Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / drivers / s390 / cio / chp.c
diff --git a/kernel/drivers/s390/cio/chp.c b/kernel/drivers/s390/cio/chp.c
new file mode 100644 (file)
index 0000000..c692dfe
--- /dev/null
@@ -0,0 +1,792 @@
+/*
+ *    Copyright IBM Corp. 1999, 2010
+ *    Author(s): Cornelia Huck (cornelia.huck@de.ibm.com)
+ *              Arnd Bergmann (arndb@de.ibm.com)
+ *              Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
+ */
+
+#include <linux/bug.h>
+#include <linux/workqueue.h>
+#include <linux/spinlock.h>
+#include <linux/export.h>
+#include <linux/sched.h>
+#include <linux/init.h>
+#include <linux/jiffies.h>
+#include <linux/wait.h>
+#include <linux/mutex.h>
+#include <linux/errno.h>
+#include <linux/slab.h>
+#include <asm/chpid.h>
+#include <asm/sclp.h>
+#include <asm/crw.h>
+
+#include "cio.h"
+#include "css.h"
+#include "ioasm.h"
+#include "cio_debug.h"
+#include "chp.h"
+
+#define to_channelpath(device) container_of(device, struct channel_path, dev)
+#define CHP_INFO_UPDATE_INTERVAL       1*HZ
+
+enum cfg_task_t {
+       cfg_none,
+       cfg_configure,
+       cfg_deconfigure
+};
+
+/* Map for pending configure tasks. */
+static enum cfg_task_t chp_cfg_task[__MAX_CSSID + 1][__MAX_CHPID + 1];
+static DEFINE_MUTEX(cfg_lock);
+static int cfg_busy;
+
+/* Map for channel-path status. */
+static struct sclp_chp_info chp_info;
+static DEFINE_MUTEX(info_lock);
+
+/* Time after which channel-path status may be outdated. */
+static unsigned long chp_info_expires;
+
+/* Workqueue to perform pending configure tasks. */
+static struct workqueue_struct *chp_wq;
+static struct work_struct cfg_work;
+
+/* Wait queue for configure completion events. */
+static wait_queue_head_t cfg_wait_queue;
+
+/* Set vary state for given chpid. */
+static void set_chp_logically_online(struct chp_id chpid, int onoff)
+{
+       chpid_to_chp(chpid)->state = onoff;
+}
+
+/* On success return 0 if channel-path is varied offline, 1 if it is varied
+ * online. Return -ENODEV if channel-path is not registered. */
+int chp_get_status(struct chp_id chpid)
+{
+       return (chpid_to_chp(chpid) ? chpid_to_chp(chpid)->state : -ENODEV);
+}
+
+/**
+ * chp_get_sch_opm - return opm for subchannel
+ * @sch: subchannel
+ *
+ * Calculate and return the operational path mask (opm) based on the chpids
+ * used by the subchannel and the status of the associated channel-paths.
+ */
+u8 chp_get_sch_opm(struct subchannel *sch)
+{
+       struct chp_id chpid;
+       int opm;
+       int i;
+
+       opm = 0;
+       chp_id_init(&chpid);
+       for (i = 0; i < 8; i++) {
+               opm <<= 1;
+               chpid.id = sch->schib.pmcw.chpid[i];
+               if (chp_get_status(chpid) != 0)
+                       opm |= 1;
+       }
+       return opm;
+}
+EXPORT_SYMBOL_GPL(chp_get_sch_opm);
+
+/**
+ * chp_is_registered - check if a channel-path is registered
+ * @chpid: channel-path ID
+ *
+ * Return non-zero if a channel-path with the given chpid is registered,
+ * zero otherwise.
