linux/drivers/xen/xen-acpi-processor.c
<<
>>
Prefs
   1/*
   2 * Copyright 2012 by Oracle Inc
   3 * Author: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
   4 *
   5 * This code borrows ideas from https://lkml.org/lkml/2011/11/30/249
   6 * so many thanks go to Kevin Tian <kevin.tian@intel.com>
   7 * and Yu Ke <ke.yu@intel.com>.
   8 *
   9 * This program is free software; you can redistribute it and/or modify it
  10 * under the terms and conditions of the GNU General Public License,
  11 * version 2, as published by the Free Software Foundation.
  12 *
  13 * This program is distributed in the hope it will be useful, but WITHOUT
  14 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
  15 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
  16 * more details.
  17 *
  18 */
  19
  20#include <linux/cpumask.h>
  21#include <linux/cpufreq.h>
  22#include <linux/freezer.h>
  23#include <linux/kernel.h>
  24#include <linux/kthread.h>
  25#include <linux/init.h>
  26#include <linux/module.h>
  27#include <linux/types.h>
  28#include <acpi/acpi_bus.h>
  29#include <acpi/acpi_drivers.h>
  30#include <acpi/processor.h>
  31
  32#include <xen/xen.h>
  33#include <xen/interface/platform.h>
  34#include <asm/xen/hypercall.h>
  35
  36#define DRV_NAME "xen-acpi-processor: "
  37
  38static int no_hypercall;
  39MODULE_PARM_DESC(off, "Inhibit the hypercall.");
  40module_param_named(off, no_hypercall, int, 0400);
  41
  42/*
  43 * Note: Do not convert the acpi_id* below to cpumask_var_t or use cpumask_bit
  44 * - as those shrink to nr_cpu_bits (which is dependent on possible_cpu), which
  45 * can be less than what we want to put in. Instead use the 'nr_acpi_bits'
  46 * which is dynamically computed based on the MADT or x2APIC table.
  47 */
  48static unsigned int nr_acpi_bits;
  49/* Mutex to protect the acpi_ids_done - for CPU hotplug use. */
  50static DEFINE_MUTEX(acpi_ids_mutex);
  51/* Which ACPI ID we have processed from 'struct acpi_processor'. */
  52static unsigned long *acpi_ids_done;
  53/* Which ACPI ID exist in the SSDT/DSDT processor definitions. */
  54static unsigned long __initdata *acpi_id_present;
  55/* And if there is an _CST definition (or a PBLK) for the ACPI IDs */
  56static unsigned long __initdata *acpi_id_cst_present;
  57
  58static int push_cxx_to_hypervisor(struct acpi_processor *_pr)
  59{
  60        struct xen_platform_op op = {
  61                .cmd                    = XENPF_set_processor_pminfo,
  62                .interface_version      = XENPF_INTERFACE_VERSION,
  63                .u.set_pminfo.id        = _pr->acpi_id,
  64                .u.set_pminfo.type      = XEN_PM_CX,
  65        };
  66        struct xen_processor_cx *dst_cx, *dst_cx_states = NULL;
  67        struct acpi_processor_cx *cx;
  68        unsigned int i, ok;
  69        int ret = 0;
  70
  71        dst_cx_states = kcalloc(_pr->power.count,
  72                                sizeof(struct xen_processor_cx), GFP_KERNEL);
  73        if (!dst_cx_states)
  74                return -ENOMEM;
  75
  76        for (ok = 0, i = 1; i <= _pr->power.count; i++) {
  77                cx = &_pr->power.states[i];
  78                if (!cx->valid)
  79                        continue;
  80
  81                dst_cx = &(dst_cx_states[ok++]);
  82
  83                dst_cx->reg.space_id = ACPI_ADR_SPACE_SYSTEM_IO;
  84                if (cx->entry_method == ACPI_CSTATE_SYSTEMIO) {
  85                        dst_cx->reg.bit_width = 8;
  86                        dst_cx->reg.bit_offset = 0;
  87                        dst_cx->reg.access_size = 1;
  88                } else {
  89                        dst_cx->reg.space_id = ACPI_ADR_SPACE_FIXED_HARDWARE;
  90                        if (cx->entry_method == ACPI_CSTATE_FFH) {
  91                                /* NATIVE_CSTATE_BEYOND_HALT */
