- 11 Jan, 2016 9 commits
-
-
Heiko Carstens authored
sclp_sync_wait() disables all external interrupt classes except for the service signal subclass. The static mask used for that however is wrong. It clears a couple of bits which shouldn't be cleared and on the other hand potentially does not clear bits which should be cleared. Fix this by using the same generic mask like we do it in our delay implementation. Signed-off-by:
Heiko Carstens <heiko.carstens@de.ibm.com> Reviewed-by:
Gerald Schaefer <gerald.schaefer@de.ibm.com> Signed-off-by:
Martin Schwidefsky <schwidefsky@de.ibm.com>
-
Ard Biesheuvel authored
The normalization pass in the sorting routine of the relative exception table serves two purposes: - it ensures that the address fields of the exception table entries are fully ordered, so that no ambiguities arise between entries with identical instruction offsets (i.e., when two instructions that are exactly 8 bytes apart each have an exception table entry associated with them) - it ensures that the offsets of both the instruction and the fixup fields of each entry are relative to their final location after sorting. Commit eb608fb3 ("s390/exceptions: switch to relative exception table entries") ported the relative exception table format from x86, but modified the sorting routine to only normalize the instruction offset field and not the fixup offset field. The result is that the fixup offset of each entry will be relative to the original location of the entry before sorting, likely leading to crashes when those entries are dereferenced. Fixes: eb608fb3 ("s390/exceptions: switch to relative exception table entries") Signed-off-by:
Ard Biesheuvel <ard.biesheuvel@linaro.org> Cc: <stable@vger.kernel.org> Signed-off-by:
-
Heiko Carstens authored
Signed-off-by:
-
Martin Schwidefsky authored
Add the CPU number to the per-cpu vdso data page and add the __kernel_getcpu function to the vdso object to retrieve the CPU number in user space. Suggested-by:
-
Michael S. Tsirkin authored
The initial s390 vdso code is heavily influenced by the powerpc version which does have a smp_wmb in vdso_init right before the vdso_ready=1 assignment. s390 has no need for that. Signed-off-by:
Michael S. Tsirkin <mst@redhat.com> Message-Id: <1452010645-25380-1-git-send-email-mst@redhat.com> Signed-off-by:
Christian Borntraeger <borntraeger@de.ibm.com> Signed-off-by:
-
Heiko Carstens authored
Finally get rid of the leading underscore. I tried this already two or three years ago, however Michael Holzheu objected since this would break the crash utility (again). However Michael integrated support for the new name into the crash utility back then, so it doesn't break if the name will be changed now. So finally get rid of the ever confusing leading underscore. Signed-off-by:
-
Heiko Carstens authored
The memory detection code historically had to use unsigned long long since the machine reported the true memory size (>4GB) even if the virtual machine was running in ESA/390 mode. Since the old code is gone use unsigned long everywhere and also get rid of an unused ADDR2G define. (this patch converts all long longs within sclp_info to longs) There are many more possible conversions, however that can be done if somebody touches the corresponding code. Since people started to convert unrelated long types to long longs because of the types within struct sclp_info convert this now. Signed-off-by:
-
Heiko Carstens authored
The long longs were introduced by me in order to have a working definition of the struct psw_bits also in 31 bit mode. Since that is gone also get rid of the long longs. Signed-off-by:
-
Heiko Carstens authored
Add missing multithreading fields of SYSIB 1.2.2 (Basic-Machine CPUs) to the output of /proc/sysinfo. Also use bitfields for SYSIB 2.2.2 to simplify the C code a bit. Signed-off-by:
-
- 30 Dec, 2015 6 commits
-
-
Heiko Carstens authored
Use CONFIG_TOPOLOGY which selects CONFIG_SCHED_* all over the place to reduce the random usage of the previous config options. Signed-off-by:
-
Heiko Carstens authored
s390 is always 64 bit. Signed-off-by:
-
Stefan Haberland authored
Enabling failfast should let request fail immediately if either an error occurred or the device gets disconnected. For disconnected devices new requests are not fetches from the block queue and therefore failfast is not triggered. Fix by letting the DASD driver fetch requests for disconnected devices with failfast active. Signed-off-by:
Stefan Haberland <stefan.haberland@de.ibm.com> Signed-off-by:
-
Pierre Morel authored
Return value from kzalloc is not tested and using a null pointer would lead to crash. Even if this should not happen at this moment, we may let the system decide if there is a better choice. Signed-off-by:
Pierre Morel <pmorel@linux.vnet.ibm.com> Signed-off-by:
-
Aya Mahfouz authored
Constifies hmcdrv_ftp_ops structures in s390's char driver since they are not modified after their initialization. Detected and found using Coccinelle. Suggested-by:
Julia Lawall <Julia.Lawall@lip6.fr> Signed-off-by:
Aya Mahfouz <mahfouz.saif.elyazal@gmail.com> Signed-off-by:
-
Julia Lawall authored
Add NULL test on call to kzalloc. The semantic match that finds this problem is as follows: (http://coccinelle.lip6.fr/) // <smpl> @@ expression x; identifier fld; @@ * x = kzalloc(...); ... when != x == NULL x->fld // </smpl> Signed-off-by:
-
- 18 Dec, 2015 20 commits
-
-
Peter Oberparleiter authored
Adding tracepoints to inline functions adds tracepoint invocation code for each instance where the function is inlined. The resulting increase in kernel code size can have negative impact: - More cache misses in instruction cache - Reduced amount of DMA-capable memory Therefore change all functions implementing I/O instructions from inline to normal functions. Bloat-o-meter summary after change (using performance_defconfig): add/remove: 24/2 grow/shrink: 4/39 up/down: 2205/-4858 (-2653) Signed-off-by:
Peter Oberparleiter <oberpar@linux.vnet.ibm.com> Acked-by:
Cornelia Huck <cornelia.huck@de.ibm.com> Signed-off-by:
-
Peter Oberparleiter authored
Add tracepoints to interrupt handler and core inline assemblies used by the s390 common I/O layer. These tracepoints can be used to monitor and validate hardware and hypervisor requests and responses. Signed-off-by:
Sebastian Ott <sebott@linux.vnet.ibm.com> Reviewed-by:
-
Peter Oberparleiter authored
Replace the current semi-arbitrary distribution of inline assemblies: - Inline assemblies used by CIO go into ioasm.h - Data definitions used by inline assemblies go into cio.h Beyond cleaning up the current structure this is also required for use of tracepoints in inline assemblies introduced by a follow-on patch. Signed-off-by:
-
Peter Oberparleiter authored
The numeric representation of the xsch instruction was incorrectly specified, resulting in reserved fields of the instruction opcode potentially being set to a non-zero value. While this doesn't currently cause any problem, a future architecture might make use of these fields so that the current specification could result in an exception or unwanted side-effects. Fix this by using the xsch instruction code for which support in binutils was added in 2003. Signed-off-by:
-
Peter Oberparleiter authored
There is no longer a need to maintain two versions of the same inline assembly - one with exception handling, and one without - so get rid of the duplicates and adjust names accordingly. This applies to stsch_err and msch_err which are now renamed to stsch and msch respectively, while the original msch function is removed. Signed-off-by:
-
Christian Borntraeger authored
Over time some machine flags got unused (e.g. MACHINE_FLAG_MVPG) or are available on all 64bit systems (MACHINE_FLAG_CSP, MACHINE_FLAG_IEEE) - let's remove them. Reorder the other ones to match the order of the MACHINE_HAS_* macros and renumber all bits to avoid holes. Also fix the comment about where the flags are detected. Signed-off-by:
-
Heiko Carstens authored
This is supposed to make debugging easier: if within a dump we can see that an external call or emergency signal IPI is pending but all cpus are idle, we have no idea for how long the interrupt is outstanding. Therefore save a timestamp into the per cpu pcpu array of the target cpu whenever such an IPI is sent. Signed-off-by:
-
Christian Borntraeger authored
max_mnest and rc are never used. Signed-off-by:
-
Christian Borntraeger authored
findseg_scode is assigned, but never used. Signed-off-by:
-
Christian Borntraeger authored
address is assigned but never used. Signed-off-by:
-
Christian Borntraeger authored
location is assigned but never used. Signed-off-by:
-
Heiko Carstens authored
head.S on s390 contains some sanity checks if the kernel will run on a machine or if the machine is too old, e.g. if the kernel contains instructions not available on the machine. If so, it will emit an error message to the console before it stops execution. Therefore head.S contains only instructions which are availanble with the earliest machine generation (z900). In order to make sure we don't accidently add instructions which are not available on z900, always compile with -march=z900. This makes sure compilation will fail if wrong instructions are used. Signed-off-by:
-
Heiko Carstens authored
