summaryrefslogtreecommitdiffstats
path: root/drivers/net/wireless/marvell/libertas/debugfs.c
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/net/wireless/marvell/libertas/debugfs.c')
-rw-r--r--drivers/net/wireless/marvell/libertas/debugfs.c936
1 files changed, 936 insertions, 0 deletions
diff --git a/drivers/net/wireless/marvell/libertas/debugfs.c b/drivers/net/wireless/marvell/libertas/debugfs.c
new file mode 100644
index 000000000000..faed1823c58e
--- /dev/null
+++ b/drivers/net/wireless/marvell/libertas/debugfs.c
@@ -0,0 +1,936 @@
+#include <linux/dcache.h>
+#include <linux/debugfs.h>
+#include <linux/delay.h>
+#include <linux/hardirq.h>
+#include <linux/mm.h>
+#include <linux/string.h>
+#include <linux/slab.h>
+#include <linux/export.h>
+
+#include "decl.h"
+#include "cmd.h"
+#include "debugfs.h"
+
+static struct dentry *lbs_dir;
+static char *szStates[] = {
+ "Connected",
+ "Disconnected"
+};
+
+#ifdef PROC_DEBUG
+static void lbs_debug_init(struct lbs_private *priv);
+#endif
+
+static ssize_t write_file_dummy(struct file *file, const char __user *buf,
+ size_t count, loff_t *ppos)
+{
+ return -EINVAL;
+}
+
+static const size_t len = PAGE_SIZE;
+
+static ssize_t lbs_dev_info(struct file *file, char __user *userbuf,
+ size_t count, loff_t *ppos)
+{
+ struct lbs_private *priv = file->private_data;
+ size_t pos = 0;
+ unsigned long addr = get_zeroed_page(GFP_KERNEL);
+ char *buf = (char *)addr;
+ ssize_t res;
+ if (!buf)
+ return -ENOMEM;
+
+ pos += snprintf(buf+pos, len-pos, "state = %s\n",
+ szStates[priv->connect_status]);
+ pos += snprintf(buf+pos, len-pos, "region_code = %02x\n",
+ (u32) priv->regioncode);
+
+ res = simple_read_from_buffer(userbuf, count, ppos, buf, pos);
+
+ free_page(addr);
+ return res;
+}
+
+static ssize_t lbs_sleepparams_write(struct file *file,
+ const char __user *user_buf, size_t count,
+ loff_t *ppos)
+{
+ struct lbs_private *priv = file->private_data;
+ ssize_t ret;
+ struct sleep_params sp;
+ int p1, p2, p3, p4, p5, p6;
+ char *buf;
+
+ buf = memdup_user_nul(user_buf, min(count, len - 1));
+ if (IS_ERR(buf))
+ return PTR_ERR(buf);
+
+ ret = sscanf(buf, "%d %d %d %d %d %d", &p1, &p2, &p3, &p4, &p5, &p6);
+ if (ret != 6) {
+ ret = -EINVAL;
+ goto out_unlock;
+ }
+ sp.sp_error = p1;
+ sp.sp_offset = p2;
+ sp.sp_stabletime = p3;
+ sp.sp_calcontrol = p4;
+ sp.sp_extsleepclk = p5;
+ sp.sp_reserved = p6;
+
+ ret = lbs_cmd_802_11_sleep_params(priv, CMD_ACT_SET, &sp);
+ if (!ret)
+ ret = count;
+ else if (ret > 0)
+ ret = -EINVAL;
+
+out_unlock:
+ kfree(buf);
+ return ret;
+}
+
+static ssize_t lbs_sleepparams_read(struct file *file, char __user *userbuf,