  92                                dst_cx->reg.bit_offset = 2;
  93                                dst_cx->reg.bit_width = 1; /* VENDOR_INTEL */
  94                        }
  95                        dst_cx->reg.access_size = 0;
  96                }
  97                dst_cx->reg.address = cx->address;
  98
  99                dst_cx->type = cx->type;
 100                dst_cx->latency = cx->latency;
 101
 102                dst_cx->dpcnt = 0;
 103                set_xen_guest_handle(dst_cx->dp, NULL);
 104        }
 105        if (!ok) {
 106                pr_debug(DRV_NAME "No _Cx for ACPI CPU %u\n", _pr->acpi_id);
 107                kfree(dst_cx_states);
 108                return -EINVAL;
 109        }
 110        op.u.set_pminfo.power.count = ok;
 111        op.u.set_pminfo.power.flags.bm_control = _pr->flags.bm_control;
 112        op.u.set_pminfo.power.flags.bm_check = _pr->flags.bm_check;
 113        op.u.set_pminfo.power.flags.has_cst = _pr->flags.has_cst;
 114        op.u.set_pminfo.power.flags.power_setup_done =
 115                _pr->flags.power_setup_done;
 116
 117        set_xen_guest_handle(op.u.set_pminfo.power.states, dst_cx_states);
 118
 119        if (!no_hypercall)
 120                ret = HYPERVISOR_dom0_op(&op);
 121
 122        if (!ret) {
 123                pr_debug("ACPI CPU%u - C-states uploaded.\n", _pr->acpi_id);
 124                for (i = 1; i <= _pr->power.count; i++) {
 125                        cx = &_pr->power.states[i];
 126                        if (!cx->valid)
 127                                continue;
 128                        pr_debug("     C%d: %s %d uS\n",
 129                                 cx->type, cx->desc, (u32)cx->latency);
 130                }
 131        } else if (ret != -EINVAL)
 132                /* EINVAL means the ACPI ID is incorrect - meaning the ACPI
 133                 * table is referencing a non-existing CPU - which can happen
 134                 * with broken ACPI tables. */
 135                pr_err(DRV_NAME "(CX): Hypervisor error (%d) for ACPI CPU%u\n",
 136                       ret, _pr->acpi_id);
 137
 138        kfree(dst_cx_states);
 139
 140        return ret;
 141}
 142static struct xen_processor_px *
 143xen_copy_pss_data(struct acpi_processor *_pr,
 144                  struct xen_processor_performance *dst_perf)
 145{
 146        struct xen_processor_px *dst_states = NULL;
 147        unsigned int i;
 148
 149        BUILD_BUG_ON(sizeof(struct xen_processor_px) !=
 150                     sizeof(struct acpi_processor_px));
 151
 152        dst_states = kcalloc(_pr->performance->state_count,
 153                             sizeof(struct xen_processor_px), GFP_KERNEL);
 154        if (!dst_states)
 155                return ERR_PTR(-ENOMEM);
 156
 157        dst_perf->state_count = _pr->performance->state_count;
 158        for (i = 0; i < _pr->performance->state_count; i++) {
 159                /* Fortunatly for us, they are both the same size */
 160                memcpy(&(dst_states[i]), &(_pr->performance->states[i]),
 161                       sizeof(struct acpi_processor_px));
 162        }
 163        return dst_states;
 164}
 165static int xen_copy_psd_data(struct acpi_processor *_pr,
 166                             struct xen_processor_performance *dst)
 167{
 168        struct acpi_psd_package *pdomain;
 169
 170        BUILD_BUG_ON(sizeof(struct xen_psd_package) !=
 171                     sizeof(struct acpi_psd_package));
 172
 173        /* This information is enumerated only if acpi_processor_preregister_performance
 174         * has been called.
 175         */
 176        dst->shared_type = _pr->performance->shared_type;
 177
 178        pdomain = &(_pr->performance->domain_info);
 179
 180        /* 'acpi_processor_preregister_performance' does not parse if the
 181         * num_processors <= 1, but Xen still requires it. Do it manually here.
