- 08 Jul, 2021 28 commits
-
-
Valentin Schneider authored
S390's init_idle_preempt_count(p, cpu) doesn't actually let us initialize the preempt_count of the requested CPU's idle task: it unconditionally writes to the current CPU's. This clearly conflicts with idle_threads_init(), which intends to initialize *all* the idle tasks, including their preempt_count (or their CPU's, if the arch uses a per-CPU preempt_count). Unfortunately, it seems the way s390 does things doesn't let us initialize every possible CPU's preempt_count early on, as the pages where this resides are only allocated when a CPU is brought up and are freed when it is brought down. Let the arch-specific code set a CPU's preempt_count when its lowcore is allocated, and turn init_idle_preempt_count() into an empty stub. Fixes: f1a0a376 ("sched/core: Initialize the idle task with preemption disabled") Reported-by:
Guenter Roeck <linux@roeck-us.net> Signed-off-by:
Valentin Schneider <valentin.schneider@arm.com> Tested-by:
Heiko Carstens <hca@linux.ibm.com> Link: https://lore.kernel.org/r/20210707163338.1623014-1-valentin.schneider@arm.comSigned-off-by:
Vasily Gorbik <gor@linux.ibm.com>
-
Vasily Gorbik authored
Both clang and gcc (for -march=z13 and later) align functions to 16 bytes at -O2 to benefit branch prediction. Make asm symbols alignment consistent with that. This also benefits potential ftrace code patching, which is currently able to patch 8 aligned bytes at once. With defconfig this currently increases .text size by 4104 bytes. Reviewed-by:
-
Heiko Carstens authored
Lower case matches the call_on_stack() macro and is easier to read. Reviewed-by:
Sven Schnelle <svens@linux.ibm.com> Signed-off-by:
-
Heiko Carstens authored
Make sure the to be called function takes no arguments (and returns void). Otherwise usage of CALL_ON_STACK_NORETURN() would generate broken code. Reviewed-by:
-
Heiko Carstens authored
Reviewed-by:
-
Heiko Carstens authored
Reviewed-by:
-
Heiko Carstens authored
Reviewed-by:
-
Heiko Carstens authored
Reviewed-by:
-
Heiko Carstens authored
Reviewed-by:
-
Heiko Carstens authored
Reviewed-by:
-
Heiko Carstens authored
Reviewed-by:
-
Heiko Carstens authored
The existing CALL_ON_STACK() macro allows for subtle bugs: - There is no type checking of the function that is being called. That is: missing or too many arguments do not cause any compile error or warning. The same is true if the return type of the called function changes. This can lead to quite random bugs. - Sign and zero extension of arguments is missing. Given that the s390 C ABI requires that the caller of a function performs proper sign and zero extension this can also lead to subtle bugs. - If arguments to the CALL_ON_STACK() macros contain functions calls register corruption can happen due to register asm constructs being used. Therefore introduce a new call_on_stack() macro which is supposed to fix all these problems. Reviewed-by:
-
Heiko Carstens authored
Make on_async_stack() a bit more readable, even though as usual it depends if one considers "!!!" readable or not. At least the new construct to check if the async stack is in use or not is a bit shorter and generates slightly better code. Reviewed-by:
-
Heiko Carstens authored
Move do_softirq_own_stack() to proper header file so it can be inlined; saving a few cycles. Reviewed-by:
-
Heiko Carstens authored
do_softirq_own_stack() is always called from task context and therefore it is not necessary to check if the async stack is currently used. Remove the check and directly switch to async stack. Reviewed-by:
-
Harald Freudenberger authored
This is the second part of the cleanup for the header file ap.h to remove the register asm statements. This patch deals with the inline ap_dqap() function where within the assembler code an odd register of an register pair is to be addressed. [hca@linux.ibm.com: this intentionally breaks compilation with any clang compilers prior to llvm-project commit 458eac257377 ("[SystemZ] Support the 'N' code for the odd register in inline-asm."). This is hopefully the last clang kernel compile breakage caused by incompatibilities between gcc and clang.] Signed-off-by:Harald Freudenberger <freude@linux.ibm.com> Signed-off-by:
-
Sven Schnelle authored
PIF_SYSCALL_RESTART is now only used to restart execve when loading PGSTE binaries. Rename the flag to reflect that, and avoid people thinking that this bit has anything to do with generic syscall restarting. Signed-off-by:
-
Sven Schnelle authored
