Age | Commit message (Collapse) | Author | Files | Lines |
|
Use union tod_clock and get rid of the kvm specific struct
kvm_s390_tod_clock_ext which apparently was introduced for the same
purpose.
Signed-off-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
Signed-off-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
Convert tod_clock_base to union tod_clock. This simplifies quite a bit
of code and also fixes a bug in read_persistent_clock64();
void read_persistent_clock64(struct timespec64 *ts)
{
__u64 delta;
delta = initial_leap_seconds + TOD_UNIX_EPOCH;
get_tod_clock_ext(clk);
*(__u64 *) &clk[1] -= delta;
if (*(__u64 *) &clk[1] > delta)
clk[0]--;
ext_to_timespec64(clk, ts);
}
Assume &clk[1] == 3 and delta == 2; then after the substraction the if
condition becomes true and the epoch part of the clock is decremented
by one because of an assumed overflow, even though there is none.
Fix this by using 128 bit arithmetics and let the compiler do the
right thing:
void read_persistent_clock64(struct timespec64 *ts)
{
union tod_clock clk;
u64 delta;
delta = initial_leap_seconds + TOD_UNIX_EPOCH;
store_tod_clock_ext(&clk);
clk.eitod -= delta;
ext_to_timespec64(&clk, ts);
}
Signed-off-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
Introduce new store_tod_clock_ext() function, which is the same like
store_tod_clock_ext_cc() except that it doesn't return a condition
code.
Signed-off-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
Rename store_tod_clock_ext() to store_tod_clock_ext_cc() to reflect
that it returns a condition code and also use union tod_clock as
parameter.
Signed-off-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
Introduce union tod_clock which is supposed to be used to decode and
access various fields of the result of STORE CLOCK EXTENDED.
Signed-off-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
s390 and alpha are the only 64 bit architectures with a 32-bit ino_t.
Since this is quite unusual this causes bugs from time to time.
See e.g. commit ebce3eb2f7ef ("ceph: fix inode number handling on
arches with 32-bit ino_t") for an example.
This (obviously) also prevents s390 and alpha to use 64-bit ino_t for
tmpfs. See commit b85a7a8bb573 ("tmpfs: disallow CONFIG_TMPFS_INODE64
on s390").
Therefore switch both s390 and alpha to 64-bit ino_t. This should only
have an effect on the ustat system call. To prevent ABI breakage
define struct ustat compatible to the old layout and change
sys_ustat() accordingly.
Acked-by: Linus Torvalds <[email protected]>
Signed-off-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
The current code uses the address in %r11 to figure out whether
it was called from the machine check handler or from a normal
interrupt handler. Instead of doing this implicit logic (which
is mostly a leftover from the old critical cleanup approach)
just add a second label and use that.
Signed-off-by: Sven Schnelle <[email protected]>
Reviewed-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
Instead of thrashing r11 which is normally our pointer to struct
pt_regs on the stack, use r13 as temporary register in the BR_EX
macro. r13 is already used in cleanup_sie, so no need to thrash
another register.
Signed-off-by: Sven Schnelle <[email protected]>
Reviewed-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
If a machine check is coming in during sie, the PU saves the
control registers to the machine check save area. Afterwards
mcck_int_handler is called, which loads __LC_KERNEL_ASCE into
%cr1. Later the code restores %cr1 from the machine check area,
but that is wrong when SIE was interrupted because the machine
check area still contains the gmap asce. Instead it should return
with either __KERNEL_ASCE in %cr1 when interrupted in SIE or
the previous %cr1 content saved in the machine check save area.
Fixes: 87d598634521 ("s390/mm: remove set_fs / rework address space handling")
Signed-off-by: Sven Schnelle <[email protected]>
Cc: <[email protected]> # v5.8+
Reviewed-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
The previous code used the normal kernel stack for machine checks.
This is problematic when a machine check interrupts a system call
or interrupt handler right at the beginning where registers are set up.
Assume system_call is interrupted at the first instruction and a machine
check is triggered. The machine check handler is called, checks the PSW
to see whether it is coming from user space, notices that it is already
in kernel mode but %r15 still contains the user space stack. This would
lead to a kernel crash.
There are basically two ways of fixing that: Either using the 'critical
cleanup' approach which compares the address in the PSW to see whether
it is already at a point where the stack has been set up, or use an extra
stack for the machine check handler.