 182         */
 183        if (pdomain->num_processors <= 1) {
 184                if (pdomain->coord_type == DOMAIN_COORD_TYPE_SW_ALL)
 185                        dst->shared_type = CPUFREQ_SHARED_TYPE_ALL;
 186                else if (pdomain->coord_type == DOMAIN_COORD_TYPE_HW_ALL)
 187                        dst->shared_type = CPUFREQ_SHARED_TYPE_HW;
 188                else if (pdomain->coord_type == DOMAIN_COORD_TYPE_SW_ANY)
 189                        dst->shared_type = CPUFREQ_SHARED_TYPE_ANY;
 190
 191        }
 192        memcpy(&(dst->domain_info), pdomain, sizeof(struct acpi_psd_package));
 193        return 0;
 194}
 195static int xen_copy_pct_data(struct acpi_pct_register *pct,
 196                             struct xen_pct_register *dst_pct)
 197{
 198        /* It would be nice if you could just do 'memcpy(pct, dst_pct') but
 199         * sadly the Xen structure did not have the proper padding so the
 200         * descriptor field takes two (dst_pct) bytes instead of one (pct).
 201         */
 202        dst_pct->descriptor = pct->descriptor;
 203        dst_pct->length = pct->length;
 204        dst_pct->space_id = pct->space_id;
 205        dst_pct->bit_width = pct->bit_width;
 206        dst_pct->bit_offset = pct->bit_offset;
 207        dst_pct->reserved = pct->reserved;
 208        dst_pct->address = pct->address;
 209        return 0;
 210}
 211static int push_pxx_to_hypervisor(struct acpi_processor *_pr)
 212{
 213        int ret = 0;
 214        struct xen_platform_op op = {
 215                .cmd                    = XENPF_set_processor_pminfo,
 216                .interface_version      = XENPF_INTERFACE_VERSION,
 217                .u.set_pminfo.id        = _pr->acpi_id,
 218                .u.set_pminfo.type      = XEN_PM_PX,
 219        };
 220        struct xen_processor_performance *dst_perf;
 221        struct xen_processor_px *dst_states = NULL;
 222
 223        dst_perf = &op.u.set_pminfo.perf;
 224
 225        dst_perf->platform_limit = _pr->performance_platform_limit;
 226        dst_perf->flags |= XEN_PX_PPC;
 227        xen_copy_pct_data(&(_pr->performance->control_register),
 228                          &dst_perf->control_register);
 229        xen_copy_pct_data(&(_pr->performance->status_register),
 230                          &dst_perf->status_register);
 231        dst_perf->flags |= XEN_PX_PCT;
 232        dst_states = xen_copy_pss_data(_pr, dst_perf);
 233        if (!IS_ERR_OR_NULL(dst_states)) {
 234                set_xen_guest_handle(dst_perf->states, dst_states);
 235                dst_perf->flags |= XEN_PX_PSS;
 236        }
 237        if (!xen_copy_psd_data(_pr, dst_perf))
 238                dst_perf->flags |= XEN_PX_PSD;
 239
 240        if (dst_perf->flags != (XEN_PX_PSD | XEN_PX_PSS | XEN_PX_PCT | XEN_PX_PPC)) {
 241                pr_warn(DRV_NAME "ACPI CPU%u missing some P-state data (%x), skipping.\n",
 242                        _pr->acpi_id, dst_perf->flags);
 243                ret = -ENODEV;
 244                goto err_free;
 245        }
 246
 247        if (!no_hypercall)
 248                ret = HYPERVISOR_dom0_op(&op);
 249
 250        if (!ret) {
 251                struct acpi_processor_performance *perf;
 252                unsigned int i;
 253
 254                perf = _pr->performance;
 255                pr_debug("ACPI CPU%u - P-states uploaded.\n", _pr->acpi_id);
 256                for (i = 0; i < perf->state_count; i++) {
 257                        pr_debug("     %cP%d: %d MHz, %d mW, %d uS\n",
 258                        (i == perf->state ? '*' : ' '), i,
 259                        (u32) perf->states[i].core_frequency,
 260                        (u32) perf->states[i].power,
 261                        (u32) perf->states[i].transition_latency);
 262                }
 263        } else if (ret != -EINVAL)
 264                /* EINVAL means the ACPI ID is incorrect - meaning the ACPI