+ size_t count, loff_t *ppos)
+{
+ struct lbs_private *priv = file->private_data;
+ ssize_t ret;
+ size_t pos = 0;
+ struct sleep_params sp;
+ unsigned long addr = get_zeroed_page(GFP_KERNEL);
+ char *buf = (char *)addr;
+ if (!buf)
+ return -ENOMEM;
+
+ ret = lbs_cmd_802_11_sleep_params(priv, CMD_ACT_GET, &sp);
+ if (ret)
+ goto out_unlock;
+
+ pos += snprintf(buf, len, "%d %d %d %d %d %d\n", sp.sp_error,
+ sp.sp_offset, sp.sp_stabletime,
+ sp.sp_calcontrol, sp.sp_extsleepclk,
+ sp.sp_reserved);
+
+ ret = simple_read_from_buffer(userbuf, count, ppos, buf, pos);
+
+out_unlock:
+ free_page(addr);
+ return ret;
+}
+
+static ssize_t lbs_host_sleep_write(struct file *file,
+ const char __user *user_buf, size_t count,
+ loff_t *ppos)
+{
+ struct lbs_private *priv = file->private_data;
+ ssize_t ret;
+ int host_sleep;
+ char *buf;
+
+ buf = memdup_user_nul(user_buf, min(count, len - 1));
+ if (IS_ERR(buf))
+ return PTR_ERR(buf);
+
+ ret = sscanf(buf, "%d", &host_sleep);
+ if (ret != 1) {
+ ret = -EINVAL;
+ goto out_unlock;
+ }
+
+ if (host_sleep == 0)
+ ret = lbs_set_host_sleep(priv, 0);
+ else if (host_sleep == 1) {
+ if (priv->wol_criteria == EHS_REMOVE_WAKEUP) {
+ netdev_info(priv->dev,
+ "wake parameters not configured\n");
+ ret = -EINVAL;
+ goto out_unlock;
+ }
+ ret = lbs_set_host_sleep(priv, 1);
+ } else {
+ netdev_err(priv->dev, "invalid option\n");
+ ret = -EINVAL;
+ }
+
+ if (!ret)
+ ret = count;
+
+out_unlock:
+ kfree(buf);
+ return ret;
+}
+
+static ssize_t lbs_host_sleep_read(struct file *file, char __user *userbuf,
+ size_t count, loff_t *ppos)
+{
+ struct lbs_private *priv = file->private_data;
+ ssize_t ret;
+ size_t pos = 0;
+ unsigned long addr = get_zeroed_page(GFP_KERNEL);
+ char *buf = (char *)addr;
+ if (!buf)
+ return -ENOMEM;
+
+ pos += snprintf(buf, len, "%d\n", priv->is_host_sleep_activated);
+
+ ret = simple_read_from_buffer(userbuf, count, ppos, buf, pos);
+
+ free_page(addr);
+ return ret;
+}
+
+/*
+ * When calling CMD_802_11_SUBSCRIBE_EVENT with CMD_ACT_GET, me might
+ * get a bunch of vendor-specific TLVs (a.k.a. IEs) back from the
+ * firmware. Here's an example:
+ * 04 01 02 00 00 00 05 01 02 00 00 00 06 01 02 00
+ * 00 00 07 01 02 00 3c 00 00 00 00 00 00 00 03 03
+ * 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
+ *
+ * The 04 01 is the TLV type (here TLV_TYPE_RSSI_LOW), 02 00 is the length,
+ * 00 00 are the data bytes of this TLV. For this TLV, their meaning is
+ * defined in mrvlietypes_thresholds
+ *
+ * This function searches in this TLV data chunk for a given TLV type
+ * and returns a pointer to the first data byte of the TLV, or to NULL
+ * if the TLV hasn't been found.