For simplicity, we will go with the second approach and allocate an extra
stack. This adds some memory overhead for large systems, but usually large
system have plenty of memory so this isn't really a concern. But it keeps
the mchk stack setup simple and less error prone.
Fixes: 0b0ed657fe00 ("s390: remove critical section cleanup from entry.S")
Signed-off-by: Sven Schnelle <[email protected]>
Cc: <[email protected]> # v5.8+
Reviewed-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
The code does:
S390_lowcore.async_stack = new + STACK_INIT_OFFSET;
But the compiler is free to first assign one value and
add the other value later. If a IRQ would be coming in
between these two operations, it would run with an invalid
stack. Prevent this by using WRITE_ONCE.
Signed-off-by: Sven Schnelle <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
This is a preparation patch for two later bugfixes. In the past both
int_handler and machine check handler used SWITCH_KERNEL to switch to
the kernel stack. However, SWITCH_KERNEL doesn't work properly in machine
check context. So instead of adding more complexity to this macro, just
remove it.
Signed-off-by: Sven Schnelle <[email protected]>
Cc: <[email protected]> # v5.8+
Reviewed-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
Use a cpu alternative to switch between stck and stckf instead of
making it compile time dependent. This will also make kernels compiled
for old machines, but running on newer machines, use stckf.
Signed-off-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
Add support for alternative inline assemblies with input and output
arguments. This is consistent to x86.
Signed-off-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
Use a cpu alternative to switch between stck and stckf instead of
making it compile time dependent. This will also make kernels compiled
for old machines, but running on newer machines, use stckf.
Reviewed-by: Christian Borntraeger <[email protected]>
Signed-off-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
Use STORE CLOCK EXTENDED instead of STORE CLOCK in early tod clock
setup. This is just to remove another usage of stck, trying to remove
all usages of STORE CLOCK. This doesn't fix anything.
Reviewed-by: Christian Borntraeger <[email protected]>
Signed-off-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
Use get_tod_clock_fast() instead of store_tod_clock(), since
store_tod_clock() can be very slow.
Reviewed-by: Christian Borntraeger <[email protected]>
Signed-off-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
The stck/stckf instruction used within the inline assembly within
do_account_vtime() changes the condition code. This is not reflected
with the clobber list, and therefore might result in incorrect code
generation.
It seems unlikely that the compiler could generate incorrect code
considering the surrounding C code, but it must still be fixed.
Cc: <[email protected]>
Reviewed-by: Christian Borntraeger <[email protected]>
Signed-off-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
This is the s390 variant of commit e6b28ec65b6d ("x86/vdso: On timens
page fault prefault also VVAR page").
Reviewed-by: Alexander Gordeev <[email protected]>
Signed-off-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
Implement generic vdso time namespace support which also enables time
namespaces for s390. This is quite similar to what arm64 has.
Reviewed-by: Alexander Gordeev <[email protected]>
Signed-off-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
Use the passed in vdso_data pointer instead of calculating it again.
This is also required as a prerequisite for vdso time namespaces: if a
process is part of a time namespace __arch_get_vdso_data() will return
a pointer to the time namespace data page instead of the vdso data
page, which is not what __arch_get_hw_counter() expects.
Reviewed-by: Alexander Gordeev <[email protected]>
Signed-off-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
For consistency with x86 and arm64 move the data page before code
pages. Similar to commit 601255ae3c98 ("arm64: vdso: move data page
before code pages").
Reviewed-by: Alexander Gordeev <[email protected]>
Signed-off-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
Add a separate "[vvar]" mapping for the vdso datapage, since it
doesn't need to be executable or COW-able.
This is actually the s390 implementation of commit 871549385278
("arm64: vdso: put vdso datapage in a separate vma")
Reviewed-by: Alexander Gordeev <[email protected]>
Signed-off-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
Implement vdso mapping similar to arm64 and powerpc.
Reviewed-by: Alexander Gordeev <[email protected]>
Signed-off-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
- remove unneeded includes
- move functions around
- remove obvious and/or incorrect comments
- shorten some if conditions
No functional change.
Signed-off-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
A few local variables exist only so the contents of a global variable
can be copied to them, and use that value only for reading.
Just remove them and rename some global variables. Also change
vdso64_[start|end] to be character arrays to be consistent with other
architectures, and get rid of the global variable vdso64_kbase.
