/* -*- linux-c -*- * * /proc transport and control * Copyright (C) 2005-2007 Red Hat Inc. * * This file is part of systemtap, and is free software. You can * redistribute it and/or modify it under the terms of the GNU General * Public License (GPL); either version 2, or (at your option) any * later version. */ #define STP_DEFAULT_BUFFERS 256 static int _stp_current_buffers = STP_DEFAULT_BUFFERS; static struct list_head _stp_ready_q; static struct list_head _stp_pool_q; spinlock_t _stp_pool_lock = SPIN_LOCK_UNLOCKED; spinlock_t _stp_ready_lock = SPIN_LOCK_UNLOCKED; #ifdef STP_RELAYFS extern int _stp_relay_flushing; /* handle the per-cpu subbuf info read for relayfs */ static ssize_t _stp_proc_read (struct file *file, char __user *buf, size_t count, loff_t *ppos) { int num; struct _stp_buf_info out; int cpu = *(int *)(PDE(file->f_dentry->d_inode)->data); if (!_stp_chan) return -EINVAL; out.cpu = cpu; #if (RELAYFS_CHANNEL_VERSION >= 4) || defined (CONFIG_RELAY) out.produced = _stp_chan->buf[cpu]->subbufs_produced; out.consumed = _stp_chan->buf[cpu]->subbufs_consumed; #else out.produced = atomic_read(&_stp_chan->buf[cpu]->subbufs_produced); out.consumed = atomic_read(&_stp_chan->buf[cpu]->subbufs_consumed); #endif /* RELAYFS_CHANNEL_VERSION >= 4 || CONFIG_RELAY */ out.flushing = _stp_relay_flushing; num = sizeof(out); if (copy_to_user(buf, &out, num)) return -EFAULT; return num; } /* handle the per-cpu subbuf info write for relayfs */ static ssize_t _stp_proc_write (struct file *file, const char __user *buf, size_t count, loff_t *ppos) { struct _stp_consumed_info info; int cpu = *(int *)(PDE(file->f_dentry->d_inode)->data); if (copy_from_user(&info, buf, count)) return -EFAULT; relay_subbufs_consumed(_stp_chan, cpu, info.consumed); return count; } static struct file_operations _stp_proc_fops = { .owner = THIS_MODULE, .read = _stp_proc_read, .write = _stp_proc_write, }; #endif static ssize_t _stp_proc_write_cmd (struct file *file, const char __user *buf, size_t count, loff_t *ppos) { int type; if (count < sizeof(int)) return 0; if (get_user(type, (int __user *)buf)) return -EFAULT; //printk ("_stp_proc_write_cmd. count:%d type:%d\n", count, type); if (type == STP_SYMBOLS) { count -= sizeof(long); buf += sizeof(long); } else { count -= sizeof(int); buf += sizeof(int); } switch (type) { case STP_START: { struct _stp_transport_start st; if (count < sizeof(struct _stp_transport_start)) return 0; if (copy_from_user (&st, buf, sizeof(struct _stp_transport_start))) return -EFAULT; _stp_handle_start (&st); break; } case STP_SYMBOLS: count = _stp_do_symbols(buf, count); break; case STP_MODULE: count = _stp_do_module(buf, count); break; case STP_EXIT: _stp_exit_flag = 1; break; case STP_TRANSPORT_INFO: { struct _stp_transport_info ti; kbug("STP_TRANSPORT_INFO %d %d\n", (int)count, (int)sizeof(struct _stp_transport_info)); if (count < sizeof(struct _stp_transport_info)) return 0; if (copy_from_user (&ti, buf, sizeof(struct _stp_transport_info))) return -EFAULT; if (_stp_transport_open (&ti) < 0) return -1; break; } default: printk ("invalid command type %d\n", type); return -EINVAL; } return count; } struct _stp_buffer { struct list_head list; int len; int type; char buf[STP_BUFFER_SIZE]; }; static DECLARE_WAIT_QUEUE_HEAD(_stp_proc_wq); static int _stp_write (int