+ */
+static void *lbs_tlv_find(uint16_t tlv_type, const uint8_t *tlv, uint16_t size)
+{
+ struct mrvl_ie_header *tlv_h;
+ uint16_t length;
+ ssize_t pos = 0;
+
+ while (pos < size) {
+ tlv_h = (struct mrvl_ie_header *) tlv;
+ if (!tlv_h->len)
+ return NULL;
+ if (tlv_h->type == cpu_to_le16(tlv_type))
+ return tlv_h;
+ length = le16_to_cpu(tlv_h->len) + sizeof(*tlv_h);
+ pos += length;
+ tlv += length;
+ }
+ return NULL;
+}
+
+
+static ssize_t lbs_threshold_read(uint16_t tlv_type, uint16_t event_mask,
+ struct file *file, char __user *userbuf,
+ size_t count, loff_t *ppos)
+{
+ struct cmd_ds_802_11_subscribe_event *subscribed;
+ struct mrvl_ie_thresholds *got;
+ struct lbs_private *priv = file->private_data;
+ ssize_t ret = 0;
+ size_t pos = 0;
+ char *buf;
+ u8 value;
+ u8 freq;
+ int events = 0;
+
+ buf = (char *)get_zeroed_page(GFP_KERNEL);
+ if (!buf)
+ return -ENOMEM;
+
+ subscribed = kzalloc(sizeof(*subscribed), GFP_KERNEL);
+ if (!subscribed) {
+ ret = -ENOMEM;
+ goto out_page;
+ }
+
+ subscribed->hdr.size = cpu_to_le16(sizeof(*subscribed));
+ subscribed->action = cpu_to_le16(CMD_ACT_GET);
+
+ ret = lbs_cmd_with_response(priv, CMD_802_11_SUBSCRIBE_EVENT, subscribed);
+ if (ret)
+ goto out_cmd;
+
+ got = lbs_tlv_find(tlv_type, subscribed->tlv, sizeof(subscribed->tlv));
+ if (got) {
+ value = got->value;
+ freq = got->freq;
+ events = le16_to_cpu(subscribed->events);
+
+ pos += snprintf(buf, len, "%d %d %d\n", value, freq,
+ !!(events & event_mask));
+ }
+
+ ret = simple_read_from_buffer(userbuf, count, ppos, buf, pos);
+
+ out_cmd:
+ kfree(subscribed);
+
+ out_page:
+ free_page((unsigned long)buf);
+ return ret;
+}
+
+
+static ssize_t lbs_threshold_write(uint16_t tlv_type, uint16_t event_mask,
+ struct file *file,
+ const char __user *userbuf, size_t count,
+ loff_t *ppos)
+{
+ struct cmd_ds_802_11_subscribe_event *events;
+ struct mrvl_ie_thresholds *tlv;
+ struct lbs_private *priv = file->private_data;
+ int value, freq, new_mask;
+ uint16_t curr_mask;
+ char *buf;
+ int ret;
+
+ buf = memdup_user_nul(userbuf, min(count, len - 1));
+ if (IS_ERR(buf))
+ return PTR_ERR(buf);
+
+ ret = sscanf(buf, "%d %d %d", &value, &freq, &new_mask);
+ if (ret != 3) {
+ ret = -EINVAL;
+ goto out_page;
+ }
+ events = kzalloc(sizeof(*events), GFP_KERNEL);
+ if (!events) {
+ ret = -ENOMEM;
+ goto out_page;
+ }
+
+ events->hdr.size = cpu_to_le16(sizeof(*events));
+ events->action = cpu_to_le16(CMD_ACT_GET);
+
+ ret = lbs_cmd_with_response(priv, CMD_802_11_SUBSCRIBE_EVENT, events);
+ if (ret)
+ goto out_events;
+
+ curr_mask = le16_to_cpu(events->events);
+
+ if (new_mask)
+ new_mask = curr_mask | event_mask;
+ else
+ new_mask = curr_mask & ~event_mask;
+
+ /* Now everything is set and we can send stuff down to the firmware */
+
+ tlv = (void *)events->tlv;
+
+ events->action = cpu_to_le16(CMD_ACT_SET);
+ events->events = cpu_to_le16(new_mask);
