- 12 Apr, 2017 2 commits
-
-
Martin Schwidefsky authored
The CAD instruction never worked quite as expected for the spinlock code. It has been disabled by default with git commit 61b0b016, if the "cad" kernel parameter is specified it is enabled for both user space and the spinlock code. Leave the option to enable the instruction for user space but remove it from the spinlock code. Signed-off-by:
Martin Schwidefsky <schwidefsky@de.ibm.com>
-
Martin Schwidefsky authored
Add a couple more __atomic_xxx function to atomic_ops.h and use them to replace the compare-and-swap inlines in the spinlock code. This changes the type of the lock value from unsigned int to int. Signed-off-by:
-
- 05 Apr, 2017 7 commits
-
-
Martin Schwidefsky authored
There are three different code levels in regard to the identification of guest samples. They differ in the way the LPP instruction is used. 1) Old kernels without the LPP instruction. The guest program parameter is always zero. 2) Newer kernels load the process pid into the program parameter with LPP. The guest program parameter is non-zero if the guest executes in a process != idle. 3) The latest kernels load ((1UL << 31) | pid) with LPP to make the value non-zero even for the idle task. The guest program parameter is non-zero if the guest is running. All kernels load the process pid to CR4 on context switch. The CPU sampling code uses the value in CR4 to decide between guest and host samples in case the guest program parameter is zero. The three cases: 1) CR4==pid, gpp==0 2) CR4==pid, gpp==pid 3) CR4==pid, gpp==((1UL << 31) | pid) The load-control instruction to load the pid into CR4 is expensive and the goal is to remove it. To distinguish the host CR4 from the guest pid for the idle process the maximum value 0xffff for the PASN is used. This adds a fourth case for a guest OS with an updated kernel: 4) CR4==0xffff, gpp=((1UL << 31) | pid) The host kernel will have CR4==0xffff and will use (gpp!=0 || CR4!==0xffff) to identify guest samples. This works nicely with all 4 cases, the only possible issue would be a guest with an old kernel (gpp==0) and a process pid of 0xffff. Well, don't do that.. Suggested-by:
Christian Borntraeger <borntraeger@de.ibm.com> Reviewed-by:
-
Sebastian Ott authored
Move a struct definition to get rid of a forward declaration. Signed-off-by:
Sebastian Ott <sebott@linux.vnet.ibm.com> Signed-off-by:
-
Sebastian Ott authored
Users complained that they are hitting the limit of 64 functions (some physical functions can spawn lots of virtual functions). Double the default limit. With the latest savings in static data usage this increases the image size by only 520 bytes. Signed-off-by:
-
Sebastian Ott authored
Commit c506fff3 ("s390/pci: resize iomap") reduced the iomap to NR_FUNCTIONS * PCI_BAR_COUNT elements. Since we only support functions with 64bit BARs we can cut that number in half. Signed-off-by:
-
Sebastian Ott authored
Address space identifiers are already defined in <asm/pci_insn.h>. Signed-off-by:
-
Sebastian Ott authored
barsize was never used. Get rid of it. Signed-off-by:
-
Martin Schwidefsky authored
The 32-bit lctl instruction is quite a bit slower than the 64-bit counter part lctlg. Use the faster instruction. Signed-off-by:
-
- 31 Mar, 2017 24 commits
-
-
Martin Schwidefsky authored
Merge branch 'vfio-ccw-for-martin' of git://git.kernel.org/pub/scm/linux/kernel/git/kvms390/vfio-ccw into features Pull vfio-ccw branch to add the basic channel I/O passthrough intrastructure based on vfio. The focus is on supporting dasd-eckd(cu_type/dev_type = 0x3990/0x3390) as the target device. Signed-off-by:
-
Dong Jia Shi authored
Add Cornelia Huck and myself as the vfio-ccw driver maintainers. Signed-off-by:
Dong Jia Shi <bjsdjshi@linux.vnet.ibm.com> Message-Id: <20170317031743.40128-17-bjsdjshi@linux.vnet.ibm.com> Signed-off-by:
Cornelia Huck <cornelia.huck@de.ibm.com>
-
Dong Jia Shi authored
