05 +0100)]
s390/kasan: move shadow mapping to decompressor

Since regular paging structs are initialized in decompressor already
move KASAN shadow mapping to decompressor as well. This helps to avoid
allocating KASAN required memory in 1 large chunk, de-duplicate paging
structs creation code and start the uncompressed kernel with KASAN
instrumentation right away. This also allows to avoid all pitfalls
accidentally calling KASAN instrumented code during KASAN initialization.

Acked-by: Heiko Carstens <hca@linux.ibm.com>
Reviewed-by: Alexander Gordeev <agordeev@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
18 months agos390/mm,pageattr: allow KASAN shadow memory
Vasily Gorbik [Tue, 14 Feb 2023 08:39:24 +0000 (09:39 +0100)]
s390/mm,pageattr: allow KASAN shadow memory

Allow changing page table attributes for KASAN shadow memory ranges.

Acked-by: Heiko Carstens <hca@linux.ibm.com>
Reviewed-by: Alexander Gordeev <agordeev@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
18 months agos390/boot: rework decompressor reserved tracking
Vasily Gorbik [Thu, 2 Feb 2023 12:59:36 +0000 (13:59 +0100)]
s390/boot: rework decompressor reserved tracking

Currently several approaches for finding unused memory in decompressor
are utilized. While "safe_addr" grows towards higher addresses, vmem
code allocates paging structures top down. The former requires careful
ordering. In addition to that ipl report handling code verifies potential
intersections with secure boot certificates on its own. Neither of two
approaches are memory holes aware and consistent with each other in low
memory conditions.

To solve that, existing approaches are generalized and combined
together, as well as online memory ranges are now taken into
consideration.

physmem_info has been extended to contain reserved memory ranges. New
set of functions allow to handle reserves and find unused memory.
All reserves and memory allocations are "typed". In case of out of
memory condition decompressor fails with detailed info on current
reserved ranges and usable online memory.

Linux version 6.2.0 ...
Kernel command line: ... mem=100M
Our of memory allocating 100000 bytes 100000 aligned in range 0:5800000
Reserved memory ranges:
0000000000000000 0000000003e33000 DECOMPRESSOR
0000000003f00000 00000000057648a3 INITRD
00000000063e0000 00000000063e8000 VMEM
00000000063eb000 00000000063f4000 VMEM
00000000063f7800 0000000006400000 VMEM
0000000005800000 0000000006300000 KASAN
Usable online memory ranges (info source: sclp read info [3]):
0000000000000000 0000000006400000
Usable online memory total: 6400000 Reserved: 61b10a3 Free: 24ef5d
Call Trace:
(sp:000000000002bd58 [<0000000000012a70>] physmem_alloc_top_down+0x60/0x14c)
 sp:000000000002bdc8 [<0000000000013756>] _pa+0x56/0x6a
 sp:000000000002bdf0 [<0000000000013bcc>] pgtable_populate+0x45c/0x65e
 sp:000000000002be90 [<00000000000140aa>] setup_vmem+0x2da/0x424
 sp:000000000002bec8 [<0000000000011c20>] startup_kernel+0x428/0x8b4
 sp:000000000002bf60 [<00000000000100f4>] startup_normal+0xd4/0xd4

physmem_alloc_range allows to find free memory in specified range. It
should be used for one time allocations only like finding position for
amode31 and vmlinux.
physmem_alloc_top_down can be used just like physmem_alloc_range, but
it also allows multiple allocations per type and tries to merge sequential
allocations together. Which is useful for paging structures allocations.
If sequential allocations cannot be merged together they are "chained",
allowing easy per type reserved ranges enumeration and migration to
memblock later. Extra "struct reserved_range" allocated for chaining are
not tracked or reserved but rely on the fact that both
physmem_alloc_range and physmem_alloc_top_down search for free memory
only below current top down allocator position. All reserved ranges
should be transferred to memblock before memblock allocations are
enabled.

The startup code has been reordered to delay any memory allocations until
online memory ranges are detected and occupied memory ranges are marked as
reserved to be excluded from follow-up allocations.
Ipl report certificates are a special case, ipl report certificates list
is checked together with other memory reserves until certificates are
saved elsewhere.
KASAN required memory for shadow memory allocation and mapping is reserved
as 1 large chunk which is later passed to KASAN early initialization code.