+ tlv->header.type = cpu_to_le16(tlv_type);
+ tlv->header.len = cpu_to_le16(sizeof(*tlv) - sizeof(tlv->header));
+ tlv->value = value;
+ if (tlv_type != TLV_TYPE_BCNMISS)
+ tlv->freq = freq;
+
+ /* The command header, the action, the event mask, and one TLV */
+ events->hdr.size = cpu_to_le16(sizeof(events->hdr) + 4 + sizeof(*tlv));
+
+ ret = lbs_cmd_with_response(priv, CMD_802_11_SUBSCRIBE_EVENT, events);
+
+ if (!ret)
+ ret = count;
+ out_events:
+ kfree(events);
+ out_page:
+ kfree(buf);
+ return ret;
+}
+
+
+static ssize_t lbs_lowrssi_read(struct file *file, char __user *userbuf,
+ size_t count, loff_t *ppos)
+{
+ return lbs_threshold_read(TLV_TYPE_RSSI_LOW, CMD_SUBSCRIBE_RSSI_LOW,
+ file, userbuf, count, ppos);
+}
+
+
+static ssize_t lbs_lowrssi_write(struct file *file, const char __user *userbuf,
+ size_t count, loff_t *ppos)
+{
+ return lbs_threshold_write(TLV_TYPE_RSSI_LOW, CMD_SUBSCRIBE_RSSI_LOW,
+ file, userbuf, count, ppos);
+}
+
+
+static ssize_t lbs_lowsnr_read(struct file *file, char __user *userbuf,
+ size_t count, loff_t *ppos)
+{
+ return lbs_threshold_read(TLV_TYPE_SNR_LOW, CMD_SUBSCRIBE_SNR_LOW,
+ file, userbuf, count, ppos);
+}
+
+
+static ssize_t lbs_lowsnr_write(struct file *file, const char __user *userbuf,
+ size_t count, loff_t *ppos)
+{
+ return lbs_threshold_write(TLV_TYPE_SNR_LOW, CMD_SUBSCRIBE_SNR_LOW,
+ file, userbuf, count, ppos);
+}
+
+
+static ssize_t lbs_failcount_read(struct file *file, char __user *userbuf,
+ size_t count, loff_t *ppos)
+{
+ return lbs_threshold_read(TLV_TYPE_FAILCOUNT, CMD_SUBSCRIBE_FAILCOUNT,
+ file, userbuf, count, ppos);
+}
+
+
+static ssize_t lbs_failcount_write(struct file *file, const char __user *userbuf,
+ size_t count, loff_t *ppos)
+{
+ return lbs_threshold_write(TLV_TYPE_FAILCOUNT, CMD_SUBSCRIBE_FAILCOUNT,
+ file, userbuf, count, ppos);
+}
+
+
+static ssize_t lbs_highrssi_read(struct file *file, char __user *userbuf,
+ size_t count, loff_t *ppos)
+{
+ return lbs_threshold_read(TLV_TYPE_RSSI_HIGH, CMD_SUBSCRIBE_RSSI_HIGH,
+ file, userbuf, count, ppos);
+}
+
+
+static ssize_t lbs_highrssi_write(struct file *file, const char __user *userbuf,
+ size_t count, loff_t *ppos)
+{
+ return lbs_threshold_write(TLV_TYPE_RSSI_HIGH, CMD_SUBSCRIBE_RSSI_HIGH,
+ file, userbuf, count, ppos);
+}
+
+
+static ssize_t lbs_highsnr_read(struct file *file, char __user *userbuf,
+ size_t count, loff_t *ppos)
+{
+ return lbs_threshold_read(TLV_TYPE_SNR_HIGH, CMD_SUBSCRIBE_SNR_HIGH,
+ file, userbuf, count, ppos);
+}
+
+
+static ssize_t lbs_highsnr_write(struct file *file, const char __user *userbuf,
+ size_t count, loff_t *ppos)
+{
+ return lbs_threshold_write(TLV_TYPE_SNR_HIGH, CMD_SUBSCRIBE_SNR_HIGH,
+ file, userbuf, count, ppos);
+}
+
+static ssize_t lbs_bcnmiss_read(struct file *file, char __user *userbuf,
+ size_t count, loff_t *ppos)
+{