type, void *data, int len) { struct _stp_buffer *bptr; unsigned long flags; unsigned numtrylock; #define WRITE_AGG #ifdef WRITE_AGG numtrylock = 0; while (!spin_trylock_irqsave (&_stp_ready_lock, flags) && (++numtrylock < MAXTRYLOCK)) ndelay (TRYLOCKDELAY); if (unlikely (numtrylock >= MAXTRYLOCK)) return 0; if (!list_empty(&_stp_ready_q)) { bptr = (struct _stp_buffer *)_stp_ready_q.prev; if (bptr->len + len <= STP_BUFFER_SIZE && type == STP_REALTIME_DATA && bptr->type == STP_REALTIME_DATA) { memcpy (bptr->buf + bptr->len, data, len); bptr->len += len; spin_unlock_irqrestore(&_stp_ready_lock, flags); return len; } } spin_unlock_irqrestore(&_stp_ready_lock, flags); #endif numtrylock = 0; while (!spin_trylock_irqsave (&_stp_pool_lock, flags) && (++numtrylock < MAXTRYLOCK)) ndelay (TRYLOCKDELAY); if (unlikely (numtrylock >= MAXTRYLOCK)) return 0; if (list_empty(&_stp_pool_q)) { spin_unlock_irqrestore(&_stp_pool_lock, flags); return -1; } /* get the next buffer from the pool */ bptr = (struct _stp_buffer *)_stp_pool_q.next; list_del_init(&bptr->list); spin_unlock_irqrestore(&_stp_pool_lock, flags); bptr->type = type; memcpy (bptr->buf, data, len); bptr->len = len; /* put it on the pool of ready buffers */ numtrylock = 0; while (!spin_trylock_irqsave (&_stp_ready_lock, flags) && (++numtrylock < MAXTRYLOCK)) ndelay (TRYLOCKDELAY); if (unlikely (numtrylock >= MAXTRYLOCK)) return 0; list_add_tail(&bptr->list, &_stp_ready_q); spin_unlock_irqrestore(&_stp_ready_lock, flags); return len; } static ssize_t _stp_proc_read_cmd (struct file *file, char __user *buf, size_t count, loff_t *ppos) { struct _stp_buffer *bptr; int len; unsigned long flags; /* FIXME FIXME FIXME. assuming count is large enough to hold buffer!! */ /* wait for nonempty ready queue */ spin_lock_irqsave(&_stp_ready_lock, flags); while (list_empty(&_stp_ready_q)) { spin_unlock_irqrestore(&_stp_ready_lock, flags); if (file->f_flags & O_NONBLOCK) return -EAGAIN; if (wait_event_interruptible(_stp_proc_wq, !list_empty(&_stp_ready_q))) return -ERESTARTSYS; spin_lock_irqsave(&_stp_ready_lock, flags); } /* get the next buffer off the ready list */ bptr = (struct _stp_buffer *)_stp_ready_q.next; list_del_init(&bptr->list); spin_unlock_irqrestore(&_stp_ready_lock, flags); /* write it out */ len = bptr->len + 4; if (copy_to_user(buf, &bptr->type, len)) { /* now what? We took it off the queue then failed to send it */ /* we can't put it back on the queue because it will likely be out-of-order */ /* fortunately this should never happen */ /* FIXME need to mark this as a transport failure */ return -EFAULT; } /* put it on the pool of free buffers */ spin_lock_irqsave(&_stp_pool_lock, flags); list_add_tail(&bptr->list, &_stp_pool_q); spin_unlock_irqrestore(&_stp_pool_lock, flags); return len; } static struct file_operations _stp_proc_fops_cmd = { .owner = THIS_MODULE, .read = _stp_proc_read_cmd, .write = _stp_proc_write_cmd, // .poll = _stp_proc_poll_cmd }; static struct proc_dir_entry *_stp_proc_root, *_stp_proc_mod; /* copy since proc_match is not MODULE_EXPORT'd */ static int my_proc_match(int len, const char *name, struct proc_dir_entry *de) { if (de->namelen != len) return 0; return !memcmp(name, de->name, len); } /* set the number of buffers to use to 'num' */ static int _stp_set_buffers(int num) { int