Acked-by: Heiko Carstens <hca@linux.ibm.com>
Reviewed-by: Alexander Gordeev <agordeev@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
18 months agos390/boot: rename mem_detect to physmem_info
Vasily Gorbik [Wed, 8 Feb 2023 17:11:25 +0000 (18:11 +0100)]
s390/boot: rename mem_detect to physmem_info

In preparation to extending mem_detect with additional information like
reserved ranges rename it to more generic physmem_info. This new naming
also help to avoid confusion by using more exact terms like "physmem
online ranges", etc.

Acked-by: Heiko Carstens <hca@linux.ibm.com>
Reviewed-by: Alexander Gordeev <agordeev@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
18 months agos390/boot: remove non-functioning image bootable check
Vasily Gorbik [Wed, 15 Feb 2023 13:27:45 +0000 (14:27 +0100)]
s390/boot: remove non-functioning image bootable check

check_image_bootable() has been introduced with commit 627c9b62058e
("s390/boot: block uncompressed vmlinux booting attempts") to make sure
that users don't try to boot uncompressed vmlinux ELF image in qemu. It
used to be possible quite some time ago. That commit prevented confusion
with uncompressed vmlinux image starting to boot and even printing
kernel messages until it crashed. Users might have tried to report the
problem without realizing they are doing something which was not intended.
Since commit f1d3c5323772 ("s390/boot: move sclp early buffer from fixed
address in asm to C") check_image_bootable() doesn't function properly
anymore, as well as booting uncompressed vmlinux image in qemu doesn't
really produce any output and crashes. Moving forward it doesn't make
sense to fix check_image_bootable() anymore, so simply remove it.

Acked-by: Alexander Gordeev <agordeev@linux.ibm.com>
Acked-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
18 months agos390/dumpstack: resolve userspace last_break
Ilya Leoshkevich [Fri, 10 Mar 2023 02:36:48 +0000 (03:36 +0100)]
s390/dumpstack: resolve userspace last_break

report_user_fault() currently does not show which library last_break
points to. Call print_vma_addr() to find out; the output now looks
like this:

    Last Breaking-Event-Address:
     [<000003ffaa2a56e4>] libc.so.6[3ffaa180000+251000]

For kernel it's unchanged:

    Last Breaking-Event-Address:
     [<000000000030fd06>] trace_hardirqs_on+0x56/0xc8

Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Acked-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
18 months agos390: simplify dynamic sysctl registration for appldata_register_ops
Luis Chamberlain [Fri, 10 Mar 2023 23:45:25 +0000 (15:45 -0800)]
s390: simplify dynamic sysctl registration for appldata_register_ops

The routine appldata_register_ops() allocates a sysctl table
with 4 entries. The firsts one,   ops->ctl_table[0] is the parent directory
with an empty entry following it, ops->ctl_table[1]. The next entry is
for the ops->name and that is     ops->ctl_table[2]. It needs an empty
entry following that, and that is ops->ctl_table[3]. And so hence the
kcalloc(4, sizeof(struct ctl_table), GFP_KERNEL).

We can simplify this considerably since sysctl_register("foo", table)
can create the parent directory for us if it does not exist. So we
can just remove the first two entries and move back the ops->name to
the first entry, and just use kcalloc(2, ...).

[gor@linux.ibm.com: appldata_generic_handler fixup ctl_table index 2->0]
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
Link: https://lore.kernel.org/r/20230310234525.3986352-7-mcgrof@kernel.org
Reviewed-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
18 months agos390: simplify one-level sysctl registration for page_table_sysctl
Luis Chamberlain [Fri, 10 Mar 2023 23:45:24 +0000 (15:45 -0800)]
s390: simplify one-level sysctl registration for page_table_sysctl

There is no need to declare an extra tables to just create directory,
this can be easily be done with a prefix path with register_sysctl().

Simplify this registration.

Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
Link: https://lore.kernel.org/r/20230310234525.3986352-6-mcgrof@kernel.org
Reviewed-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
18 months agos390: simplify one level sysctl registration for cmm_table
Luis Chamberlain [Fri, 10 Mar 2023 23:45:23 +0000 (15:45 -0800)]
s390: simplify one level sysctl registration for cmm_table

There is no need to declare an extra tables to just create directory,
this can be easily be done with a prefix path with register_sysctl().

