- 20 Nov, 2020 15 commits
-
-
Heiko Carstens authored
While allmodconfig and allyesconfig build for s390 there are also various bots running compile tests with randconfig, where PCI is disabled. This reveals that a lot of drivers should actually depend on HAS_IOMEM. Adding this to each device driver would be a never ending story, therefore just disable COMPILE_TEST for s390. The reasoning is more or less the same as described in commit bc083a64 ("init/Kconfig: make COMPILE_TEST depend on !UML"). Reported-by:
kernel test robot <lkp@intel.com> Suggested-by:
Arnd Bergmann <arnd@kernel.org> Signed-off-by:
Heiko Carstens <hca@linux.ibm.com>
-
Alexander Gordeev authored
Rename some variable and functions to better clarify what they are and what they do. Signed-off-by:
Alexander Gordeev <agordeev@linux.ibm.com> Signed-off-by:
-
Alexander Gordeev authored
Reviewed-by:
David Hildenbrand <david@redhat.com> Signed-off-by:
-
Julian Wiedmann authored
Instead of creating the sysfs attributes for the stp root_dev by hand, pass them to subsys_system_register() as parameter. This also ensures that the attributes are available when the KOBJ_ADD event is raised. Signed-off-by:
Julian Wiedmann <jwi@linux.ibm.com> Signed-off-by:
-
Vasily Gorbik authored
Add kernel command line and last breaking event. The kernel command line is taken from early_command_line and printed only if kernel is not running as protected virtualization guest and if it has been already initialized from the "COMMAND_LINE". Linux version 5.10.0-rc3-22794-gecaa72788df0-dirty (gor@tuxmaker) #28 SMP PREEMPT Mon Nov 9 17:41:20 CET 2020 Kernel command line: audit_enable=0 audit=0 selinux=0 crashkernel=296M root=/dev/dasda1 dasd=ec5b memblock=debug die Kernel fault: interruption code 0005 ilc:2 PSW : 0000000180000000 0000000000012f92 (parse_boot_command_line+0x27a/0x46c) R:0 T:0 IO:0 EX:0 Key:0 M:0 W:0 P:0 AS:0 CC:0 PM:0 RI:0 EA:3 GPRS: 0000000000000000 00ffffffffffffff 0000000000000000 000000000001a65c 000000000000bf60 0000000000000000 00000000000003c0 0000000000000000 0000000000000080 000000000002322d 000000007f29ef20 0000000000efd018 000000000311c000 0000000000010070 0000000000012f82 000000000000bea8 Call Trace: (sp:000000000000bea8 [<000000000002016e>] 000000000002016e) sp:000000000000bf18 [<0000000000012408>] startup_kernel+0x88/0x2fc sp:000000000000bf60 [<00000000000100c4>] startup_normal+0xb0/0xb0 Last Breaking-Event-Address: [<00000000000135ba>] strcmp+0x22/0x24 Reviewed-by:Alexander Egorenkov <egorenar@linux.ibm.com> Acked-by:
Viktor Mihajlovski <mihajlov@linux.ibm.com> Signed-off-by:
Vasily Gorbik <gor@linux.ibm.com> Signed-off-by:
-
Vasily Gorbik authored
Decompressor works on a single statically allocated stack. Stacktrace implementation with -mbackchain just takes few lines of code. Linux version 5.10.0-rc3-22793-g0f84a355b776-dirty (gor@tuxmaker) #27 SMP PREEMPT Mon Nov 9 17:30:18 CET 2020 Kernel fault: interruption code 0005 ilc:2 PSW : 0000000180000000 0000000000012f92 (parse_boot_command_line+0x27a/0x46c) R:0 T:0 IO:0 EX:0 Key:0 M:0 W:0 P:0 AS:0 CC:0 PM:0 RI:0 EA:3 GPRS: 0000000000000000 00ffffffffffffff 0000000000000000 000000000001a62c 000000000000bf60 0000000000000000 00000000000003c0 0000000000000000 0000000000000080 000000000002322d 000000007f29ef20 0000000000efd018 000000000311c000 0000000000010070 0000000000012f82 000000000000bea8 Call Trace: (sp:000000000000bea8 [<000000000002016e>] 000000000002016e) sp:000000000000bf18 [<0000000000012408>] startup_kernel+0x88/0x2fc sp:000000000000bf60 [<00000000000100c4>] startup_normal+0xb0/0xb0 Reviewed-by: -
Vasily Gorbik authored
