From dd3140588d9551235ebc2a0dacdca098e7677573 Mon Sep 17 00:00:00 2001 From: Daniel De Graaf Date: Mon, 7 Feb 2011 17:23:05 -0500 Subject: xen-gntalloc: Userspace grant allocation driver This allows a userspace application to allocate a shared page for implementing inter-domain communication or device drivers. These shared pages can be mapped using the gntdev device or by the kernel in another domain. Signed-off-by: Daniel De Graaf Signed-off-by: Konrad Rzeszutek Wilk --- include/xen/gntalloc.h | 50 ++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 50 insertions(+) create mode 100644 include/xen/gntalloc.h (limited to 'include/xen') diff --git a/include/xen/gntalloc.h b/include/xen/gntalloc.h new file mode 100644 index 000000000000..bc3b85e8bff7 --- /dev/null +++ b/include/xen/gntalloc.h @@ -0,0 +1,50 @@ +/****************************************************************************** + * gntalloc.h + * + * Interface to /dev/xen/gntalloc. + * + * Author: Daniel De Graaf + * + * This file is in the public domain. + */ + +#ifndef __LINUX_PUBLIC_GNTALLOC_H__ +#define __LINUX_PUBLIC_GNTALLOC_H__ + +/* + * Allocates a new page and creates a new grant reference. + */ +#define IOCTL_GNTALLOC_ALLOC_GREF \ +_IOC(_IOC_NONE, 'G', 5, sizeof(struct ioctl_gntalloc_alloc_gref)) +struct ioctl_gntalloc_alloc_gref { + /* IN parameters */ + /* The ID of the domain to be given access to the grants. */ + uint16_t domid; + /* Flags for this mapping */ + uint16_t flags; + /* Number of pages to map */ + uint32_t count; + /* OUT parameters */ + /* The offset to be used on a subsequent call to mmap(). */ + uint64_t index; + /* The grant references of the newly created grant, one per page */ + /* Variable size, depending on count */ + uint32_t gref_ids[1]; +}; + +#define GNTALLOC_FLAG_WRITABLE 1 + +/* + * Deallocates the grant reference, allowing the associated page to be freed if + * no other domains are using it. + */ +#define IOCTL_GNTALLOC_DEALLOC_GREF \ +_IOC(_IOC_NONE, 'G', 6, sizeof(struct ioctl_gntalloc_dealloc_gref)) +struct ioctl_gntalloc_dealloc_gref { + /* IN parameters */ + /* The offset returned in the map operation */ + uint64_t index; + /* Number of references to unmap */ + uint32_t count; +}; +#endif /* __LINUX_PUBLIC_GNTALLOC_H__ */ -- cgit v1.2.3 From bdc612dc6903c4ea06e40d02f84ad5e25d93459d Mon Sep 17 00:00:00 2001 From: Daniel De Graaf Date: Thu, 3 Feb 2011 12:19:04 -0500 Subject: xen/gntalloc,gntdev: Add unmap notify ioctl This ioctl allows the users of a shared page to be notified when the other end exits abnormally. [v2: updated description in structs] Signed-off-by: Daniel De Graaf Signed-off-by: Konrad Rzeszutek Wilk --- drivers/xen/gntalloc.c | 59 ++++++++++++++++++++++++++++++++++++++++++++++++ drivers/xen/gntdev.c | 61 +++++++++++++++++++++++++++++++++++++++++++++++++- include/xen/gntalloc.h | 32 ++++++++++++++++++++++++++ include/xen/gntdev.h | 31 +++++++++++++++++++++++++ 4 files changed, 182 insertions(+), 1 deletion(-) (limited to 'include/xen') diff --git a/drivers/xen/gntalloc.c b/drivers/xen/gntalloc.c index d06bf2b4cd07..a7ffdfe19fc9 100644 --- a/drivers/xen/gntalloc.c +++ b/drivers/xen/gntalloc.c @@ -60,11 +60,13 @@ #include #include #include +#include #include #include #include #include +#include static int limit = 1024; module_param(limit, int, 0644); @@ -75,6 +77,12 @@ static LIST_HEAD(gref_list); static