Simplify this registration.

Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
Link: https://lore.kernel.org/r/20230310234525.3986352-5-mcgrof@kernel.org
Reviewed-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
18 months agos390: simplify one-level sysctl registration for appldata_table
Luis Chamberlain [Fri, 10 Mar 2023 23:45:22 +0000 (15:45 -0800)]
s390: simplify one-level sysctl registration for appldata_table

There is no need to declare an extra tables to just create directory,
this can be easily be done with a prefix path with register_sysctl().

Simplify this registration.

Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
Link: https://lore.kernel.org/r/20230310234525.3986352-4-mcgrof@kernel.org
Reviewed-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
18 months agos390: simplify one-level syctl registration for s390dbf_table
Luis Chamberlain [Fri, 10 Mar 2023 23:45:21 +0000 (15:45 -0800)]
s390: simplify one-level syctl registration for s390dbf_table

There is no need to declare an extra tables to just create directory,
this can be easily be done with a prefix path with register_sysctl().

Simplify this registration.

Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
Link: https://lore.kernel.org/r/20230310234525.3986352-3-mcgrof@kernel.org
Reviewed-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
18 months agos390: simplify one-level sysctl registration for topology_ctl_table
Luis Chamberlain [Fri, 10 Mar 2023 23:45:20 +0000 (15:45 -0800)]
s390: simplify one-level sysctl registration for topology_ctl_table

There is no need to declare an extra tables to just create directory,
this can be easily be done with a prefix path with register_sysctl().

Simplify this registration.

Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
Link: https://lore.kernel.org/r/20230310234525.3986352-2-mcgrof@kernel.org
Reviewed-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
18 months agos390/pci: clean up left over special treatment for function zero
Niklas Schnelle [Mon, 6 Mar 2023 15:10:14 +0000 (16:10 +0100)]
s390/pci: clean up left over special treatment for function zero

Prior to commit 960ac3626487 ("s390/pci: allow zPCI zbus without
a function zero") enabling and scanning a PCI function had to
potentially be postponed until the function with devfn zero on that bus
was plugged. While the commit removed the waiting itself extra code to
scan all functions on the PCI bus once function zero appeared was
missed. Remove that code and the outdated comments about waiting for
function zero.

Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Reviewed-by: Matthew Rosato <mjrosato@linux.ibm.com>
Link: https://lore.kernel.org/r/20230306151014.60913-5-schnelle@linux.ibm.com
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
18 months agos390/pci: remove redundant pci_bus_add_devices() on new bus
Niklas Schnelle [Mon, 6 Mar 2023 15:10:13 +0000 (16:10 +0100)]
s390/pci: remove redundant pci_bus_add_devices() on new bus

The pci_bus_add_devices() call in zpci_bus_create_pci_bus() is without
function since at this point no device could have been added to the
freshly created PCI bus.

Suggested-by: Bjorn Helgaas <helgaas@kernel.org>
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Reviewed-by: Matthew Rosato <mjrosato@linux.ibm.com>
Link: https://lore.kernel.org/r/20230306151014.60913-4-schnelle@linux.ibm.com
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
18 months agos390/pci: only add specific device in zpci_bus_scan_device()
Niklas Schnelle [Mon, 6 Mar 2023 15:10:12 +0000 (16:10 +0100)]
s390/pci: only add specific device in zpci_bus_scan_device()

As the name suggests zpci_bus_scan_device() is used to scan a specific
device and thus pci_bus_add_device() for that device is sufficient.
Furthermore move this call inside the rescan/remove locking.

Suggested-by: Bjorn Helgaas <helgaas@kernel.org>
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Reviewed-by: Matthew Rosato <mjrosato@linux.ibm.com>
Link: https://lore.kernel.org/r/20230306151014.60913-3-schnelle@linux.ibm.com
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
18 months agos390/setup: always inline gen_lpswe()
Heiko Carstens [Thu, 9 Mar 2023 19:12:28 +0000 (20:12 +0100)]
s390/setup: always inline gen_lpswe()

gen_lpswe() contains a BUILD_BUG_ON() statement which depends on a function
parameter. If the compiler decides to generate a not inlined function this
will lead to a build error, even if all call sites pass a valid parameter.