On s390, execve might have to be restarted for PGSTE binaries like kvm. In the past this was done via the PIF_SYSCALL_RESTART bit. However, with the recent changes, syscalls are now restarted differently. Now that execve() is the only call that might get restarted via PIF_SYSCALL_RESTART, move the loop to do_syscall(). This also has the advantage that the restart is no longer visible to userspace. Signed-off-by:
-
Sven Schnelle authored
{rt_}sigreturn is now called from the vdso, so we no longer need the svc on the stack, and therefore no hack to support that mechanism on machines with non-executable stack. Signed-off-by: -
Sven Schnelle authored
with generic entry, there's a bug when it comes to restarting of signals. The failing sequence is: a) a signal is coming in, and no handler is registered, so the lower part of arch_do_signal_or_restart() in arch/s390/kernel/signal.c sets PIF_SYSCALL_RESTART. b) a second signal gets pending while the kernel is still in the exit loop, and for that one, a handler exists. c) The first part of arch_do_signal_or_restart() is called. That part calls handle_signal(), which sets up stack + registers for handling the signal. d) __do_syscall() in arch/s390/kernel/syscall.c checks for PIF_SYSCALL_RESTART right before leaving to userspace. If it is set, it restart's the syscall. However, the registers are already setup for handling a signal from c). The syscall is now restarted with the wrong arguments. Change the code to: - use vdso for syscall_restart() instead of PIF_SYSCALL_RESTART because we cannot rewind and go back to userspace on s390 because the system call number might be encoded in the svc instruction. - for all other syscalls we rewind the PSW and return to userspace. Cc: <stable@kernel.org> # v5.12+ d57778fe: s390/vdso: always enable vdso Cc: <stable@kernel.org> # v5.12+ 686341f2: s390/vdso64: add sigreturn,rt_sigreturn and restart_syscall Cc: <stable@kernel.org> # v5.12+ 43e1f76b: s390/vdso: rename VDSO64_LBASE to VDSO_LBASE Cc: <stable@kernel.org> # v5.12+ 779df224: s390/vdso: add minimal compat vdso Cc: <stable@kernel.org> # v5.12+ Signed-off-by:
-
Sven Schnelle authored
Add a small vdso for 31 bit compat application that provides trampolines for calls to sigreturn,rt_sigreturn,syscall_restart. This is requird for moving these syscalls away from the signal frame to the vdso. Note that this patch effectively disables CONFIG_COMPAT when using clang to compile the kernel. clang doesn't support 31 bit mode. We want to redirect sigreturn and restart_syscall to the vdso. However, the kernel cannot parse the ELF vdso file, so we need to generate header files which contain the offsets of the syscall instructions in the vdso page. Signed-off-by:
-
Sven Schnelle authored
Will be used by both vdso32 and vdso64, so change the name. Signed-off-by:
-
Sven Schnelle authored
Add minimalistic trampolines to vdso64 so we can return from signal without using the stack which requires pgm check handler hacks when NX is enabled. restart_syscall will be called from vdso to work around the architectural limitation that the syscall number might be encoded in the svc instruction, and therefore can not be changed. Signed-off-by:
-
Sven Schnelle authored
With the upcoming move of the svc sigreturn instruction from the signal frame to vdso we need to have vdso always enabled. Signed-off-by:
-
Heiko Carstens authored
Using register asm statements has been proven to be very error prone, especially when using code instrumentation where gcc may add function calls, which clobbers register contents in an unexpected way. Therefore get rid of register asm statements in ap code. There are also potential bugs, depending on inline decisions of the compiler. E.g. for: static inline struct ap_queue_status ap_tapq(ap_qid_t qid, unsigned long *info) { register unsigned long reg0 asm ("0") = qid; register struct ap_queue_status reg1 asm ("1"); register unsigned long reg2 asm ("2"); asm volatile(".long 0xb2af0000" /* PQAP(TAPQ) */ : "=d" (reg1), "=d" (reg2) : "d" (reg0) : "cc"); if (info) *info = reg2; return reg1; } In case of KCOV the "if (info)" line could cause a generated function call, which could clobber the contents of both reg2, and reg1. Similar can happen in case of KASAN for the "*info = reg2" line. Even though compilers will likely inline the function and optimize things away, this is not guaranteed. To get rid of this bug class, simply get rid of register asm constructs. Note: The inline function ap_dqap() will be handled in a separate patch because this one requires an addressing of the odd register of a register pair (which is done with %N[xxx] in the assembler code) and that's currently not supported by clang. Signed-off-by: -
