- 31 Jul, 2020 2 commits
-
-
Catalin Marinas authored
Merge branches 'for-next/misc', 'for-next/vmcoreinfo', 'for-next/cpufeature', 'for-next/acpi', 'for-next/perf', 'for-next/timens', 'for-next/msi-iommu' and 'for-next/trivial' into for-next/core * for-next/misc: : Miscellaneous fixes and cleanups arm64: use IRQ_STACK_SIZE instead of THREAD_SIZE for irq stack arm64/mm: save memory access in check_and_switch_context() fast switch path recordmcount: only record relocation of type R_AARCH64_CALL26 on arm64. arm64: Reserve HWCAP2_MTE as (1 << 18) arm64/entry: deduplicate SW PAN entry/exit routines arm64: s/AMEVTYPE/AMEVTYPER arm64/hugetlb: Reserve CMA areas for gigantic pages on 16K and 64K configs arm64: stacktrace: Move export for save_stack_trace_tsk() smccc: Make constants available to assembly arm64/mm: Redefine CONT_{PTE, PMD}_SHIFT arm64/defconfig: Enable CONFIG_KEXEC_FILE arm64: Document sysctls for emulated deprecated instructions arm64/panic: Unify all three existing notifier blocks arm64/module: Optimize module load time by optimizing PLT counting * for-next/vmcoreinfo: : Export the virtual and physical address sizes in vmcoreinfo arm64/crash_core: Export TCR_EL1.T1SZ in vmcoreinfo crash_core, vmcoreinfo: Append 'MAX_PHYSMEM_BITS' to vmcoreinfo * for-next/cpufeature: : CPU feature handling cleanups arm64/cpufeature: Validate feature bits spacing in arm64_ftr_regs[] arm64/cpufeature: Replace all open bits shift encodings with macros arm64/cpufeature: Add remaining feature bits in ID_AA64MMFR2 register arm64/cpufeature: Add remaining feature bits in ID_AA64MMFR1 register arm64/cpufeature: Add remaining feature bits in ID_AA64MMFR0 register * for-next/acpi: : ACPI updates for arm64 arm64/acpi: disallow writeable AML opregion mapping for EFI code regions arm64/acpi: disallow AML memory opregions to access kernel memory * for-next/perf: : perf updates for arm64 arm64: perf: Expose some new events via sysfs tools headers UAPI: Update tools's copy of linux/perf_event.h arm64: perf: Add cap_user_time_short perf: Add perf_event_mmap_page::cap_user_time_short ABI arm64: perf: Only advertise cap_user_time for arch_timer arm64: perf: Implement correct cap_user_time time/sched_clock: Use raw_read_seqcount_latch() sched_clock: Expose struct clock_read_data arm64: perf: Correct the event index in sysfs perf/smmuv3: To simplify code for ioremap page in pmcg * for-next/timens: : Time namespace support for arm64 arm64: enable time namespace support arm64/vdso: Restrict splitting VVAR VMA arm64/vdso: Handle faults on timens page arm64/vdso: Add time namespace page arm64/vdso: Zap vvar pages when switching to a time namespace arm64/vdso: use the fault callback to map vvar pages * for-next/msi-iommu: : Make the MSI/IOMMU input/output ID translation PCI agnostic, augment the : MSI/IOMMU ACPI/OF ID mapping APIs to accept an input ID bus-specific parameter : and apply the resulting changes to the device ID space provided by the : Freescale FSL bus bus: fsl-mc: Add ACPI support for fsl-mc bus/fsl-mc: Refactor the MSI domain creation in the DPRC driver of/irq: Make of_msi_map_rid() PCI bus agnostic of/irq: make of_msi_map_get_device_domain() bus agnostic dt-bindings: arm: fsl: Add msi-map device-tree binding for fsl-mc bus of/device: Add input id to of_dma_configure() of/iommu: Make of_map_rid() PCI agnostic ACPI/IORT: Add an input ID to acpi_dma_configure() ACPI/IORT: Remove useless PCI bus walk ACPI/IORT: Make iort_msi_map_rid() PCI agnostic ACPI/IORT: Make iort_get_device_domain IRQ domain agnostic ACPI/IORT: Make iort_match_node_callback walk the ACPI namespace for NC * for-next/trivial: : Trivial fixes arm64: sigcontext.h: delete duplicated word arm64: ptrace.h: delete duplicated word arm64: pgtable-hwdef.h: delete duplicated words -
Maninder Singh authored
IRQ_STACK_SIZE can be made different from THREAD_SIZE, and as IRQ_STACK_SIZE is used while irq stack allocation, same define should be used while printing information of irq stack. Signed-off-by:
Maninder Singh <maninder1.s@samsung.com> Acked-by:
Mark Rutland <mark.rutland@arm.com> Link: https://lore.kernel.org/r/1596196190-14141-1-git-send-email-maninder1.s@samsung.comSigned-off-by:
Catalin Marinas <catalin.marinas@arm.com>
-
- 30 Jul, 2020 4 commits
-
-
Pingfan Liu authored
On arm64, smp_processor_id() reads a per-cpu `cpu_number` variable, using the per-cpu offset stored in the tpidr_el1 system register. In some cases we generate a per-cpu address with a sequence like: cpu_ptr = &per_cpu(ptr, smp_processor_id()); Which potentially incurs a cache miss for both `cpu_number` and the in-memory `__per_cpu_offset` array. This can be written more optimally as: cpu_ptr = this_cpu_ptr(ptr); Which only needs the offset from tpidr_el1, and does not need to load from memory. The following two test cases show a small performance improvement measured on a 46-cpus qualcomm machine with 5.8.0-rc4 kernel. Test 1: (about 0.3% improvement) #cat b.sh make clean && make all -j138 #perf stat --repeat 10 --null --sync sh b.sh - before this patch Performance counter stats for 'sh b.sh' (10 runs): 298.62 +- 1.86 seconds time elapsed ( +- 0.62% ) - after this patch Performance counter stats for 'sh b.sh' (10 runs): 297.734 +- 0.954 seconds time elapsed ( +- 0.32% ) Test 2: (about 1.69% improvement) 'perf stat -r 10 perf bench sched messaging' Then sum the total time of 'sched/messaging' by manual. - before this patch total 0.707 sec for 10 times - after this patch totol 0.695 sec for 10 times Signed-off-by:Pingfan Liu <kernelfans@gmail.com> Acked-by:
-
Randy Dunlap authored
Drop the repeated word "the". Signed-off-by:
Randy Dunlap <rdunlap@infradead.org> Cc: Will Deacon <will@kernel.org> Cc: linux-arm-kernel@lists.infradead.org Link: https://lore.kernel.org/r/20200726003207.20253-4-rdunlap@infradead.orgSigned-off-by:
-
Randy Dunlap authored
Drop the repeated word "the". Signed-off-by:
-
Randy Dunlap authored
Drop the repeated words "at" and "the". Signed-off-by:
-
- 28 Jul, 2020 12 commits
-
-
Makarand Pawagi authored
Add ACPI support in the fsl-mc driver. Driver parses MC DSDT table to extract memory and other resources. Interrupt (GIC ITS) information is extracted from the MADT table by drivers/irqchip/irq-gic-v3-its-fsl-mc-msi.c. IORT table is parsed to configure DMA. Signed-off-by:
Makarand Pawagi <makarand.pawagi@nxp.com> Signed-off-by:
Diana Craciun <diana.craciun@oss.nxp.com> Signed-off-by:
Laurentiu Tudor <laurentiu.tudor@nxp.com> Link: https://lore.kernel.org/r/20200619082013.13661-13-lorenzo.pieralisi@arm.comSigned-off-by:
-
Diana Craciun authored
The DPRC driver is not taking into account the msi-map property and assumes that the icid is the same as the stream ID. Although this assumption is correct, generalize the code to include a translation between icid and streamID. Furthermore do not just copy the MSI domain from parent (for child containers), but use the information provided by the msi-map property. If the msi-map property is missing from the device tree retain the old behaviour for backward compatibility ie the child DPRC objects inherit the MSI domain from the parent. Signed-off-by:
Marc Zyngier <maz@kernel.org> Link: https://lore.kernel.org/r/20200619082013.13661-12-lorenzo.pieralisi@arm.comSigned-off-by:
-
Lorenzo Pieralisi authored
There is nothing PCI bus specific in the of_msi_map_rid() implementation other than the requester ID tag for the input ID space. Rename requester ID to a more generic ID so that the translation code can be used by all busses that require input/output ID translations. No functional change intended. Signed-off-by:
Lorenzo Pieralisi <lorenzo.pieralisi@arm.com> Reviewed-by:
Rob Herring <robh@kernel.org> Cc: Bjorn Helgaas <bhelgaas@google.com> Cc: Rob Herring <robh+dt@kernel.org> Cc: Marc Zyngier <maz@kernel.org> Link: https://lore.kernel.org/r/20200619082013.13661-11-lorenzo.pieralisi@arm.comSigned-off-by:
-
Diana Craciun authored
of_msi_map_get_device_domain() is PCI specific but it need not be and can be easily changed to be bus agnostic in order to be used by other busses by adding an IRQ domain bus token as an input parameter. Signed-off-by:
-
Laurentiu Tudor authored
The existing bindings cannot be used to specify the relationship between fsl-mc devices and GIC ITSes. Add a generic binding for mapping fsl-mc devices to GIC ITSes, using msi-map property. In addition, deprecate msi-parent property which no longer makes sense now that we support translating the MSIs. Signed-off-by:
-
Lorenzo Pieralisi authored
Devices sitting on proprietary busses have a device ID space that is owned by the respective bus and related firmware bindings. In order to let the generic OF layer handle the input translations to an IOMMU id, for such busses the current of_dma_configure() interface should be extended in order to allow the bus layer to provide the device input id parameter - that is retrieved/assigned in bus specific code and firmware. Augment of_dma_configure() to add an optional input_id parameter, leaving current functionality unchanged. Signed-off-by:
-
Lorenzo Pieralisi authored
There is nothing PCI specific (other than the RID - requester ID) in the of_map_rid() implementation, so the same function can be reused for input/output IDs mapping for other busses just as well. Rename the RID instances/names to a generic "id" tag. No functionality change intended. Signed-off-by:
Joerg Roedel <jroedel@suse.de> Cc: Rob Herring <robh+dt@kernel.org> Cc: Joerg Roedel <joro@8bytes.org> Cc: Robin Murphy <robin.murphy@arm.com> Cc: Marc Zyngier <maz@kernel.org> Link: https://lore.kernel.org/r/20200619082013.13661-7-lorenzo.pieralisi@arm.comSigned-off-by:
-
Lorenzo Pieralisi authored
Some HW devices are created as child devices of proprietary busses, that have a bus specific policy defining how the child devices wires representing the devices ID are translated into IOMMU and IRQ controllers device IDs. Current IORT code provides translations for: - PCI devices, where the device ID is well identified at bus level as the requester ID (RID) - Platform devices that are endpoint devices where the device ID is retrieved from the ACPI object IORT mappings (Named components single mappings). A platform device is represented in IORT as a named component node For devices that are child devices of proprietary busses the IORT firmware represents the bus node as a named component node in IORT and it is up to that named component node to define in/out bus specific ID translations for the bus child devices that are allocated and created in a bus specific manner. In order to make IORT ID translations available for proprietary bus child devices, the current ACPI (and IORT) code must be augmented to provide an additional ID parameter to acpi_dma_configure() representing the child devices input ID. This ID is bus specific and it is retrieved in bus specific code. By adding an ID parameter to acpi_dma_configure(), the IORT code can map the child device ID to an IOMMU stream ID through the IORT named component representing the bus in/out ID mappings. Signed-off-by:
-
Lorenzo Pieralisi authored
The PCI bus domain number (used in the iort_match_node_callback() - pci_domain_nr() call) is cascaded through the PCI bus hierarchy at PCI bus enumeration time, therefore there is no need in iort_find_dev_node() to walk the PCI bus upwards to grab the root bus to be passed to iort_scan_node(), the device->bus PCI bus pointer will do. Remove this useless code. Signed-off-by:
-
Lorenzo Pieralisi authored
There is nothing PCI specific in iort_msi_map_rid(). Rename the function using a bus protocol agnostic name, iort_msi_map_id(), and convert current callers to it. Signed-off-by:
Bjorn Helgaas <bhelgaas@google.com> Cc: Will Deacon <will@kernel.org> Cc: Hanjun Guo <guohanjun@huawei.com> Cc: Bjorn Helgaas <bhelgaas@google.com> Cc: Sudeep Holla <sudeep.holla@arm.com> Cc: Robin Murphy <robin.murphy@arm.com> Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net> Link: https://lore.kernel.org/r/20200619082013.13661-4-lorenzo.pieralisi@arm.comSigned-off-by:
-
Lorenzo Pieralisi authored
iort_get_device_domain() is PCI specific but it need not be, since it can be used to retrieve IRQ domain nexus of any kind by adding an irq_domain_bus_token input to it. Make it PCI agnostic by also renaming the requestor ID input to a more generic ID name. Signed-off-by:
-
Lorenzo Pieralisi authored
When the iort_match_node_callback is invoked for a named component the match should be executed upon a device with an ACPI companion. For devices with no ACPI companion set-up the ACPI device tree must be walked in order to find the first parent node with a companion set and check the parent node against the named component entry to check whether there is a match and therefore an IORT node describing the in/out ID translation for the device has been found. Signed-off-by:
-
- 24 Jul, 2020 8 commits
-
-
Andrei Vagin authored
CONFIG_TIME_NS is dependes on GENERIC_VDSO_TIME_NS. Signed-off-by:
Andrei Vagin <avagin@gmail.com> Reviewed-by:
Vincenzo Frascino <vincenzo.frascino@arm.com> Reviewed-by:
Dmitry Safonov <dima@arista.com> Link: https://lore.kernel.org/r/20200624083321.144975-7-avagin@gmail.comSigned-off-by:
-
Andrei Vagin authored
Forbid splitting VVAR VMA resulting in a stricter ABI and reducing the amount of corner-cases to consider while working further on VDSO time namespace support. As the offset from timens to VVAR page is computed compile-time, the pages in VVAR should stay together and not being partically mremap()'ed. Signed-off-by:
-
Andrei Vagin authored
If a task belongs to a time namespace then the VVAR page which contains the system wide VDSO data is replaced with a namespace specific page which has the same layout as the VVAR page. Signed-off-by:
-
Andrei Vagin authored
Allocate the time namespace page among VVAR pages. Provide __arch_get_timens_vdso_data() helper for VDSO code to get the code-relative position of VVARs on that special page. If a task belongs to a time namespace then the VVAR page which contains the system wide VDSO data is replaced with a namespace specific page which has the same layout as the VVAR page. That page has vdso_data->seq set to 1 to enforce the slow path and vdso_data->clock_mode set to VCLOCK_TIMENS to enforce the time namespace handling path. The extra check in the case that vdso_data->seq is odd, e.g. a concurrent update of the VDSO data is in progress, is not really affecting regular tasks which are not part of a time namespace as the task is spin waiting for the update to finish and vdso_data->seq to become even again. If a time namespace task hits that code path, it invokes the corresponding time getter function which retrieves the real VVAR page, reads host time and then adds the offset for the requested clock which is stored in the special VVAR page. The time-namespace page isn't allocated on !CONFIG_TIME_NAMESPACE, but vma is the same size, which simplifies criu/vdso migration between different kernel configs. Signed-off-by:
-
Andrei Vagin authored
The order of vvar pages depends on whether a task belongs to the root time namespace or not. In the root time namespace, a task doesn't have a per-namespace page. In a non-root namespace, the VVAR page which contains the system-wide VDSO data is replaced with a namespace specific page that contains clock offsets. Whenever a task changes its namespace, the VVAR page tables are cleared and then they will be re-faulted with a corresponding layout. A task can switch its time namespace only if its ->mm isn't shared with another task. Signed-off-by:
Christian Brauner <christian.brauner@ubuntu.com> Link: https://lore.kernel.org/r/20200624083321.144975-3-avagin@gmail.comSigned-off-by:
-
Andrei Vagin authored
Currently the vdso has no awareness of time namespaces, which may apply distinct offsets to processes in different namespaces. To handle this within the vdso, we'll need to expose a per-namespace data page. As a preparatory step, this patch separates the vdso data page from the code pages, and has it faulted in via its own fault callback. Subsquent patches will extend this to support distinct pages per time namespace. The vvar vma has to be installed with the VM_PFNMAP flag to handle faults via its vma fault callback. Signed-off-by:
-
Gregory Herrero authored
Currently, if a section has a relocation to '_mcount' symbol, a new __mcount_loc entry will be added whatever the relocation type is. This is problematic when a relocation to '_mcount' is in the middle of a section and is not a call for ftrace use. Such relocation could be generated with below code for example: bool is_mcount(unsigned long addr) { return (target == (unsigned long) &_mcount); } With this snippet of code, ftrace will try to patch the mcount location generated by this code on module load and fail with: Call trace: ftrace_bug+0xa0/0x28c ftrace_process_locs+0x2f4/0x430 ftrace_module_init+0x30/0x38 load_module+0x14f0/0x1e78 __do_sys_finit_module+0x100/0x11c __arm64_sys_finit_module+0x28/0x34 el0_svc_common+0x88/0x194 el0_svc_handler+0x38/0x8c el0_svc+0x8/0xc ---[ end trace d828d06b36ad9d59 ]--- ftrace failed to modify [<ffffa2dbf3a3a41c>] 0xffffa2dbf3a3a41c actual: 66:a9:3c:90 Initializing ftrace call sites ftrace record flags: 2000000 (0) expected tramp: ffffa2dc6cf66724 So Limit the relocation type to R_AARCH64_CALL26 as in perl version of recordmcount. Fixes: af64d2aa ("ftrace: Add arm64 support to recordmcount") Signed-off-by:Gregory Herrero <gregory.herrero@oracle.com> Acked-by:
Steven Rostedt (VMware) <rostedt@goodmis.org> Link: https://lore.kernel.org/r/20200717143338.19302-1-gregory.herrero@oracle.comSigned-off-by:
-
Catalin Marinas authored
While MTE is not supported in the upstream kernel yet, add a comment that HWCAP2_MTE as (1 << 18) is reserved. Glibc makes use of it for the resolving (ifunc) of the MTE-safe string routines. Signed-off-by:
-
- 23 Jul, 2020 1 commit
-
-
Ard Biesheuvel authored
Factor the 12 copies of the SW PAN entry and exit code into callable subroutines, and use alternatives patching to either emit a 'bl' instruction to call them, or a NOP if h/w PAN is found to be available at runtime. Signed-off-by:
Ard Biesheuvel <ardb@kernel.org> Link: https://lore.kernel.org/r/20200721083315.4816-1-ardb@kernel.orgSigned-off-by:
-
- 22 Jul, 2020 1 commit
-
-
Vladimir Murzin authored
Activity Monitor Event Type Registers are named as AMEVTYPER{0,1}<n> Signed-off-by:Vladimir Murzin <vladimir.murzin@arm.com> Acked-by:
-
- 21 Jul, 2020 1 commit
-
-
Shaokun Zhang authored
Some new PMU events can been detected by PMCEID1_EL0, but it can't be listed, Let's expose these through sysfs. Signed-off-by:
Shaokun Zhang <zhangshaokun@hisilicon.com> Cc: Will Deacon <will@kernel.org> Cc: Mark Rutland <mark.rutland@arm.com> Link: https://lore.kernel.org/r/1595328573-12751-2-git-send-email-zhangshaokun@hisilicon.comSigned-off-by:
Will Deacon <will@kernel.org>
-
- 20 Jul, 2020 8 commits
-
-
Leo Yan authored
To get the changes in the commit: "perf: Add perf_event_mmap_page::cap_user_time_short ABI" This update is a prerequisite to add support for short clock counters related ABI extension. Signed-off-by:
Leo Yan <leo.yan@linaro.org> Link: https://lore.kernel.org/r/20200716051130.4359-8-leo.yan@linaro.orgSigned-off-by:
-
Peter Zijlstra authored
This completes the ARM64 cap_user_time support. Signed-off-by:
Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by:
-
Peter Zijlstra authored
In order to support short clock counters, provide an ABI extension. As a whole: u64 time, delta, cyc = read_cycle_counter(); + if (cap_user_time_short) + cyc = time_cycle + ((cyc - time_cycle) & time_mask); delta = mul_u64_u32_shr(cyc, time_mult, time_shift); if (cap_user_time_zero) time = time_zero + delta; delta += time_offset; Signed-off-by: -
Peter Zijlstra authored
When sched_clock is running on anything other than arch_timer, don't advertise cap_user_time*. Signed-off-by:
-
Peter Zijlstra authored