Reviewed-by: Alexander Gordeev <[email protected]>
Signed-off-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
vdso_pages (aka vdso64_pages) is never 0, therefore remove the check.
Reviewed-by: Alexander Gordeev <[email protected]>
Signed-off-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
Handle allocation error gracefully and simply disable vdso instead of
leaving the system in an undefined state.
Reviewed-by: Alexander Gordeev <[email protected]>
Signed-off-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
The vdso is (and must) be page aligned and its size must also be
a multiple of PAGE_SIZE. Therefore no need to round upwards.
Reviewed-by: Alexander Gordeev <[email protected]>
Signed-off-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
Convert vdso_init() to arch_initcall like it is on all other architectures.
This requires to remove the vdso_getcpu_init() call from vdso_init()
since it must be called before smp is enabled.
vdso_getcpu_init() is now an early_initcall like on powerpc.
Reviewed-by: Alexander Gordeev <[email protected]>
Signed-off-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
The vdso data page actually contains an array. Fix that.
This doesn't fix a real bug, just reflects reality.
Reviewed-by: Alexander Gordeev <[email protected]>
Signed-off-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
Reviewed-by: Alexander Gordeev <[email protected]>
Signed-off-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
Since Fedora 33 the virtualization stack of Fedora requires a couple of
netfilter modules to function properly. Let's add these to defconfig and
debug_defconfig.
Signed-off-by: Halil Pasic <[email protected]>
Reported-by: Marc Hartmayer <[email protected]>
Tested-by: Bjoern Walk <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
...but set it to off by default. Use the kernel command line option
`kmemleak=on` to enable it.
Signed-off-by: Marc Hartmayer <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
The GitHub organisation name under which the s390-tools package is being
hosted has changed. Update the web link.
Signed-off-by: Jan Höppner <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
When a msg is retried because the lower ap layer returns -EAGAIN
there is a retry limit (currently 10). When this limit is reached
the last return code from the lower layer is returned, causing
the userspace to get -1 on the ioctl with errno EAGAIN.
This EAGAIN is misleading here. After 10 retry attempts the
userspace should receive a clear failure indication like EINVAL
or EIO or ENODEV. However, the reason why these retries all
fail is unclear. On an invalid message EINVAL would be returned
by the lower layer, and if devices go away or are not available
an ENODEV is seen. So this patch now reworks the retry loops
to return EIO to userspace when the retry limit is reached.
Fixes: 91ffc519c199 ("s390/zcrypt: introduce msg tracking in zcrypt functions")
Signed-off-by: Harald Freudenberger <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
Add missing forward declaration for task_struct.
The warning appears when the -Werror C compiler flag is being used.
Signed-off-by: Alexander Egorenkov <[email protected]>
Acked-by: Heiko Carstens <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
Disable CONFIG_TMPFS_INODE64 which is currently broken on s390x
because size of ino_t on s390x is 4 bytes.
This fixes the following error with kdump:
[ 9.415082] [608]: Remounting '/' read-only in with options 'size=238372k,nr_inodes=59593,inode64'.
[ 9.415093] rootfs: Cannot use inode64 with <64bit inums in kernel
[ 9.415093]
[ 9.415100] [608]: Failed to remount '/' read-only: Invalid argument
Fixes: 5c60ed283e1d ("s390: update defconfigs")
Signed-off-by: Alexander Egorenkov <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
Fix the following coccicheck warnings:
./arch/s390/include/asm/scsw.h:528:48-50: WARNING !A || A && B is
equivalent to !A || B.
Reported-by: Abaci Robot <[email protected]>
Signed-off-by: Jiapeng Zhong <[email protected]>
Reviewed-by: Vineeth Vijayan <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
Remove a superfluous semicolon after function definition.
Signed-off-by: Chengyang Fan <[email protected]>
Message-Id: <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
Currently zpci_create_device() is only called in clp_add_pci_device()
which allocates the memory for the struct zpci_dev being created. There
is little separation of concerns as only both functions together can
create a zpci_dev and the only CLP specific code in
clp_add_pci_device() is a call to clp_query_pci_fn().
Improve this by removing clp_add_pci_device() and refactor
zpci_create_device() such that it alone creates and initializes the
zpci_dev given the FID and Function Handle. For this we need to make
clp_query_pci_fn() non-static. While at it remove the function handle
parameter since we can just take that from the zpci_dev. Also move
adding to the zpci_list to after the zdev has been fully created which
eliminates a window where a partially initialized zdev can be found by
get_zdev_by_fid().