To avoid this always inline gen_lpswe().

Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
18 months agos390: trim ancient junk from copy_thread()
Al Viro [Mon, 6 Mar 2023 00:55:33 +0000 (00:55 +0000)]
s390: trim ancient junk from copy_thread()

Setting and ->psw.addr in childregs of kernel thread is a rudiment of
the old kernel_thread()/kernel_execve() implementation.  Mainline hadn't
been using them since 2012.

And clarify the assignments to frame->sf.gprs - the array stores
grp6..gpr15 values to be set by __switch_to(), so frame->sf.gprs[5]
actually affects grp11, etc.

Better spell that as frame->sf.gprs[11 - 6]...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Link: https://lore.kernel.org/r/ZAU6BYFisE8evmYf@ZenIV
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
18 months agos390/zcrypt: remove unnecessary (void *) conversions
Yu Zhe [Fri, 3 Mar 2023 05:21:55 +0000 (13:21 +0800)]
s390/zcrypt: remove unnecessary (void *) conversions

Pointer variables of void * type do not require type cast.

Signed-off-by: Yu Zhe <yuzhe@nfschina.com>
Reviewed-by: Muhammad Usama Anjum <usama.anjum@collabora.com>
Link: https://lore.kernel.org/r/20230303052155.21072-1-yuzhe@nfschina.com
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
18 months agos390/bp: remove __bpon()
Heiko Carstens [Tue, 28 Feb 2023 12:57:50 +0000 (13:57 +0100)]
s390/bp: remove __bpon()

There is no point in changing branch prediction state of a cpu shortly
before it enters stop state. Therefore remove __bpon().

Acked-by: Alexander Gordeev <agordeev@linux.ibm.com>
Reviewed-by: Sven Schnelle <svens@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
18 months agos390/bp: remove s390_isolate_bp_guest()
Heiko Carstens [Tue, 28 Feb 2023 12:57:49 +0000 (13:57 +0100)]
s390/bp: remove s390_isolate_bp_guest()

s390_isolate_bp_guest() is unused. Remove it.

Reviewed-by: Alexander Gordeev <agordeev@linux.ibm.com>
Reviewed-by: Sven Schnelle <svens@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
18 months agos390/bp: remove TIF_ISOLATE_BP
Heiko Carstens [Tue, 28 Feb 2023 12:57:48 +0000 (13:57 +0100)]
s390/bp: remove TIF_ISOLATE_BP

TIF_ISOLATE_BP is unused since it was introduced with commit 6b73044b2b00
("s390: run user space and KVM guests with modified branch prediction").
Given that there is no use case remove it again.

Acked-by: Alexander Gordeev <agordeev@linux.ibm.com>
Reviewed-by: Sven Schnelle <svens@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
18 months agos390/bp: add missing BPENTER to program check handler
Heiko Carstens [Tue, 28 Feb 2023 12:57:47 +0000 (13:57 +0100)]
s390/bp: add missing BPENTER to program check handler

When leaving interpretive execution because of a program check BPENTER
should be called like it is done on interrupt exit as well.

Acked-by: Alexander Gordeev <agordeev@linux.ibm.com>
Reviewed-by: Sven Schnelle <svens@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
18 months agos390: update defconfigs
Heiko Carstens [Fri, 10 Mar 2023 13:19:55 +0000 (14:19 +0100)]
s390: update defconfigs

Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
18 months agos390/ipl: add missing intersection check to ipl_report handling
Sven Schnelle [Tue, 7 Mar 2023 13:35:23 +0000 (14:35 +0100)]
s390/ipl: add missing intersection check to ipl_report handling

The code which handles the ipl report is searching for a free location
in memory where it could copy the component and certificate entries to.
It checks for intersection between the sections required for the kernel
and the component/certificate data area, but fails to check whether
the data structures linking these data areas together intersect.

This might cause the iplreport copy code to overwrite the iplreport
itself. Fix this by adding two addtional intersection checks.