Sven Schnelle authored
This is no longer true since we switched to generic entry. The code switches to the IRQ stack before calling do_IRQ, but switches back to the kernel stack before calling irq_exit(). Fixes: 56e62a73 ("s390: convert to generic entry") Signed-off-by:
-
Ilya Leoshkevich authored
tinyconfig fails to boot, because without CONFIG_BUG report_bug() always returns BUG_TRAP_TYPE_BUG, which causes mc 0,0 in test_monitor_call() to panic. Fix by skipping the test without CONFIG_BUG. Signed-off-by:
Ilya Leoshkevich <iii@linux.ibm.com> Reviewed-by:
-
Harald Freudenberger authored
Rework of the ap_dqap() inline function with the dqap inline assembler invocation and the caller code in ap_queue.c to be able to handle replies which exceed the receive buffer size. ap_dqap() now provides two additional parameters to handle together with the caller the case where a reply in the firmware queue entry exceeds the given message buffer size. It depends on the caller how to exactly handle this. The behavior implemented now by ap_sm_recv() in ap_queue.c is to simple purge this entry from the firmware queue and let the caller 'receive' a -EMSGSIZE for the request without delivering any reply data - not even a truncated reply message. However, the reworked ap_dqap() could now get invoked in a way that the message is received in multiple parts and the caller assembles the parts into one reply message. Signed-off-by:
Juergen Christ <jchrist@linux.ibm.com> Reviewed-by:
-
- 05 Jul, 2021 12 commits
-
-
Harald Freudenberger authored
This patch provides support for new dynamic AP bus message limit with the existing zcrypt device driver and AP bus core code. There is support for a new field 'ml' from TAPQ query. The field gives if != 0 the AP bus limit for this card in 4k chunk units. The actual message size limit per card is shown as a new read-only sysfs attribute. The sysfs attribute /sys/devices/ap/cardxx/max_msg_size shows the upper limit in bytes used by the AP bus and zcrypt device driver for requests and replies send to and received from this card. Currently up to CEX7 support only max 12kB msg size and thus the field shows 12288 meaning the upper limit of a valid msg for this card is 12kB. Please note that the usable payload is somewhat lower and depends on the msg type and thus the header struct which is to be prepended by the zcrypt dd. The dispatcher responsible for choosing the right card and queue is aware of the individual card AP bus message limit. So a request is only assigned to a queue of a card which is able to handle the size of the request (e.g. a 14kB request will never go to a max 12kB card). If no such card is found the ioctl will fail with ENODEV. The reply buffer held by the device driver is determined by the ml field of the TAPQ for this card. If a response from the card exceeds this limit however, the response is not truncated but the ioctl for this request will fail with errno EMSGSIZE to indicate that the device driver has dropped the response because it would overflow the buffer limit. If the request size does not indicate to the dispatcher that an adapter with extended limit is to be used, a random card will be chosen when no specific card is addressed (ANY addressing). This may result in an ioctl failure when the reply size needs an adapter with extended limit but the randomly chosen one is not capable of handling the broader reply size. The user space application needs to use dedicated addressing to forward such a request only to suitable cards to get requests like this processed properly. Signed-off-by:
Ingo Tuchscherer <ingo.tuchscherer@linux.ibm.com> Signed-off-by:
-
Heiko Carstens authored
"dummy" is not only used as output but also as input. Therefore use the correct "+" constraint modifier. Fixes: 8cf23c8e ("s390/lib,string: get rid of register asm") Reported-by:
-
Thomas Richter authored
Commit cf6acb8b ("s390/cpumf: Add support for complete counter set extraction") allows access to the CPU Measurement Counter Facility via character device /dev/hwctr. The access was exclusive via this device or via perf_event_open() system call. Only one path at a time was permitted. The CPU Measurement Counter Facility device driver blocked access to other processes. This patch removes this restriction and allows concurrent access to the CPU Measurement Counter Facility from multiple processes at the same time via perf_event_open() SVC and via /dev/hwctr device. The access via /dev/hwctr device is still exclusive, only one process is allowed to access this device. This patch - moves the /dev/hwctr device