As reported by Leo; the existing implementation is broken when the clock and counter don't intersect at 0. Use the sched_clock's struct clock_read_data information to correctly implement cap_user_time and cap_user_time_zero. Note that the ARM64 counter is architecturally only guaranteed to be 56bit wide (implementations are allowed to be wider) and the existing perf ABI cannot deal with wrap-around. This implementation should also be faster than the old; seeing how we don't need to recompute mult and shift all the time. [leoyan: Use mul_u64_u32_shr() to convert cyc to ns to avoid overflow] Reported-by:
-
Ahmed S. Darwish authored
sched_clock uses seqcount_t latching to switch between two storage places protected by the sequence counter. This allows it to have interruptible, NMI-safe, seqcount_t write side critical sections. Since 7fc26327 ("seqlock: Introduce raw_read_seqcount_latch()"), raw_read_seqcount_latch() became the standardized way for seqcount_t latch read paths. Due to the dependent load, it also has one read memory barrier less than the currently used raw_read_seqcount() API. Use raw_read_seqcount_latch() for the seqcount_t latch read path. Signed-off-by:
Ahmed S. Darwish <a.darwish@linutronix.de> Signed-off-by:
-
Peter Zijlstra authored
In order to support perf_event_mmap_page::cap_time features, an architecture needs, aside from a userspace readable counter register, to expose the exact clock data so that userspace can convert the counter register into a correct timestamp. Provide struct clock_read_data and two (seqcount) helpers so that architectures (arm64 in specific) can expose the numbers to userspace. Signed-off-by:
-
Shaokun Zhang authored
When PMU event ID is equal or greater than 0x4000, it will be reduced by 0x4000 and it is not the raw number in the sysfs. Let's correct it and obtain the raw event ID. Before this patch: cat /sys/bus/event_source/devices/armv8_pmuv3_0/events/sample_feed event=0x001 After this patch: cat /sys/bus/event_source/devices/armv8_pmuv3_0/events/sample_feed event=0x4001 Signed-off-by:
-
- 15 Jul, 2020 1 commit
-
-
Anshuman Khandual authored
Currently 'hugetlb_cma=' command line argument does not create CMA area on ARM64_16K_PAGES and ARM64_64K_PAGES based platforms. Instead, it just ends up with the following warning message. Reason being, hugetlb_cma_reserve() never gets called for these huge page sizes. [ 64.255669] hugetlb_cma: the option isn't supported by current arch This enables CMA areas reservation on ARM64_16K_PAGES and ARM64_64K_PAGES configs by defining an unified arm64_hugetlb_cma_reseve() that is wrapped in CONFIG_CMA. Call site for arm64_hugetlb_cma_reserve() is also protected as <asm/hugetlb.h> is conditionally included and hence cannot contain stub for the inverse config i.e !(CONFIG_HUGETLB_PAGE && CONFIG_CMA). Signed-off-by:
Anshuman Khandual <anshuman.khandual@arm.com> Cc: Will Deacon <will@kernel.org> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Mike Kravetz <mike.kravetz@oracle.com> Cc: Barry Song <song.bao.hua@hisilicon.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: linux-arm-kernel@lists.infradead.org Cc: linux-kernel@vger.kernel.org Link: https://lore.kernel.org/r/1593578521-24672-1-git-send-email-anshuman.khandual@arm.comSigned-off-by:
-
- 14 Jul, 2020 2 commits
-
-
Mark Brown authored
Due to refactoring way back in bb53c820 ("arm64: stacktrace: avoid listing stacktrace functions in stacktrace") the EXPORT_SYMBOL_GPL() for save_stack_trace_tsk() is at the end of __save_stack_trace() rather than the function it exports. Move it to the expected location. Signed-off-by:
Mark Brown <broonie@kernel.org> Acked-by:
-
Ard Biesheuvel authored
Given that the contents of EFI runtime code and data regions are provided by the firmware, as well as the DSDT, it is not unimaginable that AML code exists today that accesses EFI runtime code regions using a SystemMemory OpRegion. There is nothing fundamentally wrong with that, but since we take great care to ensure that executable code is never mapped writeable and executable at the same time, we should not permit AML to create writable mapping. Signed-off-by:
-