Information printed by print_pgm_check_info() is crucial for debugging decompressor problems. Printing instruction addresses is better than nothing, but turns further debugging into tedious job of figuring out which function those addresses correspond to. This change adds simplistic symbols resolution support. And adds %pS format specifier support to decompressor_printk(). Decompressor symbols list is extracted and sorted with nm -n -S: ... 0000000000010000 0000000000000014 T startup 0000000000010014 00000000000000b0 t startup_normal 0000000000010180 00000000000000b2 t startup_kdump ... Then functions are filtered and contracted to a form: "10000 14 startup\0""10014 b0 startup_normal\0""10180 b2 startup_kdump\0" ... Which makes it trivial to find beginning of an entry and names are 0 terminated, so could be used as is. Symbols are binary-searched. To get symbols list with final addresses and then get it into the decompressor's image the same trick as for kallsyms is used. Decompressor's vmlinux is linked twice. Symbols are stored in .decompressor.syms section, current size is about 2kb. Reviewed-by:
-
Vasily Gorbik authored
Use SYM_CODE_* annotations for asm functions, so that function lengths are recognized correctly. Also currently the most part of startup is marked as startup_kdump. Move misplaced startup_kdump where it belongs. $ nm -n -S arch/s390/boot/compressed/vmlinux Before: 0000000000010000 T startup 0000000000010010 T startup_kdump After: 0000000000010000 0000000000000014 T startup 0000000000010014 00000000000000b0 t startup_normal 0000000000010180 00000000000000b2 t startup_kdump Reviewed-by:
-
Vasily Gorbik authored
The decompressor does not have any special debug means. Running the kernel under qemu with gdb is helpful but tedious exercise if done repeatedly. It is also not applicable to debugging under LPAR and z/VM. One special thing which stands out is a working sclp_early_printk, which could be used once the kernel switches to 64-bit addressing mode. But sclp_early_printk does not provide any string formating capabilities. Formatting and printing string without printk-alike function is a not fun. The lack of printk-alike function means people would save up on testing and introduce more bugs. So, finally, provide decompressor_printk function, which fits on one screen and trades features for simplicity. It only supports "%s", "%x" and "%lx" specifiers and zero padding for hex values. Reviewed-by:
-
Vasily Gorbik authored
Currently the kernel minimal compiler requirement is gcc 4.9 or clang 10.0.1. * gcc -mhotpatch option is supported since 4.8. * A combination of -pg -mrecord-mcount -mnop-mcount -mfentry flags is supported since gcc 9 and since clang 10. Drop support for old -pg function prologues. Which leaves binary compatible -mhotpatch / -mnop-mcount -mfentry prologues in a form: brcl 0,0 Which are also do not require initial nop optimization / conversion and presence of _mcount symbol. Signed-off-by:
-
Vasily Gorbik authored
Currently we have to consider too many different values which in the end only affect identity mapping size. These are: 1. max_physmem_end - end of physical memory online or standby. Always <= end of the last online memory block (get_mem_detect_end()). 2. CONFIG_MAX_PHYSMEM_BITS - the maximum size of physical memory the kernel is able to support. 3. "mem=" kernel command line option which limits physical memory usage. 4. OLDMEM_BASE which is a kdump memory limit when the kernel is executed as crash kernel. 5. "hsa" size which is a memory limit when the kernel is executed during zfcp/nvme dump. Through out kernel startup and run we juggle all those values at once but that does not bring any amusement, only confusion and complexity. Unify all those values to a single one we should really care, that is our identity mapping size. Signed-off-by:
-
Julian Wiedmann authored
Instead of creating the sysfs attributes for the prng devices by hand, describe them in .groups and let the misdevice core handle it. This also ensures that the attributes are available when the KOBJ_ADD event is raised. Signed-off-by:
Harald Freudenberger <freude@linux.ibm.com> Signed-off-by:
-
Mauro Carvalho Chehab authored