access from file perf_cpum_cf_diag.c. to file perf_cpum_cf.c. - use only one trace buffer .../s390dbf/cpum_cf. - remove cfset_csd structure and includes its members it into the structure cpu_cf_events. This results in one data structure and simplifies the access. - rework function familiy ctr_set_enable, ctr_set_disable, ctr_set_start and ctr_set_stop which operate on a counter set number. Now they operate on a counter set bit mask. - move CF_DIAG event functionality to file perf_cpum_cf.c. It now contains the complete functionality of the CPU Measurement Counter Facility: - Performance measurement support for counters using perf stat. - Support for complete counter set extraction with device /dev/hwctr. - Support for counter set extraction event CF_DIAG attached to samples using perf record. - removes file perf_cpum_cf_diag.c Signed-off-by:
Thomas Richter <tmricht@linux.ibm.com> Reviewed-by:
Sumanth Korikkar <sumanthk@linux.ibm.com> Signed-off-by:
-
Alexander Gordeev authored
This update partially reverts commit 3037a52f ("s390/nmi: do register validation as early as possible"). Storage error checks and control registers validation are left in the assembler code, since correct ASCEs and page tables are required to enable DAT - which is done before the C handler is entered. System damage, kernel instruction address and PSW MWP checks are left in the assembler code as well, since there is no way to proceed if one of these checks is failed. The getcpu vdso syscall reads CPU number from the programmable field of the TOD clock. Disregard the TOD programmable register validity bit and load the CPU number into the TOD programmable field unconditionally. Signed-off-by:
Alexander Gordeev <agordeev@linux.ibm.com> Reviewed-by:
-
Alexander Egorenkov authored
The magic string "S390EP" at offset 0x10008 indicated to the decompressed kernel that it was booted by the decompressor. Introduce a new bootdata flag instead which conveys the same information in an explicit and a cleaner way. But keep the magic string because it is a kernel ABI. Signed-off-by:
Alexander Egorenkov <egorenar@linux.ibm.com> Reviewed-by:
-
Alexander Gordeev authored
This update puts in order MCCK_CODE_CPU_TIMER_VALID code definition. Signed-off-by:
-
Alexander Gordeev authored
The current storage errors tackling is wrong - the DAT is enabled in assembler code before the actual storage checks in C half are executed. In case the page tables themselves are damaged such approach is not going to work. With this update unrecoverable storage errors are not passed to C code for handling, but rather the machine is stopped right away. The only exception to this flow is when a machine check occurred in KVM guest - in this case the errors are reinjected by the handler. Signed-off-by:
-
Alexander Gordeev authored
The machine check handler must be entered with DAT disabled in case control registers are corrupted or a storage error happened and we can not tell if such error corresponds to a page table. Both of described conditions end up in stopping all CPUs and entering the disabled wait in C half of the handler. However, the storage errors are still checked after the DAT is enabled and C code is entered. In case a page table is damaged such flow is not expected to work. This update paves the way for moving the storage error checks from C to assembler half. All fatal errors that can only be handled with DAT disabled are handled in assembler half also. As result, the C half is only entered if the DAT is secured. Signed-off-by:
-
Alexander Gordeev authored
In case of the !CONFIG_KVM use "jz" instead of "jnz" when detecting user mode and get rid of unnecessary jump as result. Signed-off-by:
Christia Borntraeger <borntraeger@de.ibm.com> Signed-off-by:
-
Alexander Gordeev authored
Factor out SIEEXIT macro and use it instead of cleanup_sie routine. As a side effect %r13 and %r14 are spared. Signed-off-by:
-
Heiko Carstens authored
General register 0 is clobbered within the inline assembly and therefore must be listed in the clobber list. Fixes: d1e18efa ("s390/lib,uaccess: get rid of register asm") Signed-off-by:
-
Janosch Frank authored
Turns out that the bit 61 in the TEID is not always 1 and if that's the case the address space ID and the address are unpredictable. Without an address and its address space ID we can't export memory and hence we can only send a SIGSEGV to the process or panic the kernel depending on who caused the exception. Unfortunately bit 61 is only reliable if we have the "misc" UV feature bit. Signed-off-by:
Janosch Frank <frankja@linux.ibm.com> Reviewed-by:
-