+ */
+int chp_is_registered(struct chp_id chpid)
+{
+       return chpid_to_chp(chpid) != NULL;
+}
+
+/*
+ * Function: s390_vary_chpid
+ * Varies the specified chpid online or offline
+ */
+static int s390_vary_chpid(struct chp_id chpid, int on)
+{
+       char dbf_text[15];
+       int status;
+
+       sprintf(dbf_text, on?"varyon%x.%02x":"varyoff%x.%02x", chpid.cssid,
+               chpid.id);
+       CIO_TRACE_EVENT(2, dbf_text);
+
+       status = chp_get_status(chpid);
+       if (!on && !status)
+               return 0;
+
+       set_chp_logically_online(chpid, on);
+       chsc_chp_vary(chpid, on);
+       return 0;
+}
+
+/*
+ * Channel measurement related functions
+ */
+static ssize_t chp_measurement_chars_read(struct file *filp,
+                                         struct kobject *kobj,
+                                         struct bin_attribute *bin_attr,
+                                         char *buf, loff_t off, size_t count)
+{
+       struct channel_path *chp;
+       struct device *device;
+
+       device = container_of(kobj, struct device, kobj);
+       chp = to_channelpath(device);
+       if (!chp->cmg_chars)
+               return 0;
+
+       return memory_read_from_buffer(buf, count, &off,
+                               chp->cmg_chars, sizeof(struct cmg_chars));
+}
+
+static struct bin_attribute chp_measurement_chars_attr = {
+       .attr = {
+               .name = "measurement_chars",
+               .mode = S_IRUSR,
+       },
+       .size = sizeof(struct cmg_chars),
+       .read = chp_measurement_chars_read,
+};
+
+static void chp_measurement_copy_block(struct cmg_entry *buf,
+                                      struct channel_subsystem *css,
+                                      struct chp_id chpid)
+{
+       void *area;
+       struct cmg_entry *entry, reference_buf;
+       int idx;
+
+       if (chpid.id < 128) {
+               area = css->cub_addr1;
+               idx = chpid.id;
+       } else {
+               area = css->cub_addr2;
+               idx = chpid.id - 128;
+       }
+       entry = area + (idx * sizeof(struct cmg_entry));
+       do {
+               memcpy(buf, entry, sizeof(*entry));
+               memcpy(&reference_buf, entry, sizeof(*entry));
+       } while (reference_buf.values[0] != buf->values[0]);
+}
+
+static ssize_t chp_measurement_read(struct file *filp, struct kobject *kobj,
+                                   struct bin_attribute *bin_attr,
+                                   char *buf, loff_t off, size_t count)
+{
+       struct channel_path *chp;
+       struct channel_subsystem *css;
+       struct device *device;
+       unsigned int size;
+
+       device = container_of(kobj, struct device, kobj);
+       chp = to_channelpath(device);
+       css = to_css(chp->dev.parent);
+
+       size = sizeof(struct cmg_entry);
+
+       /* Only allow single reads. */
+       if (off || count < size)
+               return 0;
+       chp_measurement_copy_block((struct cmg_entry *)buf, css, chp->chpid);
+       count = size;
+       return count;
+}
+
+static struct bin_attribute chp_measurement_attr = {
+       .attr = {
+               .name = "measurement",
+               .mode = S_IRUSR,
+       },
+       .size = sizeof(struct cmg_entry),
+       .read = chp_measurement_read,
+};
+
+void chp_remove_cmg_attr(struct channel_path *chp)
+{
+       device_remove_bin_file(&chp->dev, &chp_measurement_chars_attr);
+       device_remove_bin_file(&chp->dev, &chp_measurement_attr);
+}
+
+int chp_add_cmg_attr(struct channel_path *chp)
+{
+       int ret;
+
+       ret = device_create_bin_file(&chp->dev, &chp_measurement_chars_attr);
+       if (ret)
+               return ret;
+       ret = device_create_bin_file(&chp->dev, &chp_measurement_attr);
+       if (ret)
+               device_remove_bin_file(&chp->dev, &chp_measurement_chars_attr);
+       return ret;
+}
+
+/*
+ * Files for the channel path entries.