 265                 * table is referencing a non-existing CPU - which can happen
 266                 * with broken ACPI tables. */
 267                pr_warn(DRV_NAME "(_PXX): Hypervisor error (%d) for ACPI CPU%u\n",
 268                       ret, _pr->acpi_id);
 269err_free:
 270        if (!IS_ERR_OR_NULL(dst_states))
 271                kfree(dst_states);
 272
 273        return ret;
 274}
 275static int upload_pm_data(struct acpi_processor *_pr)
 276{
 277        int err = 0;
 278
 279        mutex_lock(&acpi_ids_mutex);
 280        if (__test_and_set_bit(_pr->acpi_id, acpi_ids_done)) {
 281                mutex_unlock(&acpi_ids_mutex);
 282                return -EBUSY;
 283        }
 284        if (_pr->flags.power)
 285                err = push_cxx_to_hypervisor(_pr);
 286
 287        if (_pr->performance && _pr->performance->states)
 288                err |= push_pxx_to_hypervisor(_pr);
 289
 290        mutex_unlock(&acpi_ids_mutex);
 291        return err;
 292}
 293static unsigned int __init get_max_acpi_id(void)
 294{
 295        struct xenpf_pcpuinfo *info;
 296        struct xen_platform_op op = {
 297                .cmd = XENPF_get_cpuinfo,
 298                .interface_version = XENPF_INTERFACE_VERSION,
 299        };
 300        int ret = 0;
 301        unsigned int i, last_cpu, max_acpi_id = 0;
 302
 303        info = &op.u.pcpu_info;
 304        info->xen_cpuid = 0;
 305
 306        ret = HYPERVISOR_dom0_op(&op);
 307        if (ret)
 308                return NR_CPUS;
 309
 310        /* The max_present is the same irregardless of the xen_cpuid */
 311        last_cpu = op.u.pcpu_info.max_present;
 312        for (i = 0; i <= last_cpu; i++) {
 313                info->xen_cpuid = i;
 314                ret = HYPERVISOR_dom0_op(&op);
 315                if (ret)
 316                        continue;
 317                max_acpi_id = max(info->acpi_id, max_acpi_id);
 318        }
 319        max_acpi_id *= 2; /* Slack for CPU hotplug support. */
 320        pr_debug(DRV_NAME "Max ACPI ID: %u\n", max_acpi_id);
 321        return max_acpi_id;
 322}
 323/*
 324 * The read_acpi_id and check_acpi_ids are there to support the Xen
 325 * oddity of virtual CPUs != physical CPUs in the initial domain.
 326 * The user can supply 'xen_max_vcpus=X' on the Xen hypervisor line
 327 * which will band the amount of CPUs the initial domain can see.
 328 * In general that is OK, except it plays havoc with any of the
 329 * for_each_[present|online]_cpu macros which are banded to the virtual
 330 * CPU amount.
 331 */
 332static acpi_status __init
 333read_acpi_id(acpi_handle handle, u32 lvl, void *context, void **rv)
 334{
 335        u32 acpi_id;
 336        acpi_status status;
 337        acpi_object_type acpi_type;
 338        unsigned long long tmp;
 339        union acpi_object object = { 0 };
 340        struct acpi_buffer buffer = { sizeof(union acpi_object), &object };
 341        acpi_io_address pblk = 0;
 342
 343        status = acpi_get_type(handle, &acpi_type);
 344        if (ACPI_FAILURE(status))
 345                return AE_OK;
 346
 347        switch (acpi_type) {
 348        case ACPI_TYPE_PROCESSOR:
 349                status = acpi_evaluate_object(handle, NULL, NULL, &buffer);
 350                if (ACPI_FAILURE(status))
 351                        return AE_OK;
 352                acpi_id = object.processor.proc_id;
 353                pblk = object.processor.pblk_address;
 354                break;
 355        case ACPI_TYPE_DEVICE:
 356                status = acpi_evaluate_integer(handle, "_UID", NULL, &tmp);
 357                if (ACPI_FAILURE(status))
 358                        return AE_OK;
 359                acpi_id = tmp;
 360                break;
 361        default:
 362                return AE_OK;
 363        }
 364        /* There are more ACPI Processor objects than in x2APIC or MADT.