Cc: <stable@vger.kernel.org>
Fixes: 9641b8cc733f ("s390/ipl: read IPL report at early boot")
Signed-off-by: Sven Schnelle <svens@linux.ibm.com>
Reviewed-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
18 months agoPCI: s390: Fix use-after-free of PCI resources with per-function hotplug
Niklas Schnelle [Mon, 6 Mar 2023 15:10:11 +0000 (16:10 +0100)]
PCI: s390: Fix use-after-free of PCI resources with per-function hotplug

On s390 PCI functions may be hotplugged individually even when they
belong to a multi-function device. In particular on an SR-IOV device VFs
may be removed and later re-added.

In commit a50297cf8235 ("s390/pci: separate zbus creation from
scanning") it was missed however that struct pci_bus and struct
zpci_bus's resource list retained a reference to the PCI functions MMIO
resources even though those resources are released and freed on
hot-unplug. These stale resources may subsequently be claimed when the
PCI function re-appears resulting in use-after-free.

One idea of fixing this use-after-free in s390 specific code that was
investigated was to simply keep resources around from the moment a PCI
function first appeared until the whole virtual PCI bus created for
a multi-function device disappears. The problem with this however is
that due to the requirement of artificial MMIO addreesses (address
cookies) extra logic is then needed to keep the address cookies
compatible on re-plug. At the same time the MMIO resources semantically
belong to the PCI function so tying their lifecycle to the function
seems more logical.

Instead a simpler approach is to remove the resources of an individually
hot-unplugged PCI function from the PCI bus's resource list while
keeping the resources of other PCI functions on the PCI bus untouched.

This is done by introducing pci_bus_remove_resource() to remove an
individual resource. Similarly the resource also needs to be removed
from the struct zpci_bus's resource list. It turns out however, that
there is really no need to add the MMIO resources to the struct
zpci_bus's resource list at all and instead we can simply use the
zpci_bar_struct's resource pointer directly.

Fixes: a50297cf8235 ("s390/pci: separate zbus creation from scanning")
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Reviewed-by: Matthew Rosato <mjrosato@linux.ibm.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Link: https://lore.kernel.org/r/20230306151014.60913-2-schnelle@linux.ibm.com
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
18 months agoLinux 6.3-rc2 v6.3-rc2
Linus Torvalds [Sun, 12 Mar 2023 23:36:44 +0000 (16:36 -0700)]
Linux 6.3-rc2

18 months agowifi: cfg80211: Partial revert "wifi: cfg80211: Fix use after free for wext"
Hector Martin [Sat, 11 Mar 2023 14:19:14 +0000 (23:19 +0900)]
wifi: cfg80211: Partial revert "wifi: cfg80211: Fix use after free for wext"

This reverts part of commit 015b8cc5e7c4 ("wifi: cfg80211: Fix use after
free for wext")

This commit broke WPA offload by unconditionally clearing the crypto
modes for non-WEP connections. Drop that part of the patch.

Signed-off-by: Hector Martin <marcan@marcan.st>
Reported-by: Ilya <me@0upti.me>
Reported-and-tested-by: Janne Grunau <j@jannau.net>
Reviewed-by: Eric Curtin <ecurtin@redhat.com>
Fixes: 015b8cc5e7c4 ("wifi: cfg80211: Fix use after free for wext")
Cc: stable@kernel.org
Link: https://lore.kernel.org/linux-wireless/ZAx0TWRBlGfv7pNl@kroah.com/T/#m11e6e0915ab8fa19ce8bc9695ab288c0fe018edf
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
18 months agoMerge tag 'tpm-v6.3-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/jarkko...
Linus Torvalds [Sun, 12 Mar 2023 23:15:36 +0000 (16:15 -0700)]
Merge tag 'tpm-v6.3-rc3' of git://git./linux/kernel/git/jarkko/linux-tpmdd

Pull tpm fixes from Jarkko Sakkinen:
 "Two additional bug fixes for v6.3"

* tag 'tpm-v6.3-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/jarkko/linux-tpmdd:
  tpm: disable hwrng for fTPM on some AMD designs
  tpm/eventlog: Don't abort tpm_read_log on faulty ACPI address

18 months agotpm: disable hwrng for fTPM on some AMD designs
Mario Limonciello [Tue, 28 Feb 2023 02:44:39 +0000 (20:44 -0600)]
tpm: disable hwrng for fTPM on some AMD designs

AMD has issued an advisory indicating that having fTPM enabled in
BIOS can cause "stuttering" in the OS.  This issue has been fixed
in newer versions of the fTPM firmware, but it's up to system
designers to decide whether to distribute it.

