- 17 Feb, 2017 1 commit
-
-
Michael Holzheu authored
In binutils/libbfd (bfd/elf.c) it is enforced that all s390 specific ELF notes like e.g. NT_S390_PREFIX or NT_S390_CTRS have "LINUX" specified as note name. Otherwise the notes are ignored. For /proc/vmcore we currently use "CORE" for these notes. Up to now this has not been a real problem because the dump analysis tool "crash" does not check the note name. But it will break all programs that use libbfd for processing ELF notes. So fix this and use "LINUX" for all s390 specific notes to comply with libbfd. Cc: stable@vger.kernel.org # v4.4+ Reported-by:
Philipp Rudo <prudo@linux.vnet.ibm.com> Reviewed-by:
Michael Holzheu <holzheu@linux.vnet.ibm.com> Signed-off-by:
Martin Schwidefsky <schwidefsky@de.ibm.com>
-
- 08 Feb, 2017 8 commits
-
-
Martin Schwidefsky authored
Bit 0x100 of a page table, segment table of region table entry can be used to disallow code execution for the virtual addresses associated with the entry. There is one tricky bit, the system call to return from a signal is part of the signal frame written to the user stack. With a non-executable stack this would stop working. To avoid breaking things the protection fault handler checks the opcode that caused the fault for 0x0a77 (sys_sigreturn) and 0x0aad (sys_rt_sigreturn) and injects a system call. This is preferable to the alternative solution with a stub function in the vdso because it works for vdso=off and statically linked binaries as well. Signed-off-by: -
Martin Schwidefsky authored
Add hardware capability bits and feature tags to /proc/cpuinfo for the "Vector Packed Decimal Facility" (tag "vxd" / hwcap bit 2^12) and the "Vector Enhancements Facility 1" (tag "vxe" / hwcap bit 2^13). Signed-off-by: -
Heiko Carstens authored
The current implementation of setup_randomness uses the stack address and therefore the pointer to the SYSIB 3.2.2 block as input data address. Furthermore the length of the input data is the number of virtual-machine description blocks which is typically one. This means that typically a single zero byte is fed to add_device_randomness. Fix both of these and use the address of the first virtual machine description block as input data address and also use the correct length. Fixes: bcfcbb6b ("s390: add system information as device randomness") Signed-off-by:
Heiko Carstens <heiko.carstens@de.ibm.com> Signed-off-by:
-
Heiko Carstens authored
Get rid of common response code handling. Each command requires its own response code handling anyway. Also the retry in case of -EBUSY does not work and can be simply removed. Reviewed-by:
Peter Oberparleiter <oberpar@linux.vnet.ibm.com> Signed-off-by:
-
Heiko Carstens authored
The early vt220 sclp printk code added an extra new line to each printed multi-line text. If used for the early sclp console this will lead to numerous extra new lines. Therefore get rid of this semantic and require that each to be printed string contains a line feed character if a new line is wanted. Reviewed-by:
-
Heiko Carstens authored
This patch - unifies the old sclp early code and the sclp early printk code, so they can use common functions - makes sure all sclp early functions and variables have the same "sclp_early" prefix - converts the sclp early printk code into readable code by using existing data structures instead of hard coded magic arrays - splits the early sclp code into two files: sclp_early.c and sclp_early_core.c. The core file contains everything that is required by the kernel decompressor and may not call functions not contained within the core file. Otherwise the result would be a link error. - changes interrupt handling to be completely synchronous. The old early sclp code had a small window which allowed to receive several interrupts instead of exactly the single expected interrupt. This did hide a subtle potential bug, which is fixed with this large rework. - contains a couple of small cleanups. Reviewed-by:
-
Heiko Carstens authored
Make sure the early sclp code does not generate any sclp requests anymore as soon as the base sclp driver is active. Otherwise both drivers may see unexpected requests or may miss expected interrupts. Reviewed-by:
-
Heiko Carstens authored
Move the early sclp printk code to the drivers folder where also the rest of the sclp code can be found. This way it is possible to use the sclp private header files for further cleanups. Reviewed-by:
-
- 07 Feb, 2017 3 commits
-
-
Martin Schwidefsky authored
The debug features currently uses absolute TOD time stamps for the debug events. Given that the TOD clock can jump forward and backward due to STP sync checks the order of debug events can get obfuscated. Replace the absolute TOD time stamps with a delta to the IPL time stamp. On a STP sync check the TOD clock correction is added to the IPL time stamp as well to make the deltas unaffected by STP sync check. The readout of the debug feature entries will convert the deltas back to absolute time stamps based on the Unix epoch. Signed-off-by: -
Martin Schwidefsky authored
The data stored by the STSI instruction can be up to a page in size but the memblock_virt_alloc allocation for tl_info only specifies 16 bytes. The memory after the short allocation is overwritten every time arch_update_cpu_topology is called. Fixes: 8c910580 "s390/numa: establish cpu to node mapping early" Signed-off-by:
-
Heiko Carstens authored
Commit bcfcbb6b ("s390: add system information as device randomness") intended to add some virtual machine specific information to the randomness pool. Unfortunately it uses the page allocator before it is ready to use. In result the page allocator always returns NULL and the setup_randomness function never adds anything to the randomness pool. To fix this use memblock_alloc and memblock_free instead. Fixes: bcfcbb6b ("s390: add system information as device randomness") Signed-off-by:
-
- 03 Feb, 2017 5 commits
-
-
Julian Wiedmann authored
With multiple input queues, these DBFs turned out to be not very helpful... Signed-off-by:
Julian Wiedmann <jwi@linux.vnet.ibm.com> Signed-off-by:
-
Julian Wiedmann authored
Missed in commit f4eae94f ("s390/airq: simplify adapter interrupt code") Signed-off-by:
Steffen Maier <maier@linux.vnet.ibm.com> Signed-off-by:
-
Julian Wiedmann authored
In tiqdio_call_inq_handlers(), we're looping over all input queues on the *same* irq. So instead of using the queues' back pointer, we can just access the irq directly. No functional change. Signed-off-by:
-
Julian Wiedmann authored
For devices with multiple input queues, tiqdio_call_inq_handlers() iterates over all input queues and clears the device's DSCI during each iteration. If the DSCI is re-armed during one of the later iterations, we therefore do not scan the previous queues again. The re-arming also raises a new adapter interrupt. But its handler does not trigger a rescan for the device, as the DSCI has already been erroneously cleared. This can result in queue stalls on devices with multiple input queues. Fix it by clearing the DSCI just once, prior to scanning the queues. As the code is moved in front of the loop, we also need to access the DSCI directly (ie irq->dsci) instead of going via each queue's parent pointer to the same irq. This is not a functional change, and a follow-up patch will clean up the other users. In practice, this bug only affects CQ-enabled HiperSockets devices, ie. devices with sysfs-attribute "hsuid" set. Setting a hsuid is needed for AF_IUCV socket applications that use HiperSockets communication. Fixes: 104ea556 ("qdio: support asynchronous delivery of storage blocks") Cc: <stable@vger.kernel.org> # v3.2+ Reviewed-by:
Ursula Braun <ubraun@linux.vnet.ibm.com> Signed-off-by:
-
Gerald Schaefer authored
Since commit dd22f551 "block: Change direct_access calling convention", the device size calculation in dcssblk_direct_access() is off-by-one. This results in bdev_direct_access() always returning -ENXIO because the returned value is not page aligned. Fix this by adding 1 to the dev_sz calculation. Fixes: dd22f551 ("block: Change direct_access calling convention") Cc: <stable@vger.kernel.org> # 4.0+ Signed-off-by:
Gerald Schaefer <gerald.schaefer@de.ibm.com> Signed-off-by:
-
- 31 Jan, 2017 11 commits
-
-
Stefan Haberland authored
Signed-off-by:
Stefan Haberland <sth@linux.vnet.ibm.com> Signed-off-by:
-
Stefan Haberland authored
If safe offline is called for a DASD alias device a null pointer is passed to fsync_bdev. So check for existence of the blockdevice before calling fsync_bdev. Should not be a real world problem since safe offline for an alias device does not make sense and fsync_bdev can deal with a NULL pointer which it gets after successful NULL pointer dereferencing on s390. Signed-off-by:
-
Stefan Haberland authored
Check if the device pointer is valid. Just a sanity check since we already are in the int handler of the device. Signed-off-by:
-
Stefan Haberland authored
Allow 0 as valid input for the path_threshold attribute to deactivate the IFCC/CCC error handling. Signed-off-by:
-
Martin Schwidefsky authored
The principles of operations specifies that the breaking event address is stored to the address 0x110 in the prefix page only for program checks. The last branch in user space is lost as soon as a branch in kernel space is executed after e.g. an svc. This makes it impossible to accurately maintain the breaking event address for a user space process. Simplify the code, just copy the current breaking event address from 0x110 to the task structure for program checks from user space. Signed-off-by: -
Jan Höppner authored
The function dasd_busid() still uses simple_strtoul() to convert a string to an integer value. This function is obsolete for quite some time already and should be replaced. The whole parameter parsing semantic still relies somewhat on the fact, that simple_strtoul() parses a string containing literals without complains and just returns the parsed integer value plus the residual string. kstrtoint(), however, would return -EINVAL in such a case. Since we want to get rid of simple_strtoul() and now have a nice dasd[] containing only single elements, we can clean up and simplify a few things. Replace simple_strtoul() with kstrtouint(), improve and simplify the overall parameter parsing by the following: - instead of residual strings return proper error codes - remove dasd_parse_next_element() and decide directly what sort of element is being parsed - if we parse a device or a range of devices, split that element into separate bits with a new function - remove warning about invalid ending as it doesn't apply anymore - annotate all parsing functions and data that can be freed after initialisation with __init and __initdata respectively - clean up bits and pieces while at it Reviewed-by:
Sebastian Ott <sebott@linux.vnet.ibm.com> Signed-off-by:
Jan Höppner <hoeppner@linux.vnet.ibm.com> Signed-off-by:
-
Jan Höppner authored
When the DASD driver is built into the kernel, the entire comma separated parameter list is stored as one single element in the dasd[] array, opposed to the module build where each element is stored separately in dasd[]. There is no point in doing so. Therefore, store each part of the list as single elements in dasd[] as well when built into the kernel. Also, create a define for the maximum of 256 parameters. Reviewed-by:
-
Harald Freudenberger authored
During tests the Kernel complained about inconsistend lock state: inconsistent {IN-SOFTIRQ-W} -> {SOFTIRQ-ON-W} usage. Now all the queue locks use spin_lock_bh/spin_unlock_bh. Signed-off-by:Harald Freudenberger <freude@linux.vnet.ibm.com> Signed-off-by:
-
Harald Freudenberger authored
The generate_entropy function used a sha256 for compacting together 256 bits of entropy into 32 bytes hash. However, it is questionable if a sha256 can really be used here, as potential collisions may reduce the max entropy fitting into a 32 byte hash value. So this batch introduces the use of sha512 instead and the required buffer adjustments for the calling functions. Further more the working buffer for the generate_entropy function has been widened from one page to two pages. So now 1024 stckf invocations are used to gather 256 bits of entropy. This has been done to be on the save side if the jitters of stckf values isn't as good as supposed. Signed-off-by:
-
Harald Freudenberger authored
In fips mode only xts keys with 128 bit or 125 bit are allowed. This fix extends the xts_aes_set_key function to check for these valid key lengths in fips mode. Signed-off-by:
-
Matthew Rosato authored
Triple-DES implementations will soon be required to check for uniqueness of keys with fips mode enabled. Add checks to ensure none of the 3 keys match. Signed-off-by:
Matthew Rosato <mjrosato@linux.vnet.ibm.com> Signed-off-by:
-
- 16 Jan, 2017 12 commits
-
-
Daniel Borkmann authored
After we already allocated the jit.prg_buf image via bpf_jit_binary_alloc() and filled it out with instructions, jit.prg_buf cannot be NULL anymore. Thus, remove the unnecessary check. Tested on s390x with test_bpf module. Signed-off-by:
Daniel Borkmann <daniel@iogearbox.net> Cc: Michael Holzheu <holzheu@linux.vnet.ibm.com> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Signed-off-by:
-
Sebastian Ott authored
This is a leftover from the 31 bit era to avoid 64 bit divisions. Get rid of it. Reported-by:
Cornelia Huck <cornelia.huck@de.ibm.com> Signed-off-by:
-
Heiko Carstens authored
Use the early sclp code to provide a boot console. This boot console is available if the kernel parameter "earlyprintk" has been specified, just like it works for other architectures that also provide an early boot console. This makes debugging of early problems much easier, since now we finally have working console output even before memory detection is running. The boot console will be automatically disabled as soon as another console will be registered. Reviewed-by:
-
Heiko Carstens authored
Make sure the _sclp_print_lm function stays within bounds of the early sccb, even if the passed string is very long. If the string is too long, the remaining characters will be dropped. Suggested-by:
-
Heiko Carstens authored
Make the early sclp interrupt handler more robust: - disable all interrupt sub classes except for the service signal subclass - extend ctlreg0 union so it is easily possible to set the service signal subclass mask bit without using a magic number - disable lowcore protection before writing to it - make sure that all write accesses are done before the original content of control register 0 is restored, which could enable lowcore protection Reviewed-by:
-
Heiko Carstens authored
The "topology=off" kernel parameter is supposed to prevent the kernel to use hardware topology information to generate scheduling domains etc. For an unknown reason I implemented this in a very odd way back then: instead of simply clearing the MACHINE_HAS_TOPOLOGY flag within the lowcore I added a second variable which indicated that topology information should not be used. This is more than suboptimal since it partially doesn't work. For the fake NUMA case topology information is still considered and scheduling domains will be created based on this. To fix this and to simplify the code get rid of the extra variable and implement the "topology=off" case like it is done for other features. Signed-off-by:
-
Heiko Carstens authored
Yet another trivial patch to reduce the noise that coccinelle generates. Signed-off-by:
-
Heiko Carstens authored
Remove a couple of unneeded semicolons. This is just to reduce the noise that the coccinelle static code checker generates. Signed-off-by:
-
Heiko Carstens authored
Merge the seven printks within topology_init_early to a single one. With an early boot console this avoids printing six lines each containing only a single character. Signed-off-by:
-
Heiko Carstens authored
Fix a long-standing but currently irrelevant bug: the memory detection code performs a tprot instruction on address zero to figure out if the first memory chunk is readable or writable. Due to low address protection the result is "read-only". If the memory detection code would actually care, it would have to ignore the first memory increment, but it adds the memory increment to writable memory anyway. If memblock debugging is enabled this leads to an extra rather surprising call which registers memory. To avoid this get rid of the first misleading tprot call and simply assume that the first memory increment is writable. Otherwise we wouldn't have reached the memory detection code anyway. Signed-off-by:
-
Heiko Carstens authored
The s390 specific memory detection code does not call memblock_add, which would generate debug output if memblock=debug is specified on the kernel command line. Instead it directly calls memblock_add_range, which doesn't generate any debug output. To have a chance to debug early memblock related bugs add an s390 specific memblock_dbg call and a (missing) memblock_dump_all call. Signed-off-by:
-
Heiko Carstens authored
reserve_initrd currently calls memblock_reserve even if the to be reserved size is zero. Even though the memblock core code can handle this correctly, it still yields confusing debug messages if memblock debugging is enabled. Therefore make sure to not call memblock_reserve with a size of zero. Signed-off-by:
-