+ */
+static ssize_t chp_status_show(struct device *dev,
+                              struct device_attribute *attr, char *buf)
+{
+       struct channel_path *chp = to_channelpath(dev);
+       int status;
+
+       mutex_lock(&chp->lock);
+       status = chp->state;
+       mutex_unlock(&chp->lock);
+
+       return status ? sprintf(buf, "online\n") : sprintf(buf, "offline\n");
+}
+
+static ssize_t chp_status_write(struct device *dev,
+                               struct device_attribute *attr,
+                               const char *buf, size_t count)
+{
+       struct channel_path *cp = to_channelpath(dev);
+       char cmd[10];
+       int num_args;
+       int error;
+
+       num_args = sscanf(buf, "%5s", cmd);
+       if (!num_args)
+               return count;
+
+       if (!strncasecmp(cmd, "on", 2) || !strcmp(cmd, "1")) {
+               mutex_lock(&cp->lock);
+               error = s390_vary_chpid(cp->chpid, 1);
+               mutex_unlock(&cp->lock);
+       } else if (!strncasecmp(cmd, "off", 3) || !strcmp(cmd, "0")) {
+               mutex_lock(&cp->lock);
+               error = s390_vary_chpid(cp->chpid, 0);
+               mutex_unlock(&cp->lock);
+       } else
+               error = -EINVAL;
+
+       return error < 0 ? error : count;
+}
+
+static DEVICE_ATTR(status, 0644, chp_status_show, chp_status_write);
+
+static ssize_t chp_configure_show(struct device *dev,
+                                 struct device_attribute *attr, char *buf)
+{
+       struct channel_path *cp;
+       int status;
+
+       cp = to_channelpath(dev);
+       status = chp_info_get_status(cp->chpid);
+       if (status < 0)
+               return status;
+
+       return snprintf(buf, PAGE_SIZE, "%d\n", status);
+}
+
+static int cfg_wait_idle(void);
+
+static ssize_t chp_configure_write(struct device *dev,
+                                  struct device_attribute *attr,
+                                  const char *buf, size_t count)
+{
+       struct channel_path *cp;
+       int val;
+       char delim;
+
+       if (sscanf(buf, "%d %c", &val, &delim) != 1)
+               return -EINVAL;
+       if (val != 0 && val != 1)
+               return -EINVAL;
+       cp = to_channelpath(dev);
+       chp_cfg_schedule(cp->chpid, val);
+       cfg_wait_idle();
+
+       return count;
+}
+
+static DEVICE_ATTR(configure, 0644, chp_configure_show, chp_configure_write);
+
+static ssize_t chp_type_show(struct device *dev, struct device_attribute *attr,
+                            char *buf)
+{
+       struct channel_path *chp = to_channelpath(dev);
+       u8 type;
+
+       mutex_lock(&chp->lock);
+       type = chp->desc.desc;
+       mutex_unlock(&chp->lock);
+       return sprintf(buf, "%x\n", type);
+}
+
+static DEVICE_ATTR(type, 0444, chp_type_show, NULL);
+
+static ssize_t chp_cmg_show(struct device *dev, struct device_attribute *attr,
+                           char *buf)
+{
+       struct channel_path *chp = to_channelpath(dev);
+
+       if (!chp)
+               return 0;
+       if (chp->cmg == -1) /* channel measurements not available */
+               return sprintf(buf, "unknown\n");
+       return sprintf(buf, "%x\n", chp->cmg);
+}
+
+static DEVICE_ATTR(cmg, 0444, chp_cmg_show, NULL);
+
+static ssize_t chp_shared_show(struct device *dev,
+                              struct device_attribute *attr, char *buf)
+{
+       struct channel_path *chp = to_channelpath(dev);
+
+       if (!chp)
+               return 0;
+       if (chp->shared == -1) /* channel measurements not available */
+               return sprintf(buf, "unknown\n");
+       return sprintf(buf, "%x\n", chp->shared);
+}
+
+static DEVICE_ATTR(shared, 0444, chp_shared_show, NULL);
+
+static ssize_t chp_chid_show(struct device *dev, struct device_attribute *attr,
+                            char *buf)
+{
+       struct channel_path *chp = to_channelpath(dev);
+       ssize_t rc;
+
+       mutex_lock(&chp->lock);
+       if (chp->desc_fmt1.flags & 0x10)
+               rc = sprintf(buf, "%04x\n", chp->desc_fmt1.chid);
+       else
+               rc = 0;
+       mutex_unlock(&chp->lock);
+
+       return rc;
+}
+static DEVICE_ATTR(chid, 0444, chp_chid_show, NULL);
+
+static ssize_t chp_chid_external_show(struct device *dev,
+                                     struct device_attribute *attr, char *buf)
+{
+       struct channel_path *chp = to_channelpath(dev);
+       ssize_t rc;
+
+       mutex_lock(&chp->lock);
+       if (chp->desc_fmt1.flags & 0x10)
+               rc = sprintf(buf, "%x\n", chp->desc_fmt1.flags & 0x8 ? 1 : 0);
+       else
+               rc = 0;
+       mutex_unlock(&chp->lock);
+
+       return rc;
+}
+static DEVICE_ATTR(chid_external, 0444, chp_chid_external_show, NULL);
+
+static struct attribute *chp_attrs[] = {
+       &dev_attr_status.attr,
+       &dev_attr_configure.attr,
+       &dev_attr_type.attr,
+       &dev_attr_cmg.attr,
+       &dev_attr_shared.attr,
+       &dev_attr_chid.attr,
+       &dev_attr_chid_external.attr,
+       NULL,
+};
+static struct attribute_group chp_attr_group = {
+       .attrs = chp_attrs,
+};
+static const struct attribute_group *chp_attr_groups[] = {
+       &chp_attr_group,
+       NULL,
+};
+
+static void chp_release(struct device *dev)
+{
+       struct channel_path *cp;
+
+       cp = to_channelpath(dev);
+       kfree(cp);
+}
+
+/**
+ * chp_update_desc - update channel-path description
+ * @chp - channel-path
+ *
+ * Update the channel-path description of the specified channel-path.