+ return lbs_threshold_read(TLV_TYPE_BCNMISS, CMD_SUBSCRIBE_BCNMISS,
+ file, userbuf, count, ppos);
+}
+
+
+static ssize_t lbs_bcnmiss_write(struct file *file, const char __user *userbuf,
+ size_t count, loff_t *ppos)
+{
+ return lbs_threshold_write(TLV_TYPE_BCNMISS, CMD_SUBSCRIBE_BCNMISS,
+ file, userbuf, count, ppos);
+}
+
+
+static ssize_t lbs_rdmac_read(struct file *file, char __user *userbuf,
+ size_t count, loff_t *ppos)
+{
+ struct lbs_private *priv = file->private_data;
+ ssize_t pos = 0;
+ int ret;
+ unsigned long addr = get_zeroed_page(GFP_KERNEL);
+ char *buf = (char *)addr;
+ u32 val = 0;
+
+ if (!buf)
+ return -ENOMEM;
+
+ ret = lbs_get_reg(priv, CMD_MAC_REG_ACCESS, priv->mac_offset, &val);
+ mdelay(10);
+ if (!ret) {
+ pos = snprintf(buf, len, "MAC[0x%x] = 0x%08x\n",
+ priv->mac_offset, val);
+ ret = simple_read_from_buffer(userbuf, count, ppos, buf, pos);
+ }
+ free_page(addr);
+ return ret;
+}
+
+static ssize_t lbs_rdmac_write(struct file *file,
+ const char __user *userbuf,
+ size_t count, loff_t *ppos)
+{
+ struct lbs_private *priv = file->private_data;
+ char *buf;
+
+ buf = memdup_user_nul(userbuf, min(count, len - 1));
+ if (IS_ERR(buf))
+ return PTR_ERR(buf);
+
+ priv->mac_offset = simple_strtoul(buf, NULL, 16);
+ kfree(buf);
+ return count;
+}
+
+static ssize_t lbs_wrmac_write(struct file *file,
+ const char __user *userbuf,
+ size_t count, loff_t *ppos)
+{
+
+ struct lbs_private *priv = file->private_data;
+ ssize_t res;
+ u32 offset, value;
+ char *buf;
+
+ buf = memdup_user_nul(userbuf, min(count, len - 1));
+ if (IS_ERR(buf))
+ return PTR_ERR(buf);
+
+ res = sscanf(buf, "%x %x", &offset, &value);
+ if (res != 2) {
+ res = -EFAULT;
+ goto out_unlock;
+ }
+
+ res = lbs_set_reg(priv, CMD_MAC_REG_ACCESS, offset, value);
+ mdelay(10);
+
+ if (!res)
+ res = count;
+out_unlock:
+ kfree(buf);
+ return res;
+}
+
+static ssize_t lbs_rdbbp_read(struct file *file, char __user *userbuf,
+ size_t count, loff_t *ppos)
+{
+ struct lbs_private *priv = file->private_data;
+ ssize_t pos = 0;
+ int ret;
+ unsigned long addr = get_zeroed_page(GFP_KERNEL);
+ char *buf = (char *)addr;
+ u32 val;
+
+ if (!buf)
+ return -ENOMEM;
+
+ ret = lbs_get_reg(priv, CMD_BBP_REG_ACCESS, priv->bbp_offset, &val);
+ mdelay(10);
+ if (!ret) {
+ pos = snprintf(buf, len, "BBP[0x%x] = 0x%08x\n",
+ priv->bbp_offset, val);
+ ret = simple_read_from_buffer(userbuf, count, ppos, buf, pos);
+ }
+ free_page(addr);
+
+ return ret;
+}
+
+static ssize_t lbs_rdbbp_write(struct file *file,
+ const char __user *userbuf,
+ size_t count, loff_t *ppos)
+{
+ struct lbs_private *priv = file->private_data;
+ char *buf;
+
+ buf = memdup_user_nul(userbuf, min(count, len - 1));
+ if (IS_ERR(buf))
+ return PTR_ERR(buf);
+
+ priv->bbp_offset = simple_strtoul(buf, NULL, 16);
+ kfree(buf);
+
+ return count;
+}
+
+static ssize_t lbs_wrbbp_write(struct file *file,