It is already possible to remove CFLAGS with the CFLAGS_REMOVE option that was introduced with commit 656ee82c ("kbuild: create new CFLAGS_REMOVE_(basename).o option"). However it is not possible to remove AFLAGS for assembler files. So this patch just adds the AFLAGS_REMOVE option which works the same like CFLAGS_REMOVE. Signed-off-by:
Michal Marek <mmarek@suse.com> Signed-off-by:
-
Heiko Carstens authored
If configured for z13 assume the kernel makes use of the instructions that are part of the load-and-zero-rightmost-byte facility and load/store-on-condition facility 2. Signed-off-by:
-
Heiko Carstens authored
test_facility() can be optimized for bits which must be set anyway, due to the check in head.S. This removes a couple of superfluous runtime checks. Signed-off-by:
-
Heiko Carstens authored
The facility lists contain a lot of bits which are not necessary to run the kernel. Therefore remove them and keep only those bits which are required for the kernel. Signed-off-by:
-
Heiko Carstens authored
Change head.S to make use of the generated facility list. Signed-off-by:
-
Heiko Carstens authored
Modifying the architecture level set facility lists was always very error prone. Given the numbering of the facility bits within the Principles of Operation, where the most significant bit number is 0, it happened a lot of times that wrong bits were set or cleared. Therefore this patch adds a tool "gen_facilities" which generates include/generated/facilites.h. The definition of the bits to be set is contained within arch/s390/include/asm/facilities_src.h and can be easily extended to e.g. also generate such lists for the KVM module. The generated file looks like this: #define FACILITIES_ALS _AC(0xc1006450f0040000,UL) #define FACILITIES_ALS_DWORDS 1 The facility bits defined in this patch match 1:1 to the current masks that can be found in head.S. That is if the tool gets executed with -march=z990 then the generated masks will equal the masks in head.S for CONFIG_MARCH_Z990. Signed-off-by:
-
Heiko Carstens authored
head.s contains an stfle instruction which stores it result at the storage location that is assigned to the stfl instruction. This is currently no problem, since we only care about one double word. However if the number of double words in the ALS bitfield grows the current code is not very stable. E.g. before issuing the stfle command the memory to which it stores must be cleared, since the instruction may or may not clear memory contents where no bits are set. In order to simplify the code a bit always use the storage location that we reserved for the stfle result. Signed-off-by:
-
Heiko Carstens authored
Now that 31 bit support is gone, the assembler always knows about the stfl instruction. Therefore lets use a readable mnemonic. Also remove the not needed extable entry for the inline assembly and fix the output constraint. Signed-off-by:
-
- 27 Nov, 2015 5 commits
-
-
Martin Schwidefsky authored
It does not make sense to try to relinquish the time slice with diag 0x9c to a CPU in a state that does not allow to schedule the CPU. The scenario where this can happen is a CPU waiting in udelay/mdelay while holding a spin-lock. Add a CIF bit to tag a CPU in enabled wait and use it to detect that the yield of a CPU will not be successful and skip the diagnose call. Reviewed-by:
-
Jochen Schweflinghaus authored
Provide a user space interface and an enhancement to the sclp device driver which allows to send an 'Open for Business' event from the operating system to the Support Element. The 'Open for Business' event is used to signal the Support Element that the operating system (or an application running on top of it) is up and running. Signed-off-by:
Jochen Schweflinghaus <schwefel@de.ibm.com> Signed-off-by:
-
Heiko Carstens authored
is_32bit_task() used to be helpful when we still had CONFIG_32BIT. Since that is gone, it is nowadays identical to is_compat_task(). So remove it. Signed-off-by:
-
Michael Holzheu authored
Signed-off-by:
Michael Holzheu <holzheu@linux.vnet.ibm.com> Signed-off-by:
-
Sascha Silbe authored
When running under qemu with the default configuration (-nographic), there is only a VT220 SCLP console, no line-mode SCLP console. Add VT220 support to the early SCLP console so the user has a chance to see critical error messages during early boot. None of the existing users of _sclp_print_early() check the return code. Instead of trying to come up with return code semantics when printing to multiple consoles (any or all of which may fail), we just drop the return code entirely. Tested on z/VM (line mode console) and LPAR (VT220 and line mode console). Tested on qemu/KVM with VT220 console and / or line mode console. Signed-off-by:
Sascha Silbe <silbe@linux.vnet.ibm.com> Reviewed-by:
-