Acked-by: Pierre Morel <[email protected]>
Signed-off-by: Niklas Schnelle <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
We currently track the time of the most recent QDIO Adapter Interrupt.
This is a system-wide timestamp (as such interrupts are not bound to
one specific qdio device).
If interrupt processing stalls on one device but is functional for a
different device, the timestamp continues to be updated and is of no
help for problem diagnosis.
So for debugging purposes also track the time of the last Data IRQ on
a per-device level. Collect this data in the legacy non-AI path as well.
Signed-off-by: Julian Wiedmann <[email protected]>
Reviewed-by: Benjamin Block <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
tiqdio_add_device() adds the device to the tiq_list of eligible targets
for a data IRQ, which gets walked on each QDIO Adapter Interrupt to
inspect their DSCIs.
But currently the tiqdio_add_device() / tiqdio_remove_device() calls
are not symmetric - the device is removed within qdio_shutdown(),
but only added by qdio_activate().
So depending on the call sequence and encountered errors, we might
be trying to remove a list entry in qdio_shutdown() that was never even
added to the list. This required additional INIT_LIST_HEAD() calls to
ensure that the list entry was always in a consistent state.
All drivers now fence the IRQ delivery via qdio_start_irq() /
qdio_stop_irq(), so we can nicely integrate this tiq_list management
with the other steps needed for QDIO Adapter IRQ (de-)registration
(qdio_establish_thinint() / qdio_shutdown_thinint()).
As the naming suggests these get called during qdio_establish() and
qdio_shutdown(), with proper symmetry and roll-back after errors.
With this we longer need to worry about misplaced list removals, and
thus can clean up the list API abuse (INIT_LIST_HEAD() should not be
called on list entries).
Signed-off-by: Julian Wiedmann <[email protected]>
Reviewed-by: Benjamin Block <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
Convert the Output Queue tasklet code to take a tasklet_struct as
parameter. Then initialize the tasklet with tasklet_setup() to indicate
that we follow the new model.
Signed-off-by: Julian Wiedmann <[email protected]>
Reviewed-by: Benjamin Block <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
It's used in just one place, inline it.
Signed-off-by: Julian Wiedmann <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
Both qeth and zfcp have fully moved to the polling-driven flow for
Input Queues with commit 0a6e634535f1 ("s390/qdio: extend polling
support to multiple queues") and commit 0b524abc2dd1 ("scsi: zfcp: Lift
Input Queue tasklet from qdio").
So remove the tasklet code for Input Queues, streamline the IRQ handlers
and push the tasklet struct into struct qdio_output_q.
Signed-off-by: Julian Wiedmann <[email protected]>
Reviewed-by: Benjamin Block <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
A master key change on a CCA card may cause an immediately
following request to derive an protected key from a secure
key to fail with error condition 8/2290. The recommendation
from firmware is to retry with 1 second sleep.
So now the low level cca functions return -EAGAIN when this
error condition is seen and the paes retry function will
evaluate the return value. Seeing EAGAIN and running in
process context results in trying to sleep for 1 s now.
Signed-off-by: Harald Freudenberger <[email protected]>
Reviewed-by: Ingo Franzki <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
Checking zdev->zbus for NULL in __zpci_event_availability() is
superfluous as it can never be NULL at this point. While harmless this
check causes smatch warnings because we later access zdev->zbus with
only having checked zdev != NULL which is sufficient.
The reason zdev->zbus can never be NULL is since with zdev != NULL given
we know the zdev came from get_zdev_by_fid() and thus the zpci_list.
Now on first glance at zpci_create_device() one may assume that there is
a window where the zdev is in the list without a zdev, however this
window can't overlap with __zpci_event_availability() as
zpci_create_device() either runs on the same kthread as part of
availability events, or during the initial CLP List PCI at which point
the __zpci_event_availability() is not yet called as zPCI is not yet
initialized.
Reported-by: kernel test robot <[email protected]>
Reported-by: Dan Carpenter <[email protected]>
Signed-off-by: Niklas Schnelle <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|
|
This fixes the following warning:
CHECK linux/arch/s390/kernel/signal.c
linux/arch/s390/kernel/signal.c:465:6: warning: symbol 'arch_do_signal_or_restart' was not declared. Should it be static?
Signed-off-by: Sven Schnelle <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>
|