This issue has existed for a while, but is more prevalent starting
with kernel 6.1 because commit b006c439d58db ("hwrng: core - start
hwrng kthread also for untrusted sources") started to use the fTPM
for hwrng by default. However, all uses of /dev/hwrng result in
unacceptable stuttering.

So, simply disable registration of the defective hwrng when detecting
these faulty fTPM versions.  As this is caused by faulty firmware, it
is plausible that such a problem could also be reproduced by other TPM
interactions, but this hasn't been shown by any user's testing or reports.

It is hypothesized to be triggered more frequently by the use of the RNG
because userspace software will fetch random numbers regularly.

Intentionally continue to register other TPM functionality so that users
that rely upon PCR measurements or any storage of data will still have
access to it.  If it's found later that another TPM functionality is
exacerbating this problem a module parameter it can be turned off entirely
and a module parameter can be introduced to allow users who rely upon
fTPM functionality to turn it on even though this problem is present.

Link: https://www.amd.com/en/support/kb/faq/pa-410
Link: https://bugzilla.kernel.org/show_bug.cgi?id=216989
Link: https://lore.kernel.org/all/20230209153120.261904-1-Jason@zx2c4.com/
Fixes: b006c439d58d ("hwrng: core - start hwrng kthread also for untrusted sources")
Cc: stable@vger.kernel.org
Cc: Jarkko Sakkinen <jarkko@kernel.org>
Cc: Thorsten Leemhuis <regressions@leemhuis.info>
Cc: James Bottomley <James.Bottomley@hansenpartnership.com>
Tested-by: reach622@mailcuk.com
Tested-by: Bell <1138267643@qq.com>
Co-developed-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Reviewed-by: Jarkko Sakkinen <jarkko@kernel.org>
Signed-off-by: Jarkko Sakkinen <jarkko@kernel.org>
18 months agotpm/eventlog: Don't abort tpm_read_log on faulty ACPI address
Morten Linderud [Wed, 15 Feb 2023 09:25:52 +0000 (10:25 +0100)]
tpm/eventlog: Don't abort tpm_read_log on faulty ACPI address

tpm_read_log_acpi() should return -ENODEV when no eventlog from the ACPI
table is found. If the firmware vendor includes an invalid log address
we are unable to map from the ACPI memory and tpm_read_log() returns -EIO
which would abort discovery of the eventlog.

Change the return value from -EIO to -ENODEV when acpi_os_map_iomem()
fails to map the event log.

The following hardware was used to test this issue:
    Framework Laptop (Pre-production)
    BIOS: INSYDE Corp, Revision: 3.2
    TPM Device: NTC, Firmware Revision: 7.2

Dump of the faulty ACPI TPM2 table:
    [000h 0000   4]                    Signature : "TPM2"    [Trusted Platform Module hardware interface Table]
    [004h 0004   4]                 Table Length : 0000004C
    [008h 0008   1]                     Revision : 04
    [009h 0009   1]                     Checksum : 2B
    [00Ah 0010   6]                       Oem ID : "INSYDE"
    [010h 0016   8]                 Oem Table ID : "TGL-ULT"
    [018h 0024   4]                 Oem Revision : 00000002
    [01Ch 0028   4]              Asl Compiler ID : "ACPI"
    [020h 0032   4]        Asl Compiler Revision : 00040000

    [024h 0036   2]               Platform Class : 0000
    [026h 0038   2]                     Reserved : 0000
    [028h 0040   8]              Control Address : 0000000000000000
    [030h 0048   4]                 Start Method : 06 [Memory Mapped I/O]

    [034h 0052  12]            Method Parameters : 00 00 00 00 00 00 00 00 00 00 00 00
    [040h 0064   4]           Minimum Log Length : 00010000
    [044h 0068   8]                  Log Address : 000000004053D000

Fixes: 0cf577a03f21 ("tpm: Fix handling of missing event log")
Tested-by: Erkki Eilonen <erkki@bearmetal.eu>
Signed-off-by: Morten Linderud <morten@linderud.pw>
Reviewed-by: Jarkko Sakkinen <jarkko@kernel.org>
Signed-off-by: Jarkko Sakkinen <jarkko@kernel.org>