Fix typo in the kernel-doc markups 1. ccw driver -> ccw_driver 2. ccw_device_id_is_equal() -> ccw_dev_id_is_equal Signed-off-by:
Mauro Carvalho Chehab <mchehab+huawei@kernel.org> Reviewed-by:
Cornelia Huck <cohuck@redhat.com> [vneethv@linux.ibm.com: slight modification in the changelog] Reviewed-by:
Vineeth Vijayan <vneethv@linux.ibm.com> Signed-off-by:
-
Heiko Carstens authored
System call and program check handler both use the system call exit path when returning to previous context. However the program check handler jumps right to the end of the system call exit path if the previous context is kernel context. This lead to the quite odd double disabling of interrupts in the system call exit path introduced with commit ce9dfafe ("s390: fix system call exit path"). To avoid that have a separate program check handler exit path if the previous context is kernel context. Reviewed-by:
Sven Schnelle <svens@linux.ibm.com> Signed-off-by:
-
Heiko Carstens authored
* fixes: s390/cpum_sf.c: fix file permission for cpum_sfb_size s390: update defconfigs s390: fix system call exit path Signed-off-by:
-
- 18 Nov, 2020 4 commits
-
-
Sumanth Korikkar authored
As the number of cpus increases, the sccb response can exceed 4k for read cpu and read scp info sclp commands. Hence, all cpu info entries cant be embedded within a sccb response Solution: To overcome this limitation, extended sccb facility is provided by sclp. 1. Check if the extended sccb facility is installed. 2. If extended sccb is installed, perform the read scp and read cpu command considering a max sccb length of three page size. This max length is based on factors like max cpus, sccb header. 3. If extended sccb is not installed, perform the read scp and read cpu sclp command considering a max sccb length of one page size. Signed-off-by:
Sumanth Korikkar <sumanthk@linux.ibm.com> Reviewed-by:
-
Sumanth Korikkar authored
For extended sccb support, sccb size could be up to 3 pages. Hence avoid copy of sclp_info_sccb. Signed-off-by:
-
Sumanth Korikkar authored
sclp early read cpu info is used to detect the number of configured cpus, which is utilized by smp_detect_cpus() in early startup. * For read cpu info, the sccb block should be below 2gb. * smp_detect_cpus() utilizes read cpu info early, but after memblock initialization. Thus use memblock_allow_low() instead. * Avoid copy of sclp_core_info structure. * sclp_early_init_core_info(), sclp_early_core_info and sclp_early_core_info_valid initdata are no longer required. * smp_get_core_info() is called only once during early stage. Hence for early sclp_get_core_info(), directly call read cpu command. No need to maintain sclp_early_core_info_valid. Signed-off-by:
-
Niklas Schnelle authored
when we're missing the necessary machine facilities zPCI can not function. Until now it would silently fail to be initialized, add an informational print. Signed-off-by:
Niklas Schnelle <schnelle@linux.ibm.com> Signed-off-by:
-
- 12 Nov, 2020 3 commits
-
-
Heiko Carstens authored
Signed-off-by: -
Thomas Richter authored
This file is installed by the s390 CPU Measurement sampling facility device driver to export supported minimum and maximum sample buffer sizes. This file is read by lscpumf tool to display the details of the device driver capabilities. The lscpumf tool might be invoked by a non-root user. In this case it does not print anything because the file contents can not be read. Fix this by allowing read access for all users. Reading the file contents is ok, changing the file contents is left to the root user only. For further reference and details see: [1] https://github.com/ibm-s390-tools/s390-tools/issues/97 Fixes: 69f239ed ("s390/cpum_sf: Dynamically extend the sampling buffer if overflows occur") Cc: <stable@vger.kernel.org> # 3.14 Signed-off-by:
Thomas Richter <tmricht@linux.ibm.com> Acked-by:
-
Heiko Carstens authored
Signed-off-by:
-
- 09 Nov, 2020 18 commits
-
-