DEFINE_SPINLOCK(gref_lock); static int gref_size; +struct notify_info { + uint16_t pgoff:12; /* Bits 0-11: Offset of the byte to clear */ + uint16_t flags:2; /* Bits 12-13: Unmap notification flags */ + int event; /* Port (event channel) to notify */ +}; + /* Metadata on a grant reference. */ struct gntalloc_gref { struct list_head next_gref; /* list entry gref_list */ @@ -83,6 +91,7 @@ struct gntalloc_gref { uint64_t file_index; /* File offset for mmap() */ unsigned int users; /* Use count - when zero, waiting on Xen */ grant_ref_t gref_id; /* The grant reference number */ + struct notify_info notify; /* Unmap notification */ }; struct gntalloc_file_private_data { @@ -164,6 +173,16 @@ undo: static void __del_gref(struct gntalloc_gref *gref) { + if (gref->notify.flags & UNMAP_NOTIFY_CLEAR_BYTE) { + uint8_t *tmp = kmap(gref->page); + tmp[gref->notify.pgoff] = 0; + kunmap(gref->page); + } + if (gref->notify.flags & UNMAP_NOTIFY_SEND_EVENT) + notify_remote_via_evtchn(gref->notify.event); + + gref->notify.flags = 0; + if (gref->gref_id > 0) { if (gnttab_query_foreign_access(gref->gref_id)) return; @@ -349,6 +368,43 @@ dealloc_grant_out: return rc; } +static long gntalloc_ioctl_unmap_notify(struct gntalloc_file_private_data *priv, + void __user *arg) +{ + struct ioctl_gntalloc_unmap_notify op; + struct gntalloc_gref *gref; + uint64_t index; + int pgoff; + int rc; + + if (copy_from_user(&op, arg, sizeof(op))) + return -EFAULT; + + index = op.index & ~(PAGE_SIZE - 1); + pgoff = op.index & (PAGE_SIZE - 1); + + spin_lock(&gref_lock); + + gref = find_grefs(priv, index, 1); + if (!gref) { + rc = -ENOENT; + goto unlock_out; + } + + if (op.action & ~(UNMAP_NOTIFY_CLEAR_BYTE|UNMAP_NOTIFY_SEND_EVENT)) { + rc = -EINVAL; + goto unlock_out; + } + + gref->notify.flags = op.action; + gref->notify.pgoff = pgoff; + gref->notify.event = op.event_channel_port; + rc = 0; + unlock_out: + spin_unlock(&gref_lock); + return rc; +} + static long gntalloc_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) { @@ -361,6 +417,9 @@ static long gntalloc_ioctl(struct file *filp, unsigned int cmd, case IOCTL_GNTALLOC_DEALLOC_GREF: return gntalloc_ioctl_dealloc(priv, (void __user *)arg); + case IOCTL_GNTALLOC_SET_UNMAP_NOTIFY: + return gntalloc_ioctl_unmap_notify(priv, (void __user *)arg); + default: return -ENOIOCTLCMD; } diff --git a/drivers/xen/gntdev.c b/drivers/xen/gntdev.c index bcaf797216d1..9694a1a8b2e2 100644 --- a/drivers/xen/gntdev.c +++ b/drivers/xen/gntdev.c @@ -37,6 +37,7 @@ #include #include #include +#include #include #include #include @@ -63,6 +64,13 @@ struct gntdev_priv { struct mmu_notifier mn; }; +struct unmap_notify { + int flags; + /* Address relative to the start of the grant_map */ + int addr; + int event; +}; + struct grant_map { struct list_head next; struct vm_area_struct *vma; @@ -71,6 +79,7 @@ struct grant_map { int flags; int is_mapped; atomic_t users; + struct unmap_notify notify; struct ioctl_gntdev_grant_ref *grants; struct gnttab_map_grant_ref *map_ops; struct gnttab_unmap_grant_ref *unmap_ops; @@ -165,7 +174,7 @@ static struct grant_map *gntdev_find_map_index(struct gntdev_priv *priv, list_for_each_entry(map, &priv->maps, next) { if (map->index != index) continue; - if (map->count != count) + if (count && map->count != count) continue; return map; } @@ -184,6 +193,10 @@ static void gntdev_put_map(struct grant_map *map) atomic_sub(map->count, &pages_mapped); + if (map->notify.flags & UNMAP_NOTIFY_SEND_EVENT) { + notify_remote_via_evtchn(map->notify.event); + } + if (map->pages) { if (!use_ptemod) unmap_grant_pages(map, 0, map->count); @@ -274,6 +287,16 @@ static int unmap_grant_pages(struct grant_map *map, int offset, int pages) { int i, err = 0; + if (map->notify.flags & UNMAP_NOTIFY_CLEAR_BYTE) { + int pgno = (map->notify.addr >> PAGE_SHIFT); + if (pgno >= offset && pgno < offset + pages) { + uint8_t *tmp = kmap(map->pages[pgno]); + tmp[map->notify.addr & (PAGE_SIZE-1)] = 0; + kunmap(map->pages[pgno]); + map->notify.flags &= ~UNMAP_NOTIFY_CLEAR_BYTE; + } + } + pr_debug("map %d+%d [%d+%d]\n", map->index, map->count, offset, pages); err = gnttab_unmap_refs(map->unmap_ops + offset, map->pages, pages); if (err) @@ -519,6 +542,39 @@ static long gntdev_ioctl_get_offset_for_vaddr(struct gntdev_priv *priv, return 0; } +static long gntdev_ioctl_notify(struct gntdev_priv *priv, void __user *u) +{ + struct ioctl_gntdev_unmap_notify op; + struct grant_map *map; + int rc; + + if (copy_from_user(&op, u, sizeof(op))) + return -EFAULT; + + if (op.action & ~(UNMAP_NOTIFY_CLEAR_BYTE|UNMAP_NOTIFY_SEND_EVENT)) + return -EINVAL; + + spin_lock(&priv->lock); + + list_for_each_entry(map, &priv->maps, next) { + uint64_t begin = map->index << PAGE_SHIFT; + uint64_t end = (map->index + map->count) << PAGE_SHIFT; + if (op.index >= begin && op.index < end) + goto found; + } + rc = -ENOENT; + goto unlock_out; + + found: + map->notify.flags = op.action; + map->notify.addr = op.index - (map->index << PAGE_SHIFT); + map->notify.event = op.event_channel_port; + rc = 0; + unlock_out: + spin_unlock(&priv->lock); + return rc; +} + static long gntdev_ioctl(struct file *flip, unsigned int cmd, unsigned long arg) { @@ -535,6 +591,9 @@ static long gntdev_ioctl(struct file *flip, case IOCTL_GNTDEV_GET_OFFSET_FOR_VADDR: return gntdev_ioctl_get_offset_for_vaddr(priv, ptr); + case IOCTL_GNTDEV_SET_UNMAP_NOTIFY: + return gntdev_ioctl_notify(priv, ptr); + default: pr_debug("priv %p, unknown cmd %x\n", priv, cmd); return -ENOIOCTLCMD; diff --git a/include/xen/gntalloc.h b/include/xen/gntalloc.h index bc3b85e8bff7..76bd58065f4f 100644 --- a/include/xen/gntalloc.h +++ b/include/xen/gntalloc.h @@ -47,4 +47,36 @@ struct ioctl_gntalloc_dealloc_gref { /* Number of references to unmap */ uint32_t count; }; + +/* + * Sets up an unmap notification within the page, so that the other side can do + * cleanup if this side crashes. Required to implement cross-domain robust + * mutexes or close notification on communication channels. + * + * Each mapped page only supports one notification; multiple calls referring to + * the same page overwrite the previous notification. You must clear the + * notification prior to the IOCTL_GNTALLOC_DEALLOC_GREF if you do not want it + * to occur. + */ +#define IOCTL_GNTALLOC_SET_UNMAP_NOTIFY \ +_IOC(_IOC_NONE, 'G', 7, sizeof(struct ioctl_gntalloc_unmap_notify)) +struct ioctl_gntalloc_unmap_notify { + /* IN parameters */ + /* Offset in the file descriptor for a byte within the page (same as + * used in mmap). If using UNMAP_NOTIFY_CLEAR_BYTE, this is the byte to + * be cleared. Otherwise, it can be any byte in the page whose + * notification we are adjusting. + */ + uint64_t index; + /* Action(s) to take on unmap */ + uint32_t action; + /* Event channel to notify */ + uint32_t event_channel_port; +}; + +/* Clear (set