+ const char __user *userbuf,
+ size_t count, loff_t *ppos)
+{
+
+ struct lbs_private *priv = file->private_data;
+ ssize_t res;
+ u32 offset, value;
+ char *buf;
+
+ buf = memdup_user_nul(userbuf, min(count, len - 1));
+ if (IS_ERR(buf))
+ return PTR_ERR(buf);
+
+ res = sscanf(buf, "%x %x", &offset, &value);
+ if (res != 2) {
+ res = -EFAULT;
+ goto out_unlock;
+ }
+
+ res = lbs_set_reg(priv, CMD_BBP_REG_ACCESS, offset, value);
+ mdelay(10);
+
+ if (!res)
+ res = count;
+out_unlock:
+ kfree(buf);
+ return res;
+}
+
+static ssize_t lbs_rdrf_read(struct file *file, char __user *userbuf,
+ size_t count, loff_t *ppos)
+{
+ struct lbs_private *priv = file->private_data;
+ ssize_t pos = 0;
+ int ret;
+ unsigned long addr = get_zeroed_page(GFP_KERNEL);
+ char *buf = (char *)addr;
+ u32 val;
+
+ if (!buf)
+ return -ENOMEM;
+
+ ret = lbs_get_reg(priv, CMD_RF_REG_ACCESS, priv->rf_offset, &val);
+ mdelay(10);
+ if (!ret) {
+ pos = snprintf(buf, len, "RF[0x%x] = 0x%08x\n",
+ priv->rf_offset, val);
+ ret = simple_read_from_buffer(userbuf, count, ppos, buf, pos);
+ }
+ free_page(addr);
+
+ return ret;
+}
+
+static ssize_t lbs_rdrf_write(struct file *file,
+ const char __user *userbuf,
+ size_t count, loff_t *ppos)
+{
+ struct lbs_private *priv = file->private_data;
+ char *buf;
+
+ buf = memdup_user_nul(userbuf, min(count, len - 1));
+ if (IS_ERR(buf))
+ return PTR_ERR(buf);
+
+ priv->rf_offset = simple_strtoul(buf, NULL, 16);
+ kfree(buf);
+ return count;
+}
+
+static ssize_t lbs_wrrf_write(struct file *file,
+ const char __user *userbuf,
+ size_t count, loff_t *ppos)
+{
+
+ struct lbs_private *priv = file->private_data;
+ ssize_t res;
+ u32 offset, value;
+ char *buf;
+
+ buf = memdup_user_nul(userbuf, min(count, len - 1));
+ if (IS_ERR(buf))
+ return PTR_ERR(buf);
+
+ res = sscanf(buf, "%x %x", &offset, &value);
+ if (res != 2) {
+ res = -EFAULT;
+ goto out_unlock;
+ }
+
+ res = lbs_set_reg(priv, CMD_RF_REG_ACCESS, offset, value);
+ mdelay(10);
+
+ if (!res)
+ res = count;
+out_unlock:
+ kfree(buf);
+ return res;
+}
+
+#define FOPS(fread, fwrite) { \
+ .owner = THIS_MODULE, \
+ .open = simple_open, \
+ .read = (fread), \
+ .write = (fwrite), \
+ .llseek = generic_file_llseek, \
+}
+
+struct lbs_debugfs_files {
+ const char *name;
+ umode_t perm;
+ struct file_operations fops;
+};
+
+static const struct lbs_debugfs_files debugfs_files[] = {
+ { "info", 0444, FOPS(lbs_dev_info, write_file_dummy), },
+ { "sleepparams", 0644, FOPS(lbs_sleepparams_read,
+ lbs_sleepparams_write), },
+ { "hostsleep", 0644, FOPS(lbs_host_sleep_read,
+ lbs_host_sleep_write), },
+};
+
+static const struct lbs_debugfs_files debugfs_events_files[] = {
+ {"low_rssi", 0644, FOPS(lbs_lowrssi_read,
+ lbs_lowrssi_write), },
+ {"low_snr", 0644, FOPS(lbs_lowsnr_read,
+ lbs_lowsnr_write), },
+ {"failure_count", 0644, FOPS(lbs_failcount_read,
+ lbs_failcount_write), },
