- 29 Aug, 2017 7 commits
-
-
Christian Borntraeger authored
Right now there is a potential hang situation for postcopy migrations, if the guest is enabling storage keys on the target system during the postcopy process. For storage key virtualization, we have to forbid the empty zero page as the storage key is a property of the physical page frame. As we enable storage key handling lazily we then drop all mappings for empty zero pages for lazy refaulting later on. This does not work with the postcopy migration, which relies on the empty zero page never triggering a fault again in the future. The reason is that postcopy migration will simply read a page on the target system if that page is a known zero page to fault in an empty zero page. At the same time postcopy remembers that this page was already transferred - so any future userfault on that page will NOT be retransmitted again to avoid races. If now the guest enters the storage key mode while in postcopy, we will break this assumption of postcopy. The solution is to disable the empty zero page for KVM guests early on and not during storage key enablement. With this change, the postcopy migration process is guaranteed to start after no zero pages are left. As guest pages are very likely not empty zero pages anyway the memory overhead is also pretty small. While at it this also adds proper page table locking to the zero page removal. Signed-off-by:
Christian Borntraeger <borntraeger@de.ibm.com> Acked-by:
Janosch Frank <frankja@linux.vnet.ibm.com> Cc: stable@vger.kernel.org Signed-off-by:
Martin Schwidefsky <schwidefsky@de.ibm.com>
-
Jan Höppner authored
The z/VM hypervisor provides virtual disks (VDISK) which are backed by main memory of the hypervisor. Those devices are seen as DASD FBA disks within the Linux guest. Whenever data is written to such a device, memory is allocated on-the-fly by z/VM accordingly. This memory, however, is not being freed if data on the device is deleted by the guest OS. In order to make memory usable after deletion again, add discard support to the FBA discipline. While at it, update comments regarding the DASD_FEATURE_* flags. Reviewed-by:
Stefan Haberland <sth@linux.vnet.ibm.com> Signed-off-by:
Jan Höppner <hoeppner@linux.vnet.ibm.com> Signed-off-by:
-
Bhumika Goyal authored
Make this const as it is only used in a copy operation. Signed-off-by:
Bhumika Goyal <bhumirks@gmail.com> Signed-off-by:
Harald Freudenberger <freude@linux.vnet.ibm.com> Signed-off-by:
-
Martin Schwidefsky authored
If the kernel is compiled for z10 or later machines the uaccess code inlines the mvcos instruction. The facility bit 27 which indicates the availability of MVCOS has to be set. The have_mvcos jump label will always be true. Make the generation of the have_mvcos jump label conditional on !CONFIG_HAVE_MARCH_Z10_FEATURES. Signed-off-by: -
Martin Schwidefsky authored
With git commit 3446c13b "s390/mm: four page table levels vs. fork" s390 dropped its architecture specific version of arch_dup_mmap. Now all functions defined by include/asm-generic/mm_hooks.h are identical to the s390 versions. Use the generic header. Signed-off-by:
-
Heiko Carstens authored
There really is no "general extension" facility available. Use the correct name "general instructions extension" facility instead. Signed-off-by:
Heiko Carstens <heiko.carstens@de.ibm.com> Signed-off-by:
-
Heiko Carstens authored
Get rid of the goto and "out" label within vmcp_response_free() which I added. This just makes the code harder to read than necessary. Signed-off-by:
-
- 23 Aug, 2017 5 commits
-
-
Dou Liyang authored
Commit a7be6e5a ("mm: drop useless local parameters of __register_one_node()") removes the last user of parent_node(). The parent_node() macro in S390 platform is unnecessary. Remove it for cleanup. Reported-by:
Michael Ellerman <mpe@ellerman.id.au> Signed-off-by:
Dou Liyang <douly.fnst@cn.fujitsu.com> Signed-off-by:
-
Jan Höppner authored
Unsigned long long and unsigned long were different in size for 31-bit. For 64-bit the size for both datatypes is 8 Bytes and since the support for 31-bit is long gone we can clean up a little and change everything to unsigned long. Change get_phys_clock() along the way to accept unsigned long as well so that the DASD code can be consistent. Acked-by:
-
Heiko Carstens authored
cpuhp_setup_state() returns a state number on CPUHP_AP_ONLINE_DYN if no error occurred. Therefore convert the return code to zero before using it as exit code for s390_smp_init(). Signed-off-by:
-
Heiko Carstens authored
Specifying the 'novx' kernel parameter always results in a warning: Malformed early option 'novx' The reason for this is that the novx early parameter handling function always returns a non-zero value which means that an error occurred. Fix this and return the correct zero value instead. Signed-off-by:
-
Stefan Haberland authored
Add average times to the DASD statistics interface. Signed-off-by:
-
- 09 Aug, 2017 10 commits
-
-
Sebastian Ott authored
Since commit caf7df12 ("block: remove the errors field from struct request") rq->errors can't be (mis)used by block device drivers to store the error condition for usage during async completion. Because of that I simply used async completion only for the non-error paths. This patch places the error within the private data of struct request and uses async completion for all paths again. Signed-off-by:
Sebastian Ott <sebott@linux.vnet.ibm.com> Signed-off-by:
-
Sebastian Ott authored
Some hypervisors allow to toggle uid checking at runtime. Log changes to uid checking in s390dbf. Signed-off-by:
-
Heiko Carstens authored