Although Linux does not use format-0 channel command words (CCW0) these are a non-optional part of the platform spec, and for the sake of platform compliance, and possibly some non-Linux guests, we have to support CCW0. Making the kernel execute a format 0 channel program is too much hassle because we would need to allocate and use memory which can be addressed by 24 bit physical addresses (because of CCW0.cda). So we implement CCW0 support by translating the channel program into an equivalent CCW1 program instead. Based upon an orginal patch by Kai Yue Wang. Signed-off-by:
-
Dong Jia Shi authored
Add file Documentation/s390/vfio-ccw.txt that includes details of vfio-ccw. Acked-by:
Pierre Morel <pmorel@linux.vnet.ibm.com> Signed-off-by:
-
Dong Jia Shi authored
The current implementation doesn't check if the subchannel is in a proper device state when handling an event. Let's introduce a finite state machine to manage the state/event change. Signed-off-by:
-
Dong Jia Shi authored
Introduce a singlethreaded workqueue to handle the I/O interrupts. With the work added to this queue, we store the I/O results to the io_region of the subchannel, then signal the userspace program to handle the results. Signed-off-by:
-
Dong Jia Shi authored
Realize VFIO_DEVICE_GET_IRQ_INFO ioctl to retrieve VFIO_CCW_IO_IRQ information. Realize VFIO_DEVICE_SET_IRQS ioctl to set an eventfd fd for VFIO_CCW_IO_IRQ. Once a write operation to the ccw_io_region was performed, trigger a signal on this fd. Reviewed-by:
Alex Williamson <alex.williamson@redhat.com> Message-Id: <20170317031743.40128-12-bjsdjshi@linux.vnet.ibm.com> Signed-off-by:
-
Dong Jia Shi authored
Introduce VFIO_DEVICE_RESET ioctl for vfio-ccw to make it possible to hot-reset the device. We try to achieve a reset by first disabling the subchannel and then enabling it again: this should clear all state at the subchannel. Signed-off-by:
-
Dong Jia Shi authored
Introduce device information about vfio-ccw: VFIO_DEVICE_FLAGS_CCW. Realize VFIO_DEVICE_GET_REGION_INFO ioctl for vfio-ccw. Reviewed-by:
-
Dong Jia Shi authored
We implement the basic ccw command handling infrastructure here: 1. Translate the ccw commands. 2. Issue the translated ccw commands to the device. 3. Once we get the execution result, update the guest SCSW with it. Acked-by:
-
Dong Jia Shi authored
To provide user-space a set of interfaces to: 1. pass in a ccw program to perform an I/O operation. 2. read back I/O results of the completed I/O operations. We introduce an MMIO region for the vfio-ccw device here. This region is defined to content: 1. areas to store arguments that an ssch required. 2. areas to store the I/O results. Using pwrite/pread to the device on this region, a user-space program could write/read data to/from the vfio-ccw device. Reviewed-by:
-
Dong Jia Shi authored
To make vfio support subchannel devices, we need to leverage the mediated device framework to create a mediated device for the subchannel device. This registers the subchannel device to the mediated device framework during probe to enable mediated device creation. Reviewed-by:
-
Dong Jia Shi authored
Introduce ccwchain structure and helper functions that can be used to handle a channel program issued from a virtual machine. The following limitations apply: 1. Supports only prefetch enabled mode. 2. Supports idal(c64) ccw chaining. 3. Supports 4k idaw. 4. Supports ccw1. 5. Supports direct ccw chaining by translating them to idal ccws. CCW translation requires to leverage the vfio_(un)pin_pages interfaces to pin/unpin sets of mem pages frequently. Currently we have a lack of support to do this in an efficient way. So we introduce pfn_array data structure and helper functions to handle pin/unpin operations here. Signed-off-by:
-
Dong Jia Shi authored
To make vfio support subchannel devices, we need a css driver for the vfio subchannels. This patch adds a basic vfio-ccw subchannel driver for this purpose. To enable VFIO for vfio-ccw, enable S390_CCW_IOMMU config option and configure VFIO as required. Acked-by:
-
Dong Jia Shi authored
Define vfio-ccw device API strings. CCW vendor driver using mediated device framework should use this string for device_api attribute. Reviewed-by:
-
Dong Jia Shi authored
Export the common I/O interfaces those are needed by an I/O subchannel driver to actually talk to the subchannel. Reviewed-by:
-
Dong Jia Shi authored