+ {"beacon_missed", 0644, FOPS(lbs_bcnmiss_read,
+ lbs_bcnmiss_write), },
+ {"high_rssi", 0644, FOPS(lbs_highrssi_read,
+ lbs_highrssi_write), },
+ {"high_snr", 0644, FOPS(lbs_highsnr_read,
+ lbs_highsnr_write), },
+};
+
+static const struct lbs_debugfs_files debugfs_regs_files[] = {
+ {"rdmac", 0644, FOPS(lbs_rdmac_read, lbs_rdmac_write), },
+ {"wrmac", 0600, FOPS(NULL, lbs_wrmac_write), },
+ {"rdbbp", 0644, FOPS(lbs_rdbbp_read, lbs_rdbbp_write), },
+ {"wrbbp", 0600, FOPS(NULL, lbs_wrbbp_write), },
+ {"rdrf", 0644, FOPS(lbs_rdrf_read, lbs_rdrf_write), },
+ {"wrrf", 0600, FOPS(NULL, lbs_wrrf_write), },
+};
+
+void lbs_debugfs_init(void)
+{
+ if (!lbs_dir)
+ lbs_dir = debugfs_create_dir("lbs_wireless", NULL);
+}
+
+void lbs_debugfs_remove(void)
+{
+ debugfs_remove(lbs_dir);
+}
+
+void lbs_debugfs_init_one(struct lbs_private *priv, struct net_device *dev)
+{
+ int i;
+ const struct lbs_debugfs_files *files;
+ if (!lbs_dir)
+ goto exit;
+
+ priv->debugfs_dir = debugfs_create_dir(dev->name, lbs_dir);
+ if (!priv->debugfs_dir)
+ goto exit;
+
+ for (i=0; i<ARRAY_SIZE(debugfs_files); i++) {
+ files = &debugfs_files[i];
+ priv->debugfs_files[i] = debugfs_create_file(files->name,
+ files->perm,
+ priv->debugfs_dir,
+ priv,
+ &files->fops);
+ }
+
+ priv->events_dir = debugfs_create_dir("subscribed_events", priv->debugfs_dir);
+ if (!priv->events_dir)
+ goto exit;
+
+ for (i=0; i<ARRAY_SIZE(debugfs_events_files); i++) {
+ files = &debugfs_events_files[i];
+ priv->debugfs_events_files[i] = debugfs_create_file(files->name,
+ files->perm,
+ priv->events_dir,
+ priv,
+ &files->fops);
+ }
+
+ priv->regs_dir = debugfs_create_dir("registers", priv->debugfs_dir);
+ if (!priv->regs_dir)
+ goto exit;
+
+ for (i=0; i<ARRAY_SIZE(debugfs_regs_files); i++) {
+ files = &debugfs_regs_files[i];
+ priv->debugfs_regs_files[i] = debugfs_create_file(files->name,
+ files->perm,
+ priv->regs_dir,
+ priv,
+ &files->fops);
+ }
+
+#ifdef PROC_DEBUG
+ lbs_debug_init(priv);
+#endif
+exit:
+ return;
+}
+
+void lbs_debugfs_remove_one(struct lbs_private *priv)
+{
+ int i;
+
+ for(i=0; i<ARRAY_SIZE(debugfs_regs_files); i++)
+ debugfs_remove(priv->debugfs_regs_files[i]);
+
+ debugfs_remove(priv->regs_dir);
+
+ for(i=0; i<ARRAY_SIZE(debugfs_events_files); i++)
+ debugfs_remove(priv->debugfs_events_files[i]);
+
+ debugfs_remove(priv->events_dir);
+#ifdef PROC_DEBUG
+ debugfs_remove(priv->debugfs_debug);
+#endif
+ for(i=0; i<ARRAY_SIZE(debugfs_files); i++)
+ debugfs_remove(priv->debugfs_files[i]);
+ debugfs_remove(priv->debugfs_dir);
+}
+
+
+
+/* debug entry */
+
+#ifdef PROC_DEBUG
+
+#define item_size(n) (FIELD_SIZEOF(struct lbs_private, n))
+#define item_addr(n) (offsetof(struct lbs_private, n))
+
+
+struct debug_data {
+ char name[32];
+ u32 size;
+ size_t addr;
+};
+
+/* To debug any member of struct lbs_private, simply add one line here.