+ * Return zero on success, non-zero otherwise.
+ */
+int chp_update_desc(struct channel_path *chp)
+{
+       int rc;
+
+       rc = chsc_determine_base_channel_path_desc(chp->chpid, &chp->desc);
+       if (rc)
+               return rc;
+
+       rc = chsc_determine_fmt1_channel_path_desc(chp->chpid, &chp->desc_fmt1);
+
+       return rc;
+}
+
+/**
+ * chp_new - register a new channel-path
+ * @chpid - channel-path ID
+ *
+ * Create and register data structure representing new channel-path. Return
+ * zero on success, non-zero otherwise.
+ */
+int chp_new(struct chp_id chpid)
+{
+       struct channel_path *chp;
+       int ret;
+
+       if (chp_is_registered(chpid))
+               return 0;
+       chp = kzalloc(sizeof(struct channel_path), GFP_KERNEL);
+       if (!chp)
+               return -ENOMEM;
+
+       /* fill in status, etc. */
+       chp->chpid = chpid;
+       chp->state = 1;
+       chp->dev.parent = &channel_subsystems[chpid.cssid]->device;
+       chp->dev.groups = chp_attr_groups;
+       chp->dev.release = chp_release;
+       mutex_init(&chp->lock);
+
+       /* Obtain channel path description and fill it in. */
+       ret = chp_update_desc(chp);
+       if (ret)
+               goto out_free;
+       if ((chp->desc.flags & 0x80) == 0) {
+               ret = -ENODEV;
+               goto out_free;
+       }
+       /* Get channel-measurement characteristics. */
+       if (css_chsc_characteristics.scmc && css_chsc_characteristics.secm) {
+               ret = chsc_get_channel_measurement_chars(chp);
+               if (ret)
+                       goto out_free;
+       } else {
+               chp->cmg = -1;
+       }
+       dev_set_name(&chp->dev, "chp%x.%02x", chpid.cssid, chpid.id);
+
+       /* make it known to the system */
+       ret = device_register(&chp->dev);
+       if (ret) {
+               CIO_MSG_EVENT(0, "Could not register chp%x.%02x: %d\n",
+                             chpid.cssid, chpid.id, ret);
+               put_device(&chp->dev);
+               goto out;
+       }
+       mutex_lock(&channel_subsystems[chpid.cssid]->mutex);
+       if (channel_subsystems[chpid.cssid]->cm_enabled) {
+               ret = chp_add_cmg_attr(chp);
+               if (ret) {
+                       device_unregister(&chp->dev);
+                       mutex_unlock(&channel_subsystems[chpid.cssid]->mutex);
+                       goto out;
+               }
+       }
+       channel_subsystems[chpid.cssid]->chps[chpid.id] = chp;
+       mutex_unlock(&channel_subsystems[chpid.cssid]->mutex);
+       goto out;
+out_free:
+       kfree(chp);
+out:
+       return ret;
+}
+
+/**
+ * chp_get_chp_desc - return newly allocated channel-path description
+ * @chpid: channel-path ID
+ *
+ * On success return a newly allocated copy of the channel-path description
+ * data associated with the given channel-path ID. Return %NULL on error.
+ */
+struct channel_path_desc *chp_get_chp_desc(struct chp_id chpid)
+{
+       struct channel_path *chp;
+       struct channel_path_desc *desc;
+
+       chp = chpid_to_chp(chpid);
+       if (!chp)
+               return NULL;
+       desc = kmalloc(sizeof(struct channel_path_desc), GFP_KERNEL);
+       if (!desc)
+               return NULL;
+
+       mutex_lock(&chp->lock);
+       memcpy(desc, &chp->desc, sizeof(struct channel_path_desc));
+       mutex_unlock(&chp->lock);
+       return desc;
+}
+
+/**
+ * chp_process_crw - process channel-path status change
+ * @crw0: channel report-word to handler
+ * @crw1: second channel-report word (always NULL)
+ * @overflow: crw overflow indication
+ *
+ * Handle channel-report-words indicating that the status of a channel-path
+ * has changed.