 365         * This can happen with incorrect ACPI SSDT declerations. */
 366        if (acpi_id > nr_acpi_bits) {
 367                pr_debug(DRV_NAME "We only have %u, trying to set %u\n",
 368                         nr_acpi_bits, acpi_id);
 369                return AE_OK;
 370        }
 371        /* OK, There is a ACPI Processor object */
 372        __set_bit(acpi_id, acpi_id_present);
 373
 374        pr_debug(DRV_NAME "ACPI CPU%u w/ PBLK:0x%lx\n", acpi_id,
 375                 (unsigned long)pblk);
 376
 377        status = acpi_evaluate_object(handle, "_CST", NULL, &buffer);
 378        if (ACPI_FAILURE(status)) {
 379                if (!pblk)
 380                        return AE_OK;
 381        }
 382        /* .. and it has a C-state */
 383        __set_bit(acpi_id, acpi_id_cst_present);
 384
 385        return AE_OK;
 386}
 387static int __init check_acpi_ids(struct acpi_processor *pr_backup)
 388{
 389
 390        if (!pr_backup)
 391                return -ENODEV;
 392
 393        /* All online CPUs have been processed at this stage. Now verify
 394         * whether in fact "online CPUs" == physical CPUs.
 395         */
 396        acpi_id_present = kcalloc(BITS_TO_LONGS(nr_acpi_bits), sizeof(unsigned long), GFP_KERNEL);
 397        if (!acpi_id_present)
 398                return -ENOMEM;
 399
 400        acpi_id_cst_present = kcalloc(BITS_TO_LONGS(nr_acpi_bits), sizeof(unsigned long), GFP_KERNEL);
 401        if (!acpi_id_cst_present) {
 402                kfree(acpi_id_present);
 403                return -ENOMEM;
 404        }
 405
 406        acpi_walk_namespace(ACPI_TYPE_PROCESSOR, ACPI_ROOT_OBJECT,
 407                            ACPI_UINT32_MAX,
 408                            read_acpi_id, NULL, NULL, NULL);
 409        acpi_get_devices("ACPI0007", read_acpi_id, NULL, NULL);
 410
 411        if (!bitmap_equal(acpi_id_present, acpi_ids_done, nr_acpi_bits)) {
 412                unsigned int i;
 413                for_each_set_bit(i, acpi_id_present, nr_acpi_bits) {
 414                        pr_backup->acpi_id = i;
 415                        /* Mask out C-states if there are no _CST or PBLK */
 416                        pr_backup->flags.power = test_bit(i, acpi_id_cst_present);
 417                        (void)upload_pm_data(pr_backup);
 418                }
 419        }
 420        kfree(acpi_id_present);
 421        acpi_id_present = NULL;
 422        kfree(acpi_id_cst_present);
 423        acpi_id_cst_present = NULL;
 424        return 0;
 425}
 426static int __init check_prereq(void)
 427{
 428        struct cpuinfo_x86 *c = &cpu_data(0);
 429
 430        if (!xen_initial_domain())
 431                return -ENODEV;
 432
 433        if (!acpi_gbl_FADT.smi_command)
 434                return -ENODEV;
 435
 436        if (c->x86_vendor == X86_VENDOR_INTEL) {
 437                if (!cpu_has(c, X86_FEATURE_EST))
 438                        return -ENODEV;
 439
 440                return 0;
 441        }
 442        if (c->x86_vendor == X86_VENDOR_AMD) {
 443                /* Copied from powernow-k8.h, can't include ../cpufreq/powernow
 444                 * as we get compile warnings for the static functions.