i; struct list_head *p; unsigned long flags; //printk("stp_set_buffers %d\n", num); if (num == 0 || num == _stp_current_buffers) return _stp_current_buffers; if (num > _stp_current_buffers) { for (i = 0; i < num - _stp_current_buffers; i++) { p = (struct list_head *)kmalloc(sizeof(struct _stp_buffer),STP_ALLOC_FLAGS); if (!p) { _stp_current_buffers += i; goto err; } _stp_allocated_net_memory += sizeof(struct _stp_buffer); spin_lock_irqsave(&_stp_pool_lock, flags); list_add (p, &_stp_pool_q); spin_unlock_irqrestore(&_stp_pool_lock, flags); } } else { for (i = 0; i < _stp_current_buffers - num; i++) { spin_lock_irqsave(&_stp_pool_lock, flags); p = _stp_pool_q.next; list_del(p); spin_unlock_irqrestore(&_stp_pool_lock, flags); kfree(p); } } _stp_current_buffers = num; err: return _stp_current_buffers; } static int _stp_register_procfs (void) { int i; #ifdef STP_RELAYFS int j; char buf[8]; #endif struct proc_dir_entry *de; struct list_head *p, *tmp; INIT_LIST_HEAD(&_stp_ready_q); INIT_LIST_HEAD(&_stp_pool_q); /* allocate buffers */ for (i = 0; i < STP_DEFAULT_BUFFERS; i++) { p = (struct list_head *)kmalloc(sizeof(struct _stp_buffer),STP_ALLOC_FLAGS); // printk("allocated buffer at %lx\n", (long)p); if (!p) goto err0; _stp_allocated_net_memory += sizeof(struct _stp_buffer); list_add (p, &_stp_pool_q); } /* Formerly, we allocated /proc/systemtap, but unfortunately that's racy with multiple concurrent probes. So now we set _stp_proc_root to proc_root. This way, /proc/stap_XXXX rather than /proc/systemtap/stap_XXXX will be the directory under which cmd/ etc. will show up. */ _stp_proc_root = NULL; /* now create /proc/systemtap/module_name */ _stp_proc_mod = proc_mkdir (THIS_MODULE->name, _stp_proc_root); if (_stp_proc_mod == NULL) goto err0; #ifdef STP_RELAYFS /* now for each cpu "n", create /proc/systemtap/module_name/n */ for_each_cpu(i) { sprintf(buf, "%d", i); de = create_proc_entry (buf, S_IFREG|S_IRUSR, _stp_proc_mod); if (de == NULL) goto err1; de->proc_fops = &_stp_proc_fops; de->data = _stp_kmalloc(sizeof(int)); if (de->data == NULL) { remove_proc_entry (buf, _stp_proc_mod); goto err1; } *(int *)de->data = i; } #endif /* finally create /proc/systemtap/module_name/cmd */ de = create_proc_entry ("cmd", S_IFREG|S_IRUSR, _stp_proc_mod); if (de == NULL) goto err1; de->proc_fops = &_stp_proc_fops_cmd; return 0; err1: #ifdef STP_RELAYFS for (de = _stp_proc_mod->subdir; de; de = de->next) kfree (de->data); for_each_cpu(j) { if (j == i) break; sprintf(buf, "%d", i); remove_proc_entry (buf, _stp_proc_mod); } #endif remove_proc_entry (THIS_MODULE->name, _stp_proc_root); err0: list_for_each_safe(p, tmp, &_stp_pool_q) { list_del(p); kfree(p); } printk (KERN_ERR "Error creating systemtap /proc entries.\n"); return -1; } static void _stp_unregister_procfs (void) { struct list_head *p, *tmp; #ifdef STP_RELAYFS int i; char buf[8]; struct proc_dir_entry *de; for (de = _stp_proc_mod->subdir; de; de = de->next) kfree (de->data); for_each_cpu(i) { sprintf(buf, "%d", i); remove_proc_entry (buf, _stp_proc_mod); } #endif remove_proc_entry ("cmd", _stp_proc_mod); remove_proc_entry (THIS_MODULE->name, _stp_proc_root); /* free memory pools */ list_for_each_safe(p, tmp, &_stp_pool_q) { list_del(p); kfree(p); } list_for_each_safe(p, tmp, &_stp_ready_q) { list_del(p); kfree(p); } }