+ */
+static void chp_process_crw(struct crw *crw0, struct crw *crw1,
+                           int overflow)
+{
+       struct chp_id chpid;
+
+       if (overflow) {
+               css_schedule_eval_all();
+               return;
+       }
+       CIO_CRW_EVENT(2, "CRW reports slct=%d, oflw=%d, "
+                     "chn=%d, rsc=%X, anc=%d, erc=%X, rsid=%X\n",
+                     crw0->slct, crw0->oflw, crw0->chn, crw0->rsc, crw0->anc,
+                     crw0->erc, crw0->rsid);
+       /*
+        * Check for solicited machine checks. These are
+        * created by reset channel path and need not be
+        * handled here.
+        */
+       if (crw0->slct) {
+               CIO_CRW_EVENT(2, "solicited machine check for "
+                             "channel path %02X\n", crw0->rsid);
+               return;
+       }
+       chp_id_init(&chpid);
+       chpid.id = crw0->rsid;
+       switch (crw0->erc) {
+       case CRW_ERC_IPARM: /* Path has come. */
+               if (!chp_is_registered(chpid))
+                       chp_new(chpid);
+               chsc_chp_online(chpid);
+               break;
+       case CRW_ERC_PERRI: /* Path has gone. */
+       case CRW_ERC_PERRN:
+               chsc_chp_offline(chpid);
+               break;
+       default:
+               CIO_CRW_EVENT(2, "Don't know how to handle erc=%x\n",
+                             crw0->erc);
+       }
+}
+
+int chp_ssd_get_mask(struct chsc_ssd_info *ssd, struct chp_link *link)
+{
+       int i;
+       int mask;
+
+       for (i = 0; i < 8; i++) {
+               mask = 0x80 >> i;
+               if (!(ssd->path_mask & mask))
+                       continue;
+               if (!chp_id_is_equal(&ssd->chpid[i], &link->chpid))
+                       continue;
+               if ((ssd->fla_valid_mask & mask) &&
+                   ((ssd->fla[i] & link->fla_mask) != link->fla))
+                       continue;
+               return mask;
+       }
+       return 0;
+}
+EXPORT_SYMBOL_GPL(chp_ssd_get_mask);
+
+static inline int info_bit_num(struct chp_id id)
+{
+       return id.id + id.cssid * (__MAX_CHPID + 1);
+}
+
+/* Force chp_info refresh on next call to info_validate(). */
+static void info_expire(void)
+{
+       mutex_lock(&info_lock);
+       chp_info_expires = jiffies - 1;
+       mutex_unlock(&info_lock);
+}
+
+/* Ensure that chp_info is up-to-date. */
+static int info_update(void)
+{
+       int rc;
+
+       mutex_lock(&info_lock);
+       rc = 0;
+       if (time_after(jiffies, chp_info_expires)) {
+               /* Data is too old, update. */
+               rc = sclp_chp_read_info(&chp_info);
+               chp_info_expires = jiffies + CHP_INFO_UPDATE_INTERVAL ;
+       }
+       mutex_unlock(&info_lock);
+
+       return rc;
+}
+
+/**
+ * chp_info_get_status - retrieve configure status of a channel-path
+ * @chpid: channel-path ID
+ *
+ * On success, return 0 for standby, 1 for configured, 2 for reserved,
+ * 3 for not recognized. Return negative error code on error.