 445                 */
 446#define CPUID_FREQ_VOLT_CAPABILITIES    0x80000007
 447#define USE_HW_PSTATE                   0x00000080
 448                u32 eax, ebx, ecx, edx;
 449                cpuid(CPUID_FREQ_VOLT_CAPABILITIES, &eax, &ebx, &ecx, &edx);
 450                if ((edx & USE_HW_PSTATE) != USE_HW_PSTATE)
 451                        return -ENODEV;
 452                return 0;
 453        }
 454        return -ENODEV;
 455}
 456/* acpi_perf_data is a pointer to percpu data. */
 457static struct acpi_processor_performance __percpu *acpi_perf_data;
 458
 459static void free_acpi_perf_data(void)
 460{
 461        unsigned int i;
 462
 463        /* Freeing a NULL pointer is OK, and alloc_percpu zeroes. */
 464        for_each_possible_cpu(i)
 465                free_cpumask_var(per_cpu_ptr(acpi_perf_data, i)
 466                                 ->shared_cpu_map);
 467        free_percpu(acpi_perf_data);
 468}
 469
 470static int __init xen_acpi_processor_init(void)
 471{
 472        struct acpi_processor *pr_backup = NULL;
 473        unsigned int i;
 474        int rc = check_prereq();
 475
 476        if (rc)
 477                return rc;
 478
 479        nr_acpi_bits = get_max_acpi_id() + 1;
 480        acpi_ids_done = kcalloc(BITS_TO_LONGS(nr_acpi_bits), sizeof(unsigned long), GFP_KERNEL);
 481        if (!acpi_ids_done)
 482                return -ENOMEM;
 483
 484        acpi_perf_data = alloc_percpu(struct acpi_processor_performance);
 485        if (!acpi_perf_data) {
 486                pr_debug(DRV_NAME "Memory allocation error for acpi_perf_data.\n");
 487                kfree(acpi_ids_done);
 488                return -ENOMEM;
 489        }
 490        for_each_possible_cpu(i) {
 491                if (!zalloc_cpumask_var_node(
 492                        &per_cpu_ptr(acpi_perf_data, i)->shared_cpu_map,
 493                        GFP_KERNEL, cpu_to_node(i))) {
 494                        rc = -ENOMEM;
 495                        goto err_out;
 496                }
 497        }
 498
 499        /* Do initialization in ACPI core. It is OK to fail here. */
 500        (void)acpi_processor_preregister_performance(acpi_perf_data);
 501
 502        for_each_possible_cpu(i) {
 503                struct acpi_processor_performance *perf;
 504
 505                perf = per_cpu_ptr(acpi_perf_data, i);
 506                rc = acpi_processor_register_performance(perf, i);
 507                if (rc)
 508                        goto err_out;
 509        }
 510        rc = acpi_processor_notify_smm(THIS_MODULE);
 511        if (rc)
 512                goto err_unregister;
 513
 514        for_each_possible_cpu(i) {
 515                struct acpi_processor *_pr;
 516                _pr = per_cpu(processors, i /* APIC ID */);
 517                if (!_pr)
 518                        continue;
 519
 520                if (!pr_backup) {
 521                        pr_backup = kzalloc(sizeof(struct acpi_processor), GFP_KERNEL);
 522                        if (pr_backup)
 523                                memcpy(pr_backup, _pr, sizeof(struct acpi_processor));
 524                }
 525                (void)upload_pm_data(_pr);
 526        }
 527        rc = check_acpi_ids(pr_backup);
 528
 529        kfree(pr_backup);
 530        pr_backup = NULL;
 531
 532        if (rc)
 533                goto err_unregister;
 534
 535        return 0;
 536err_unregister:
 537        for_each_possible_cpu(i) {
 538                struct acpi_processor_performance *perf;
 539                perf = per_cpu_ptr(acpi_perf_data, i);
 540                acpi_processor_unregister_performance(perf, i);
 541        }
 542err_out:
 543        /* Freeing a NULL pointer is OK: alloc_percpu zeroes. */
 544        free_acpi_perf_data();
 545        kfree(acpi_ids_done);
 546        return rc;
 547}
 548static void __exit xen_acpi_processor_exit(void)
 549{
 550        int i;
 551
 552        kfree(acpi_ids_done);
 553        for_each_possible_cpu(i) {
 554                struct acpi_processor_performance *perf;
 555                perf = per_cpu_ptr(acpi_perf_data, i);
 556                acpi_processor_unregister_performance(perf, i);
 557        }
 558        free_acpi_perf_data();
 559}
 560
 561MODULE_AUTHOR("Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>");
 562MODULE_DESCRIPTION("Xen ACPI Processor P-states (and Cx) driver which uploads PM data to Xen hypervisor");
 563MODULE_LICENSE("GPL");
 564
 565/* We want to be loaded before the CPU freq scaling drivers are loaded.
 566 * They are loaded in late_initcall. */
 567device_initcall(xen_acpi_processor_init);
 568module_exit(xen_acpi_processor_exit);
 569
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.