Harald Freudenberger authored
The zcrypt api provides a new function to wait until the zcrypt api is operational: int zcrypt_wait_api_operational(void); The AP bus scan and the binding of ap devices to device drivers is an asynchronous job. This function waits until these initial jobs are done and so the zcrypt api should be ready to serve crypto requests - if there are resources available. The function uses an internal timeout of 60s. The very first caller will either wait for ap bus bindings complete or the timeout happens. This state will be remembered for further callers which will only be blocked until a decision is made (timeout or bindings complete). Reviewed-by:
Ingo Franzki <ifranzki@linux.ibm.com> Signed-off-by:
-
Harald Freudenberger authored
This patch adds notifications to userspace for two important conditions of the ap bus: I) Initial ap bus scan done. This indicates that the initial scan of all the ap devices (cards, queues) is complete and ap devices have been build up for all the hardware found. This condition is signaled with 1) An ap bus change uevent send to userspace with an environment key/value pair "INITSCAN=done": # udevadm monitor -k -p ... KERNEL[97.830919] change /devices/ap (ap) ACTION=change DEVPATH=/devices/ap SUBSYSTEM=ap INITSCAN=done SEQNUM=10421 2) A sysfs attribute /sys/bus/ap/scans which shows the number of completed ap bus scans done since bus init. So a value of 1 or greater signals that the initial ap bus scan is complete. Note: The initial ap bus scan complete condition is fulfilled and will be signaled even if there was no ap resource found. II) APQN driver bindings complete. This indicates that all APQNs have been bound to an zcrypt or alternate device driver. Only with the help of an device driver an APQN can be used for crypto load. So the binding complete condition is the starting point for user space to be sure all crypto resources on the ap bus are available for use. This condition is signaled with 1) An ap bus change uevent send to userspace with an environment key/value pair "BINDINGS=complete": # udevadm monitor -k -p ... KERNEL[97.830975] change /devices/ap (ap) ACTION=change DEVPATH=/devices/ap SUBSYSTEM=ap BINDINGS=complete SEQNUM=10422 2) A sysfs attribute /sys/bus/ap/bindings showing "<nr of bound apqns>/<total nr of apqns> (complete)" when all available apqns have been bound to device drivers, or "<nr of bound apqns>/<total nr of apqns>" when there are some apqns not bound to an device driver. Note: The binding complete condition is also fulfilled, when there are no apqns available to bind any device driver. In this case the binding complete will be signaled AFTER init scan is done. Note: This condition may arise multiple times when after initial scan modifications on the bindings take place. For example a manual unbind of an APQN switches the binding complete condition off. When at a later time the unbound APQNs are bound with an device driver the binding is (again) complete resulting in another uevent and marking the bindings sysfs attribute with '(complete)'. There is also a new function to be used within the kernel: int ap_wait_init_apqn_bindings_complete(unsigned long timeout) Interface to wait for the AP bus to have done one initial ap bus scan and all detected APQNs have been bound to device drivers. If these both conditions are not fulfilled, this function blocks on a condition with wait_for_completion_interruptible_timeout(). If these both conditions are fulfilled (before the timeout hits) the return value is 0. If the timeout (in jiffies) hits instead -ETIME is returned. On failures negative return values are returned to the caller. Please note that further unbind/bind actions after initial binding complete is through do not cause this function to block again. Reviewed-by: -
Christian Borntraeger authored