+ */
+static struct debug_data items[] = {
+ {"psmode", item_size(psmode), item_addr(psmode)},
+ {"psstate", item_size(psstate), item_addr(psstate)},
+};
+
+static int num_of_items = ARRAY_SIZE(items);
+
+/**
+ * lbs_debugfs_read - proc read function
+ *
+ * @file: file to read
+ * @userbuf: pointer to buffer
+ * @count: number of bytes to read
+ * @ppos: read data starting position
+ *
+ * returns: amount of data read or negative error code
+ */
+static ssize_t lbs_debugfs_read(struct file *file, char __user *userbuf,
+ size_t count, loff_t *ppos)
+{
+ int val = 0;
+ size_t pos = 0;
+ ssize_t res;
+ char *p;
+ int i;
+ struct debug_data *d;
+ unsigned long addr = get_zeroed_page(GFP_KERNEL);
+ char *buf = (char *)addr;
+ if (!buf)
+ return -ENOMEM;
+
+ p = buf;
+
+ d = file->private_data;
+
+ for (i = 0; i < num_of_items; i++) {
+ if (d[i].size == 1)
+ val = *((u8 *) d[i].addr);
+ else if (d[i].size == 2)
+ val = *((u16 *) d[i].addr);
+ else if (d[i].size == 4)
+ val = *((u32 *) d[i].addr);
+ else if (d[i].size == 8)
+ val = *((u64 *) d[i].addr);
+
+ pos += sprintf(p + pos, "%s=%d\n", d[i].name, val);
+ }
+
+ res = simple_read_from_buffer(userbuf, count, ppos, p, pos);
+
+ free_page(addr);
+ return res;
+}
+
+/**
+ * lbs_debugfs_write - proc write function
+ *
+ * @f: file pointer
+ * @buf: pointer to data buffer
+ * @cnt: data number to write
+ * @ppos: file position
+ *
+ * returns: amount of data written
+ */
+static ssize_t lbs_debugfs_write(struct file *f, const char __user *buf,
+ size_t cnt, loff_t *ppos)
+{
+ int r, i;
+ char *pdata;
+ char *p;
+ char *p0;
+ char *p1;
+ char *p2;
+ struct debug_data *d = f->private_data;
+
+ if (cnt == 0)
+ return 0;
+
+ pdata = memdup_user_nul(buf, cnt);
+ if (IS_ERR(pdata))
+ return PTR_ERR(pdata);
+
+ p0 = pdata;
+ for (i = 0; i < num_of_items; i++) {
+ do {
+ p = strstr(p0, d[i].name);
+ if (p == NULL)
+ break;
+ p1 = strchr(p, '\n');
+ if (p1 == NULL)
+ break;
+ p0 = p1++;
+ p2 = strchr(p, '=');
+ if (!p2)
+ break;
+ p2++;
+ r = simple_strtoul(p2, NULL, 0);
+ if (d[i].size == 1)
+ *((u8 *) d[i].addr) = (u8) r;
+ else if (d[i].size == 2)
+ *((u16 *) d[i].addr) = (u16) r;
+ else if (d[i].size == 4)
+ *((u32 *) d[i].addr) = (u32) r;
+ else if (d[i].size == 8)
+ *((u64 *) d[i].addr) = (u64) r;
+ break;
+ } while (1);
+ }
+ kfree(pdata);
+
+ return (ssize_t)cnt;
+}
+
+static const struct file_operations lbs_debug_fops = {
+ .owner = THIS_MODULE,
+ .open = simple_open,
+ .write = lbs_debugfs_write,
+ .read = lbs_debugfs_read,
+ .llseek = default_llseek,
+};
+
+/**
+ * lbs_debug_init - create debug proc file
+ *
+ * @priv: pointer to &struct lbs_private
+ *
+ * returns: N/A
+ */
+static void lbs_debug_init(struct lbs_private *priv)
+{
+ int i;
+
+ if (!priv->debugfs_dir)
+ return;
+
+ for (i = 0; i < num_of_items; i++)
+ items[i].addr += (size_t) priv;
+
+ priv->debugfs_debug = debugfs_create_file("debug", 0644,
+ priv->debugfs_dir, &items[0],
+ &lbs_debug_fops);
+}
+#endif