- 01 Jun, 2022 13 commits
-
-
Heiko Carstens authored
Add a union which describes how the empty stack slots are being used by kvm and perf. This should help to avoid another bug like the one which was fixed with commit c9bfb460 ("s390/perf: obtain sie_block from the right address"). Reviewed-by:
Nico Boehr <nrb@linux.ibm.com> Tested-by:
Heiko Carstens <hca@linux.ibm.com>
-
Heiko Carstens authored
Merge empty1 and empty2 arrays within the stack frame to one single array. This is possible since with commit 42b01a55 ("s390: always use the packed stack layout") the alternative stack frame layout is gone. Reviewed-by:
Alexander Gordeev <agordeev@linux.ibm.com> Signed-off-by:
-
Heiko Carstens authored
Whitespace cleanup to get rid if some checkpatch findings, but mainly to have consistent coding style within the header file again. Signed-off-by: -
Heiko Carstens authored
Signed-off-by: -
Heiko Carstens authored
Historically the uaccess code pre-initializes the result of get_user() (and now also __get_kernel_nofault()) to zero and uses the result as input parameter for inline assemblies. This is different to what most, if not all, other architectures are doing, which set the result to zero within the exception handler in case of a fault. Use the new extable mechanism and handle zeroing of the result within the exception handler in case of a fault. Signed-off-by: -
Heiko Carstens authored
Make code easier to read by using symbolic names. Signed-off-by: -
Alexander Gordeev authored
Commit d768bd89 ("s390: add options to change branch prediction behaviour for the kernel") introduced .Lsie_exit label - supposedly to fence off SIE instruction. However, the corresponding address range length .Lsie_crit_mcck_length was not updated, which led to BPON code potentionally marked with CIF_MCCK_GUEST flag. Both .Lsie_exit and .Lsie_crit_mcck_length were removed with commit 0b0ed657 ("s390: remove critical section cleanup from entry.S"), but the issue persisted - currently BPOFF and BPENTER macros might get wrongly considered by the machine check handler as a guest. Fixes: d768bd89 ("s390: add options to change branch prediction behaviour for the kernel") Reviewed-by:
Sven Schnelle <svens@linux.ibm.com> Reviewed-by:
Christian Borntraeger <borntraeger@linux.ibm.com> Signed-off-by:
-
Christian Borntraeger authored
The switch to a keyed guest does not require a classic sske as the other guest CPUs are not accessing the key before the switch is complete. By using the NQ SSKE things are faster especially with multiple guests. Signed-off-by:
Janis Schoetterl-Glausch <scgl@linux.ibm.com> Reviewed-by:
Claudio Imbrenda <imbrenda@linux.ibm.com> Link: https://lore.kernel.org/r/20220530092706.11637-3-borntraeger@linux.ibm.comSigned-off-by:
-
Christian Borntraeger authored
With large and many guest with storage keys it is possible to create large latencies or stalls during initial key setting: rcu: INFO: rcu_sched self-detected stall on CPU rcu: 18-....: (2099 ticks this GP) idle=54e/1/0x4000000000000002 softirq=35598716/35598716 fqs=998 (t=2100 jiffies g=155867385 q=20879) Task dump for CPU 18: CPU 1/KVM R running task 0 1030947 256019 0x06000004 Call Trace: sched_show_task rcu_dump_cpu_stacks rcu_sched_clock_irq update_process_times tick_sched_handle tick_sched_timer __hrtimer_run_queues hrtimer_interrupt do_IRQ ext_int_handler ptep_zap_key The mmap lock is held during the page walking but since this is a semaphore scheduling is still possible. Same for the kvm srcu. To minimize overhead do this on every segment table entry or large page. Signed-off-by: -
Eric Farman authored
Add myself to the kernel side of virtio-ccw Signed-off-by:
Eric Farman <farman@linux.ibm.com> Reviewed-by:
Cornelia Huck <cohuck@redhat.com> Acked-by:
Halil Pasic <pasic@linux.ibm.com> Link: https://lore.kernel.org/r/20220525144028.2714489-2-farman@linux.ibm.comSigned-off-by:
-
Heiko Carstens authored
Avoid invoking the OOM-killer when allocating the control page. This is the s390 variant of commit dc5cccac ("kexec: don't invoke OOM-killer for control page allocation"). Signed-off-by:
-
Juerg Haefliger authored
The convention for indentation seems to be a single tab. Help text is further indented by an additional two whitespaces. Fix the lines that violate these rules. Signed-off-by:
Juerg Haefliger <juerg.haefliger@canonical.com> Link: https://lore.kernel.org/r/20220525120151.39594-1-juerg.haefliger@canonical.comSigned-off-by:
-
Juerg Haefliger authored
The convention for indentation seems to be a single tab. Help text is further indented by an additional two whitespaces. Fix the lines that violate these rules. Signed-off-by:
-
- 25 May, 2022 4 commits
-
-
Nico Boehr authored
Since commit 1179f170 ("s390: fix fpu restore in entry.S"), the sie_block pointer is located at empty1[1], but in sie_block() it was taken from empty1[0]. This leads to a random pointer being dereferenced, possibly causing system crash. This problem can be observed when running a simple guest with an endless loop and recording the cpu-clock event: sudo perf kvm --guestvmlinux=<guestkernel> --guest top -e cpu-clock With this fix, the correct guest address is shown. Fixes: 1179f170 ("s390: fix fpu restore in entry.S") Cc: stable@vger.kernel.org Acked-by:
Christian Borntraeger <borntraeger@de.ibm.com> Acked-by:
-
Heiko Carstens authored
Use asm offsets method to generate register offsets into pt_regs, instead of open-coding at several places. Reviewed-by:
-
Heiko Carstens authored
Due to historic reasons the base program check handler calls a configurable function. Given that there is only the early program check handler left, simplify the code by directly calling that function. The only other user was removed with commit d485235b ("s390: assume diag308 set always works"). Also rename all functions and the asm file to reflect this. Reviewed-by:
-
Jann Horn authored
The argument of scatterwalk_unmap() is supposed to be the void* that was returned by the previous scatterwalk_map() call. The s390 AES-GCM implementation was instead passing the pointer to the struct scatter_walk. This doesn't actually break anything because scatterwalk_unmap() only uses its argument under CONFIG_HIGHMEM and ARCH_HAS_FLUSH_ON_KUNMAP. Fixes: bf7fa038 ("s390/crypto: add s390 platform specific aes gcm support.") Signed-off-by:
Jann Horn <jannh@google.com> Acked-by:
Harald Freudenberger <freude@linux.ibm.com> Link: https://lore.kernel.org/r/20220517143047.3054498-1-jannh@google.comSigned-off-by:
-
- 18 May, 2022 1 commit
-
-
Heiko Carstens authored
Get rid of old 31 bit leftovers within ipl code: - convert everything to pc relative code - use 64 bit addressing mode as early as possible - use 64 bit arithmetics wherever possible This way the code doesn't look as odd as before anymore. Reviewed-by:
-
- 17 May, 2022 7 commits
-
-
Heiko Carstens authored
Before version 14.0.0 llvm's integrated assembler fails to handle some displacement variants: arch/s390/purgatory/head.S:108:10: error: invalid operand for instruction lg %r11,kernel_type-.base_crash(%r13) Instead of working around this and given that this is already fixed raise the minimum clang version from 13.0.0 to 14.0.0. Acked-by:
Nick Desaulniers <ndesaulniers@google.com> Tested-by:
Nathan Chancellor <nathan@kernel.org> Tested-by:
-
Heiko Carstens authored
Commit ee6d777d ("s390/decompressor: support extra debug flags") added extra debug flags, in particular debug info is created, depending on config options. With llvm's IAS this causes this compile warning: arch/s390/boot/head.S:38:1: warning: DWARF2 only supports one section per compilation unit .section ".head.text","ax" ^ This is a known problem and was addressed with commit b8a90923 ("Kbuild: do not emit debug info for assembly with LLVM_IAS=1"). Just do the same for s390 to get rid of this warning. Tested-by:
-
Heiko Carstens authored
For at least the mvc and clc instructions llvm's integrated assembler can generate incorrect code. In particular this happens with decompressor boot code. The reason seems to be that relocations for the second displacement of each instruction are at incorrect locations (-/+: gas vs llvm IAS): mvc __LC_IO_NEW_PSW(16),.Lnewpsw results in 4: d2 0f 01 f0 00 00 mvc 496(16,%r0),0 - 8: R_390_12 .head.text+0x10 + 6: R_390_12 .head.text+0x10 and clc 0(3,%r4),.L_hdr results in 258: d5 02 40 00 00 00 clc 0(3,%r4),0 - 25c: R_390_12 .head.text+0x324 + 25a: R_390_12 .head.text+0x324 Workaround this by writing the code in a different way. Tested-by: -
Heiko Carstens authored
llvm's integrated assembler cannot handle immediate values which are calculated with two local labels: arch/s390/purgatory/head.S:139:11: error: invalid operand for instruction aghi %r8,-(.base_crash-purgatory_start) Workaround this by partially rewriting the code. Link: https://lore.kernel.org/r/20220511120532.2228616-6-hca@linux.ibm.comSigned-off-by:
-
Heiko Carstens authored
llvm's integrated assembler cannot handle immediate values which are calculated with two local labels: <instantiation>:3:13: error: invalid operand for instruction clgfi %r14,.Lsie_done - .Lsie_gmap Workaround this by adding clang specific code which reads the specific value from memory. Since this code is within the hot paths of the kernel and adds an additional memory reference, keep the original code, and add ifdef'ed code. Acked-by:
-
Heiko Carstens authored
clang fails to handle ".if" statements in inline assembly which are heavily used in the alternatives code. To work around this remove this code, and enforce that users of alternatives must specify original and alternative instruction sequences which have identical sizes. Add a compile time check with two ".org" statements similar to arm64. In result not only clang can handle this, but also quite a lot of code can be removed. Acked-by:
Vasily Gorbik <gor@linux.ibm.com> Tested-by:
-
Heiko Carstens authored
Explicitly provide identical sized original/alternative instruction sequences. This way there is no need for the s390 specific alternatives infrastructure to generate padding sequences. The code which generates such sequences will be removed with a follow on patch. Acked-by:
-
- 16 May, 2022 1 commit
-
-
Thomas Richter authored
Export the extended counter set counters of the IBM z16 via sysfs. Signed-off-by:
Thomas Richter <tmricht@linux.ibm.com> Acked-by:
Sumanth Korikkar <sumanthk@linux.ibm.com> Signed-off-by:
-
- 11 May, 2022 3 commits
-
-
Heiko Carstens authored
gcc 12 does not (always) optimize away code that should only be generated if parameters are constant and within in a certain range. This depends on various obscure kernel config options, however in particular PROFILE_ALL_BRANCHES can trigger this compile error: In function ‘__atomic_add_const’, inlined from ‘__preempt_count_add.part.0’ at ./arch/s390/include/asm/preempt.h:50:3: ./arch/s390/include/asm/atomic_ops.h:80:9: error: impossible constraint in ‘asm’ 80 | asm volatile( \ | ^~~ Workaround this by simply disabling the optimization for PROFILE_ALL_BRANCHES, since the kernel will be so slow, that this optimization won't matter at all. Reported-by: -
Sven Schnelle authored
clock_delta is declared as unsigned long in various places. However, the clock sync delta can be negative. This would add a huge positive offset in clock_sync_global where clock_delta is added to clk.eitod which is a 72 bit integer. Declare it as signed long to fix this. Cc: stable@vger.kernel.org Signed-off-by:
-
Sven Schnelle authored
The size of the TOD offset field in the stp info response is 64 bits. Signed-off-by:
Heiko Carstens <hca@de.ibm.com> Signed-off-by:
-
- 09 May, 2022 2 commits
-
-
Thomas Richter authored
PMU device driver perf_pai_crypto supports Processor Activity Instrumentation (PAI), available with IBM z16: - maps a full page to lowcore address 0x1500. - uses CR0 bit 13 to turn PAI crypto counting on and off. - creates a sample with raw data on each context switch out when at context switch some mapped counters have a value of nonzero. This device driver only supports CPU wide context, no task context is allowed. Support for counting: - one or more counters can be specified using perf stat -e pai_crypto/xxx/ where xxx stands for the counter event name. Multiple invocation of this command is possible. The counter names are listed in /sys/devices/pai_crypto/events directory. - one special counters can be specified using perf stat -e pai_crypto/CRYPTO_ALL/ which returns the sum of all incremented crypto counters. - one event pai_crypto/CRYPTO_ALL/ is reserved for sampling. No multiple invocations are possible. The event collects data at context switch out and saves them in the ring buffer. Add qpaci assembly instruction to query supported memory mapped crypto counters. It returns the number of counters (no holes allowed in that range). The PAI crypto counter events are system wide and can not be executed in parallel. Therefore some restrictions documented in function paicrypt_busy apply. In particular event CRYPTO_ALL for sampling must run exclusive. Only counting events can run in parallel. PAI crypto counter events can not be created when a CPU hot plug add is processed. This means a CPU hot plug add does not get the necessary PAI event to record PAI cryptography counter increments on the newly added CPU. CPU hot plug remove removes the event and terminates the counting of PAI counters immediately. Co-developed-by:
Juergen Christ <jchrist@linux.ibm.com> Signed-off-by:
-
Sven Schnelle authored
arch_check_user_regs() is used at the moment to verify that struct pt_regs contains valid values when entering the kernel from userspace. s390 needs a place in the generic entry code to modify a cpu data structure when switching from userspace to kernel mode. As arch_check_user_regs() is exactly this, rename it to arch_enter_from_user_mode(). When entering the kernel from userspace, arch_check_user_regs() is used to verify that struct pt_regs contains valid values. Note that the NMI codepath doesn't call this function. s390 needs a place in the generic entry code to modify a cpu data structure when switching from userspace to kernel mode. As arch_check_user_regs() is exactly this, rename it to arch_enter_from_user_mode(). Signed-off-by:
Thomas Gleixner <tglx@linutronix.de> Acked-by:
Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Andy Lutomirski <luto@kernel.org> Link: https://lore.kernel.org/r/20220504062351.2954280-2-tmricht@linux.ibm.comSigned-off-by:
-
- 06 May, 2022 9 commits
-
-
Heiko Carstens authored
Remove various declarations from former s390 specific compat system calls which have been removed with commit fef747ba ("s390: use generic UID16 implementation"). While at it clean up the whole small header file. Signed-off-by:
-
Heiko Carstens authored
LLVM's integrated assembler reports the following error when compiling entry.S: <instantiation>:38:5: error: unknown token in expression tm %r8,0x0001 # coming from user space? The correct instruction would have been tmhh instead of tm. The current code is doing nothing, since (with gas) it get's translated to a tm instruction which reads from real address 8, which again contains always zero, and therefore the conditional code is never executed. Note that due to the missing displacement gas translates "%r8" into "8(%r0)". Also code inspection reveals that this conditional code is not needed. Therefore remove it. Reviewed-by:
-
Heiko Carstens authored
Convert parmarea to C, which makes it much easier to initialize it. No need to keep offsets in assembler code in sync with struct parmarea anymore. Reviewed-by:
-
Heiko Carstens authored
Convert initial lowcore to C and use proper defines and structures to initialize it. This should make the z/VM ipl procedure a bit less magic. Acked-by:
Peter Oberparleiter <oberpar@linux.ibm.com> Reviewed-by:
-
Heiko Carstens authored
The short psw definitions are contained in compat header files, however short psws are not compat specific. Therefore move the definitions to ptrace header file. This also gets rid of a compat header include in kvm code. Acked-by:
-
Heiko Carstens authored
Initialize all new psws with disabled wait psws, except for the restart new psw. This way every unexpected exception, svc, machine check, or interrupt is handled properly. Reviewed-by:
-
Heiko Carstens authored
The program check handler of the kernel image points to startup_pgm_check_handler. However an early program check which happens while loading the kernel image will jump to potentially random code, since the code of the program check handler is not yet loaded; leading to a program check loop. Therefore initialize it to a disabled wait psw and let the startup code set the proper psw when everything is in memory. Reviewed-by:
-
Heiko Carstens authored
Move iplstart entry point to 0x200 again, instead of the middle of the ipl code. This way even the comment describing the ccw program is correct again. Acked-by:
-
Heiko Carstens authored
The final kernel image is created by linking decompressor object files with a startup archive. The startup archive file however does not contain only optional code and data which can be discarded if not referenced. It also contains mandatory object data like head.o which must never be discarded, even if not referenced. Move the decompresser code and linker script to the boot directory and get rid of the startup archive so everything is kept during link time. Acked-by:
-