vmcp_ioctl() has many different return statements and duplicates a lot of mutex_unlock() calls. Simplify this so that only one return statement and one mutex_unlock() call is left. Signed-off-by:
-
Heiko Carstens authored
Return -ENOTTY for unknown ioctl commands instead of -ENOIOCTLCMD. This isn't that much of difference, since common code will translate -ENOIOCTLCMD to -ENOTTY anyway, but this way it seems to be more obvious what is happening (at least to me). Signed-off-by:
-
Heiko Carstens authored
Split the vmcp header file and move the device driver internal structure to the C file, and move the ioctl definitions to the uapi directory. Signed-off-by:
-
Heiko Carstens authored
If memory is fragmented it is unlikely that large order memory allocations succeed. This has been an issue with the vmcp device driver since a long time, since it requires large physical contiguous memory ares for large responses. To hopefully resolve this issue make use of the contiguous memory allocator (cma). This patch adds a vmcp specific vmcp cma area with a default size of 4MB. The size can be changed either via the VMCP_CMA_SIZE config option at compile time or with the "vmcp_cma" kernel parameter (e.g. "vmcp_cma=16m"). For any vmcp response buffers larger than 16k memory from the cma area will be allocated. If such an allocation fails, there is a fallback to the buddy allocator. Signed-off-by:
-
Heiko Carstens authored
According to the CP Programming Services manual Diagnose Code 8 "Virtual Console Function" can be used in all addressing modes. Also the input and output buffers do not have a limitation which specifies they need to be below the 2GB line. This is true at least since z/VM 5.4. Therefore remove the sam31/64 instructions and allow for simple GFP_KERNEL allocations. This makes it easier to allocate a 1MB page if the user requested such a large return buffer. Signed-off-by:
-
Heiko Carstens authored
The vmcp device driver should return -EFAULT if get_user() fails, due to an invalid user space address. In addition the buffer size value from user space is passed unchecked to get_order(). The return value of get_order(0) undefined. Therefore explicitly test for zero before calling get_order() and also return -EFAULT if get_user() fails. Signed-off-by:
-
Heiko Carstens authored
Memory blocks that contain areas for the contiguous memory allocator (cma) should not be allowed to go offline. Otherwise this would render cma completely useless. This might make sense on other architectures where memory might be taken offline due to hardware errors, but not on architectures which support memory hotplug for load balancing. Signed-off-by:
-
Heiko Carstens authored
Add all the missing little helper functions like virt_to_pfn(), phys_to_page(), etc. While we had a couple of these helper functions like e.g. page_to_phys() the other functions were missing, which is quite annoying if one is looking for exactly such a function. Therefore finally add all those little helper functions. Signed-off-by:
-
- 03 Aug, 2017 4 commits
-
-
Bhumika Goyal authored
Add const to bin_attribute structures as they are only passed to the functions device_{remove/create}_bin_file. The corresponding arguments are of type const, so declare the structures to be const. Cross compiled for s390 architecture. Signed-off-by: -
Bhumika Goyal authored
Declare bin_attribute structure as const as it is only passed as an argument to the function sysfs_create_bin_file. This argument is of type const, so declare the structure as const. Cross compiled for s390 architecture. Signed-off-by:
-
Heiko Carstens authored
And another header file for which we can use the generic variant, even though it doesn't look obvious at first glance. Signed-off-by:
-
Heiko Carstens authored
clang doesn't like s390 specific inline assembler constraints. These are present in our arch specific uapi/asm/swab.h which again is required by some ebpf test cases. For current compiler versions the generic swab.h already makes use of gcc's builtin functions. Therefore we can simply remove our own header file and use the generic one. This will generate worse code if used with compilers before gcc 4.8, which has no __builtin_bswap16(); or before gcc v4.4, which has no __builtin_bswap[32|64](). For these cases a C implementation fallback would be used which generates more code, but is still correct (170KB extra code for gcc 4.3 with performance_defconfig). However given that we need (and want) to get rid of the inline assemblies anyway in order to be able to use clang, the above is just a minor drawback if old gcc compilers are used. With current compilers there is close to zero difference, except for three btrfs bit functions which generate more out-of-line code. The generated code looks still correct and also uses the s390 specific byteswap instructions. Reported-and-tested-by:
Thomas Richter <tmricht@linux.vnet.ibm.com> Signed-off-by:
-
- 02 Aug, 2017 2 commits
-
-
Heiko Carstens authored
Both header files only include the corresponding uapi header file and therefore can be removed. Signed-off-by:
-
Heiko Carstens authored
The name of bit 36 of the machine check interruption code is "guarded storage registers validity". Add the missing "validity" part in order to be consistent with all other comments, which include this piece of information. Signed-off-by:
-
- 26 Jul, 2017 12 commits
-
-
Martin Schwidefsky authored
Merge tag 'vfio-ccw-20170724' of git://git.kernel.org/pub/scm/linux/kernel/git/kvms390/vfio-ccw into features Pull vfio-ccw fix from Cornelia Huck: "A bugfix in the ccw translation code."