The s390-trng does provide 100% entropy. The quality value is supported to be between 1 and 1024 and not 1..1000. Use 1024 to make this driver the preferred one. If we ever have a better driver that has the same quality but is faster we can change this again when merging the new driver. No need to be conservative. This makes sure that the hw variant is preferred over things like virtio-rng, where the hypervisor has a potential to be misconfigured and thus should have a slightly lower confidence. Cc: Harald Freudenberger <freude@linux.ibm.com> Signed-off-by:
Christian Borntraeger <borntraeger@de.ibm.com> Signed-off-by:
-
Daniel Vetter authored
Way back it was a reasonable assumptions that iomem mappings never change the pfn range they point at. But this has changed: - gpu drivers dynamically manage their memory nowadays, invalidating ptes with unmap_mapping_range when buffers get moved - contiguous dma allocations have moved from dedicated carvetouts to cma regions. This means if we miss the unmap the pfn might contain pagecache or anon memory (well anything allocated with GFP_MOVEABLE) - even /dev/mem now invalidates mappings when the kernel requests that iomem region when CONFIG_IO_STRICT_DEVMEM is set, see commit 3234ac66 ("/dev/mem: Revoke mappings when a driver claims the region") Accessing pfns obtained from ptes without holding all the locks is therefore no longer a good idea. Fix this. Since zpci_memcpy_from|toio seems to not do anything nefarious with locks we just need to open code get_pfn and follow_pfn and make sure we drop the locks only after we're done. The write function also needs the copy_from_user move, since we can't take userspace faults while holding the mmap sem. Reviewed-by:
Gerald Schaefer <gerald.schaefer@linux.ibm.com> Signed-off-by:
Daniel Vetter <daniel.vetter@intel.com> Cc: Jason Gunthorpe <jgg@ziepe.ca> Cc: Dan Williams <dan.j.williams@intel.com> Cc: Kees Cook <keescook@chromium.org> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: John Hubbard <jhubbard@nvidia.com> Cc: JÃ
© rôme Glisse <jglisse@redhat.com> Cc: Jan Kara <jack@suse.cz> Cc: linux-mm@kvack.org Cc: linux-arm-kernel@lists.infradead.org Cc: linux-samsung-soc@vger.kernel.org Cc: linux-media@vger.kernel.org Cc: Gerald Schaefer <gerald.schaefer@linux.ibm.com> Cc: linux-s390@vger.kernel.org Cc: Niklas Schnelle <schnelle@linux.ibm.com> Signed-off-by:Daniel Vetter <daniel.vetter@ffwll.ch> Signed-off-by:
-
Vasily Gorbik authored
And move it earlier in the decompressor. Reviewed-by:
-
Vasily Gorbik authored
Also correct rounding downs in estimation calculations. Reviewed-by:
-
Vasily Gorbik authored
It is relying on _REGION1_SHIFT / _REGION2_SHIFT values which come from asm/pgtable.h, so include it. Reviewed-by:
-
Vasily Gorbik authored
Kasan early code is only working on init_mm, remove unneeded pgd parameter from kasan_copy_shadow and rename it to kasan_copy_shadow_mapping. Reviewed-by:
-
Vasily Gorbik authored
Kasan has nothing to do with vmemmap, strip vmemmap from function names to avoid confusing people. Reviewed-by:
-
Vasily Gorbik authored
Fixes the following warning with CONFIG_KERNEL_UNCOMPRESSED=y arch/s390/boot/compressed/decompressor.h:6:46: warning: non-void function does not return a value [-Wreturn-type] static inline void *decompress_kernel(void) {} ^ Signed-off-by: -
Heiko Carstens authored
To make sure that the vmalloc area size is for almost all cases large enough let it depend on the (potential) physical memory size. There is still the possibility to override this with the vmalloc kernel command line parameter. Reviewed-by:
Claudio Imbrenda <imbrenda@linux.ibm.com> Reviewed-by:
-
Heiko Carstens authored
We've seen several occurences in the past where the default vmalloc size of 128GB is not sufficient. Therefore extend the default size. Reviewed-by:
-
Vasily Gorbik authored