For future code reuse purpose, this decouples the cio code with the ccw device specific parts from ccw_device_cancel_halt_clear, and makes a new common I/O interface named cio_cancel_halt_clear. Reviewed-by:
-
Hendrik Brueckner authored
The return code of hw_perf_event_update() is not evaluated by its callers. Hence, simplify the function by removing the return code. Reported-by:
Heiko Carstens <heiko.carstens@de.ibm.com> Signed-off-by:
Hendrik Brueckner <brueckner@linux.vnet.ibm.com> Signed-off-by:
-
Hendrik Brueckner authored
Using a register variable for r4 is not necessary. Let the compiler decide the register to be used. Reported-by:
-
Hendrik Brueckner authored
Make clear that the event definitions relate to the counter facility (cf) and not to the sampling facility (sf). Signed-off-by:
-
Hendrik Brueckner authored
Add the event names for the IBM z13/z13s specific CPU-MF counters. Also improve the merging of the generic and model specific events so that their sysfs attribute definitions completely reside in memory. Hence, flagging the generic event attribute definitions as initdata too. Signed-off-by:
-
Hendrik Brueckner authored
Complete the IBM z13 support and support counters from the MT-diagnostic counter set. Note that this counter set is available only if SMT is enabled. Signed-off-by:
-
Hendrik Brueckner authored
The validate_event() function just checked for reserved counters in particular CPU-MF counter sets. Because the number of counters in counter sets vary among different hardware models, remove the explicit check to tolerate new models. Reserved counters are not accounted and, thus, will return zero. Signed-off-by:
-
Hendrik Brueckner authored
Use the highest counter number that can be specified for the ecctr (extract CPU counter) instruction for perf. Signed-off-by:
-
- 30 Mar, 2017 2 commits
-
-
Heiko Carstens authored
Signed-off-by:
-
Heiko Carstens authored
Signed-off-by:
-
- 28 Mar, 2017 5 commits
-
-
Heiko Carstens authored
Signed-off-by:
-
Heiko Carstens authored
Deferred struct page initialization works on s390. However it makes only sense for the fake numa case, since the kthreads that initialize struct pages are started per node. Without fake numa there is just a single node and therefore no gain. However there is no reason to not enable this feature. Therefore select the config option and enable the feature in all config files. Signed-off-by:
-
Christian Borntraeger authored
gmap.c deals mostly with KVM-related memory management, so a lot of changes to this file will come via the KVM tree. Reflect this in MAINTAINERS. Please note that there are intricate ties to arch/s390/mm/pgtable.c. If changes are needed in both files, this will continue to be submitted via the s390 tree (or a topic branch if necessary). Signed-off-by:
-
Michael Holzheu authored
Since linux v3.14 with commit 38dfac84 ("vmcore: prevent PT_NOTE p_memsz overflow during header update") on s390 we get the following message in the kdump kernel: Warning: Exceeded p_memsz, dropping PT_NOTE entry n_namesz=0x6b6b6b6b, n_descsz=0x6b6b6b6b The reason for this is that we don't create a final zero note in the ELF header which the proc/vmcore code uses to find out the end of the notes section (see also kernel/kexec_core.c:final_note()). It still worked on s390 by chance because we (most of the time?) have the byte pattern 0x6b6b6b6b after the notes section which also makes the notes parsing code stop in update_note_header_size_elf64() because 0x6b6b6b6b is interpreded as note size: if ((real_sz + sz) > max_sz) { pr_warn("Warning: Exceeded p_memsz, dropping P ...); break; } So fix this and add the missing final note to the ELF header. We don't have to adjust the memory size for ELF header ("alloc_size") because the new ELF note still fits into the 0x1000 base memory. Cc: stable@vger.kernel.org # v4.4+ Signed-off-by:
Michael Holzheu <holzheu@linux.vnet.ibm.com> Signed-off-by:
-
Heiko Carstens authored
HAVE_ARCH_EARLY_PFN_TO_NID selects a not present Kconfig option. Therefore remove it. Given that the first call of early_pfn_to_nid() happens after numa_setup() finished to establish the memory to node mapping, there is no need to implement an architecture private version of __early_pfn_to_nid() like (only) ia64 does. Signed-off-by:
-