+ */
+int chp_info_get_status(struct chp_id chpid)
+{
+       int rc;
+       int bit;
+
+       rc = info_update();
+       if (rc)
+               return rc;
+
+       bit = info_bit_num(chpid);
+       mutex_lock(&info_lock);
+       if (!chp_test_bit(chp_info.recognized, bit))
+               rc = CHP_STATUS_NOT_RECOGNIZED;
+       else if (chp_test_bit(chp_info.configured, bit))
+               rc = CHP_STATUS_CONFIGURED;
+       else if (chp_test_bit(chp_info.standby, bit))
+               rc = CHP_STATUS_STANDBY;
+       else
+               rc = CHP_STATUS_RESERVED;
+       mutex_unlock(&info_lock);
+
+       return rc;
+}
+
+/* Return configure task for chpid. */
+static enum cfg_task_t cfg_get_task(struct chp_id chpid)
+{
+       return chp_cfg_task[chpid.cssid][chpid.id];
+}
+
+/* Set configure task for chpid. */
+static void cfg_set_task(struct chp_id chpid, enum cfg_task_t cfg)
+{
+       chp_cfg_task[chpid.cssid][chpid.id] = cfg;
+}
+
+/* Perform one configure/deconfigure request. Reschedule work function until
+ * last request. */
+static void cfg_func(struct work_struct *work)
+{
+       struct chp_id chpid;
+       enum cfg_task_t t;
+       int rc;
+
+       mutex_lock(&cfg_lock);
+       t = cfg_none;
+       chp_id_for_each(&chpid) {
+               t = cfg_get_task(chpid);
+               if (t != cfg_none) {
+                       cfg_set_task(chpid, cfg_none);
+                       break;
+               }
+       }
+       mutex_unlock(&cfg_lock);
+
+       switch (t) {
+       case cfg_configure:
+               rc = sclp_chp_configure(chpid);
+               if (rc)
+                       CIO_MSG_EVENT(2, "chp: sclp_chp_configure(%x.%02x)="
+                                     "%d\n", chpid.cssid, chpid.id, rc);
+               else {
+                       info_expire();
+                       chsc_chp_online(chpid);
+               }
+               break;
+       case cfg_deconfigure:
+               rc = sclp_chp_deconfigure(chpid);
+               if (rc)
+                       CIO_MSG_EVENT(2, "chp: sclp_chp_deconfigure(%x.%02x)="
+                                     "%d\n", chpid.cssid, chpid.id, rc);
+               else {
+                       info_expire();
+                       chsc_chp_offline(chpid);
+               }
+               break;
+       case cfg_none:
+               /* Get updated information after last change. */
+               info_update();
+               mutex_lock(&cfg_lock);
+               cfg_busy = 0;
+               mutex_unlock(&cfg_lock);
+               wake_up_interruptible(&cfg_wait_queue);
+               return;
+       }
+       queue_work(chp_wq, &cfg_work);
+}
+
+/**
+ * chp_cfg_schedule - schedule chpid configuration request
+ * @chpid - channel-path ID
+ * @configure - Non-zero for configure, zero for deconfigure
+ *
+ * Schedule a channel-path configuration/deconfiguration request.
+ */
+void chp_cfg_schedule(struct chp_id chpid, int configure)
+{
+       CIO_MSG_EVENT(2, "chp_cfg_sched%x.%02x=%d\n", chpid.cssid, chpid.id,
+                     configure);
+       mutex_lock(&cfg_lock);
+       cfg_set_task(chpid, configure ? cfg_configure : cfg_deconfigure);
+       cfg_busy = 1;
+       mutex_unlock(&cfg_lock);
+       queue_work(chp_wq, &cfg_work);
+}
+
+/**
+ * chp_cfg_cancel_deconfigure - cancel chpid deconfiguration request
+ * @chpid - channel-path ID
+ *
+ * Cancel an active channel-path deconfiguration request if it has not yet
+ * been performed.
+ */
+void chp_cfg_cancel_deconfigure(struct chp_id chpid)
+{
+       CIO_MSG_EVENT(2, "chp_cfg_cancel:%x.%02x\n", chpid.cssid, chpid.id);
+       mutex_lock(&cfg_lock);
+       if (cfg_get_task(chpid) == cfg_deconfigure)
+               cfg_set_task(chpid, cfg_none);
+       mutex_unlock(&cfg_lock);
+}
+
+static int cfg_wait_idle(void)
+{
+       if (wait_event_interruptible(cfg_wait_queue, !cfg_busy))
+               return -ERESTARTSYS;
+       return 0;
+}
+
+static int __init chp_init(void)
+{
+       struct chp_id chpid;
+       int ret;
+
+       ret = crw_register_handler(CRW_RSC_CPATH, chp_process_crw);
+       if (ret)
+               return ret;
+       chp_wq = create_singlethread_workqueue("cio_chp");
+       if (!chp_wq) {
+               crw_unregister_handler(CRW_RSC_CPATH);
+               return -ENOMEM;
+       }
+       INIT_WORK(&cfg_work, cfg_func);
+       init_waitqueue_head(&cfg_wait_queue);
+       if (info_update())
+               return 0;
+       /* Register available channel-paths. */
+       chp_id_for_each(&chpid) {
+               if (chp_info_get_status(chpid) != CHP_STATUS_NOT_RECOGNIZED)
+                       chp_new(chpid);
+       }
+
+       return 0;
+}
+
+subsys_initcall(chp_init);