Since commit 29d37e5b ("s390/protvirt: add ultravisor initialization") vmax is adjusted to the ultravisor secure storage limit. This limit is currently applied when 4-level paging is used. Later vmax is also used to align vmemmap address to the top region table entry border. When vmax is set to the ultravisor secure storage limit this is no longer the case. Instead of changing vmax, make only MODULES_END be affected by the secure storage limit, so that vmax stays intact for further vmemmap address alignment. Reviewed-by:
-
Vasily Gorbik authored
Compiling the kernel with Kasan disables automatic 3-level vs 4-level kernel space paging selection, because the shadow memory offset has to be known at compile time and there is no such offset which would be acceptable for both 3 and 4-level paging. Instead S390_4_LEVEL_PAGING option was introduced which allowed to pick how many paging levels to use under Kasan. With the introduction of protected virtualization, kernel memory layout may be affected due to ultravisor secure storage limit. This adds additional complexity into how memory layout would look like in combination with Kasan predefined shadow memory offsets. To simplify this make Kasan 4-level paging default and remove Kasan 3-level paging support. Suggested-by:
-
Vasily Gorbik authored
s390_base_ext_handler_fn haven't been used since its introduction in commit ab14de6c ("[S390] Convert memory detection into C code."). s390_base_ext_handler itself is currently falsely storing 16 registers at __LC_SAVE_AREA_ASYNC rewriting several following lowcore values: cpu_flags, return_psw, return_mcck_psw, sync_enter_timer and async_enter_timer. Besides that s390_base_ext_handler itself is only potentially hiding EXT interrupts which should not have happen in the first place. Any piece of code which requires EXT interrupts before fully functional ext_int_handler is enabled has to do it on its own, like this is done by sclp_early_cmd() which is doing EXT interrupts handling synchronously in sclp_early_wait_irq(). With s390_base_ext_handler removed unexpected EXT interrupt leads to disabled wait with the address 0x1b0 (__LC_EXT_NEW_PSW), which is currently setup in the decompressor. Reviewed-by:
-
Vasily Gorbik authored
Currently udelay relies on working EXT interrupts handler, which is not the case during early startup. In such cases udelay_simple() has to be used instead. To avoid mistakes of calling udelay too early, which could happen from the common code as well - make udelay work for the early code by introducing static branch and redirecting all udelay calls to udelay_simple until EXT interrupts handler is fully initialized and async stack is allocated. Reviewed-by:
-
Vasily Gorbik authored
Set io/ext handlers to disabled wait in the initial lowcore, so that they are effective right from the kernel start, when a boot method used does not rewrite this part of the lowcore for its own needs (i.e. kexec, z/vm ipl reader boot, qemu direct boot, load from removable media or server). When the kernel is loaded by zipl, scsi loader or qemu loader, some or all of the io/ext/pgm handlers addresses might be rewritten. Rewrite them to initial values again as early as possible. Reviewed-by:
-
Heiko Carstens authored
The system call exit path is running with interrupts enabled while checking for TIF/PIF/CIF bits which require special handling. If all bits have been checked interrupts are disabled and the kernel exits to user space. The problem is that after checking all bits and before interrupts are disabled bits can be set already again, due to interrupt handling. This means that the kernel can exit to user space with some TIF/PIF/CIF bits set, which should never happen. E.g. TIF_NEED_RESCHED might be set, which might lead to additional latencies, since that bit will only be recognized with next exit to user space. Fix this by checking the corresponding bits only when interrupts are disabled. Fixes: 0b0ed657 ("s390: remove critical section cleanup from entry.S") Cc: <stable@vger.kernel.org> # 5.8 Acked-by:
-