-
Martin Schwidefsky authored
The z14 machine introduces new mode of the next-instruction-access-intent NIAI instruction. With NIAI-8 it is possible to pin a cache-line on a CPU for a small amount of time, NIAI-7 releases the cache-line again. Finally NIAI-4 can be used to prevent the CPU to speculatively access memory beyond the compare-and-swap instruction to get the lock. Use these instruction in the spinlock code. Signed-off-by: -
Arvind Yadav authored
attribute_group are not supposed to change at runtime. All functions working with attribute_group provided by <linux/sysfs.h> work with const attribute_group. So mark the non-const structs as const. Signed-off-by:
Arvind Yadav <arvind.yadav.cs@gmail.com> Signed-off-by:
-
Arvind Yadav authored
attribute_group are not supposed to change at runtime. All functions working with attribute_group provided by <linux/sysfs.h> work with const attribute_group. So mark the non-const structs as const. File size before: text data bss dec hex filename 11511 656 16 12183 2f97 drivers/s390/char/tape_core.o File size After adding 'const': text data bss dec hex filename 11575 592 16 12183 2f97 drivers/s390/char/tape_core.o Signed-off-by:
-
Arvind Yadav authored
attribute_group are not supposed to change at runtime. All functions working with attribute_group provided by <linux/sysfs.h> work with const attribute_group. So mark the non-const structs as const. File size before: text data bss dec hex filename 8069 816 16 8901 22c5 drivers/s390/char/raw3270.o File size After adding 'const': text data bss dec hex filename 8133 752 16 8901 22c5 drivers/s390/char/raw3270.o Signed-off-by:
-
Arvind Yadav authored
attribute_group are not supposed to change at runtime. All functions working with attribute_group provided by <linux/sysfs.h> work with const attribute_group. So mark the non-const structs as const. File size before: text data bss dec hex filename 6763 1216 0 7979 1f2b drivers/s390/net/qeth_l3_sys.o File size After adding 'const': text data bss dec hex filename 7019 960 0 7971 1f2b drivers/s390/net/qeth_l3_sys.o Signed-off-by:
-
Arvind Yadav authored
attribute_group are not supposed to change at runtime. All functions working with attribute_group provided by <linux/sysfs.h> work with const attribute_group. So mark the non-const structs as const. Signed-off-by:
-
Arvind Yadav authored
attribute_group are not supposed to change at runtime. All functions working with attribute_group provided by <linux/sysfs.h> work with const attribute_group. So mark the non-const structs as const. Signed-off-by:
-
Arvind Yadav authored
attribute_group are not supposed to change at runtime. All functions working with attribute_group provided by <linux/sysfs.h> work with const attribute_group. So mark the non-const structs as const. File size before: text data bss dec hex filename 1019 160 0 1179 49b drivers/s390/crypto/zcrypt_card.o File size After adding 'const': text data bss dec hex filename 1083 96 0 1179 49b drivers/s390/crypto/zcrypt_card.o Signed-off-by:
-
Arvind Yadav authored
attribute_group are not supposed to change at runtime. All functions working with attribute_group provided by <linux/sysfs.h> work with const attribute_group. So mark the non-const structs as const. File size before: text data bss dec hex filename 1361 96 0 1457 5b1 s390/crypto/zcrypt_queue.o File size After adding 'const': text data bss dec hex filename 1425 32 0 1457 5b1 s390/crypto/zcrypt_queue.o Signed-off-by:
-
Martin Schwidefsky authored
Add detection for machine type 0x3906 and set the ELF platform name to z14. Add the miscellaneous-instruction-extension 2 facility to the list of facilities for z14. And allow to generate code that only runs on a z14 machine. Signed-off-by: -
Martin Schwidefsky authored
The TOD epoch extension adds 8 epoch bits to the TOD clock to provide a continuous clock after 2042/09/17. The store-clock-extended (STCKE) instruction will store the epoch index in the first byte of the 16 bytes stored by the instruction. The read_boot_clock64 and the read_presistent_clock64 functions need to take the additional bits into account to give the correct result after 2042/09/17. The clock-comparator register will stay 64 bit wide. The comparison of the clock-comparator with the TOD clock is limited to bytes 1 to 8 of the extended TOD format. To deal with the overflow problem due to an epoch change the clock-comparator sign control in CR0 can be used to switch the comparison of the 64-bit TOD clock with the clock-comparator to a signed comparison. The decision between the signed vs. unsigned clock-comparator comparisons is done at boot time. Only if the TOD clock is in the second half of a 142 year epoch the signed comparison is used. This solves the epoch overflow issue as long as the machine is booted at least once in an epoch. Signed-off-by:
-