to zero) the byte specified by index */ +#define UNMAP_NOTIFY_CLEAR_BYTE 0x1 +/* Send an interrupt on the indicated event channel */ +#define UNMAP_NOTIFY_SEND_EVENT 0x2 + #endif /* __LINUX_PUBLIC_GNTALLOC_H__ */ diff --git a/include/xen/gntdev.h b/include/xen/gntdev.h index eb23f4188f5a..5304bd3c84c5 100644 --- a/include/xen/gntdev.h +++ b/include/xen/gntdev.h @@ -116,4 +116,35 @@ struct ioctl_gntdev_set_max_grants { uint32_t count; }; +/* + * Sets up an unmap notification within the page, so that the other side can do + * cleanup if this side crashes. Required to implement cross-domain robust + * mutexes or close notification on communication channels. + * + * Each mapped page only supports one notification; multiple calls referring to + * the same page overwrite the previous notification. You must clear the + * notification prior to the IOCTL_GNTALLOC_DEALLOC_GREF if you do not want it + * to occur. + */ +#define IOCTL_GNTDEV_SET_UNMAP_NOTIFY \ +_IOC(_IOC_NONE, 'G', 7, sizeof(struct ioctl_gntdev_unmap_notify)) +struct ioctl_gntdev_unmap_notify { + /* IN parameters */ + /* Offset in the file descriptor for a byte within the page (same as + * used in mmap). If using UNMAP_NOTIFY_CLEAR_BYTE, this is the byte to + * be cleared. Otherwise, it can be any byte in the page whose + * notification we are adjusting. + */ + uint64_t index; + /* Action(s) to take on unmap */ + uint32_t action; + /* Event channel to notify */ + uint32_t event_channel_port; +}; + +/* Clear (set to zero) the byte specified by index */ +#define UNMAP_NOTIFY_CLEAR_BYTE 0x1 +/* Send an interrupt on the indicated event channel */ +#define UNMAP_NOTIFY_SEND_EVENT 0x2 + #endif /* __LINUX_PUBLIC_GNTDEV_H__ */ -- cgit v1.2.3 From 803eb047a28d239809fff1f87274cdaa94e0d8ea Mon Sep 17 00:00:00 2001 From: Daniel De Graaf Date: Mon, 14 Mar 2011 11:29:37 -0400 Subject: xen-balloon: Move core balloon functionality out of module The basic functionality of ballooning pages is useful for Xen drivers in general. Rather than require a dependency on the balloon module, split the functionality that is reused into the core. The balloon module is still required to follow ballooning requests from xenstore or to view balloon statistics in sysfs. Acked-by: Ian Campbell Signed-off-by: Daniel De Graaf Signed-off-by: Konrad Rzeszutek Wilk --- drivers/xen/Makefile | 4 +- drivers/xen/balloon.c | 229 ++--------------------------------------- drivers/xen/xen-balloon.c | 256 ++++++++++++++++++++++++++++++++++++++++++++++ include/xen/balloon.h | 22 ++++ 4 files changed, 286 insertions(+), 225 deletions(-) create mode 100644 drivers/xen/xen-balloon.c create mode 100644 include/xen/balloon.h (limited to 'include/xen') diff --git a/drivers/xen/Makefile b/drivers/xen/Makefile index 9585a1da52c6..f420f1ff7f13 100644 --- a/drivers/xen/Makefile +++ b/drivers/xen/Makefile @@ -1,4 +1,4 @@ -obj-y += grant-table.o features.o events.o manage.o +obj-y += grant-table.o features.o events.o manage.o balloon.o obj-y += xenbus/ nostackp := $(call cc-option, -fno-stack-protector) @@ -7,7 +7,7 @@ CFLAGS_features.o := $(nostackp) obj-$(CONFIG_BLOCK) += biomerge.o obj-$(CONFIG_HOTPLUG_CPU) += cpu_hotplug.o obj-$(CONFIG_XEN_XENCOMM) += xencomm.o -obj-$(CONFIG_XEN_BALLOON) += balloon.o +obj-$(CONFIG_XEN_BALLOON) += xen-balloon.o obj-$(CONFIG_XEN_DEV_EVTCHN) += xen-evtchn.o obj-$(CONFIG_XEN_GNTDEV) += xen-gntdev.o obj-$(CONFIG_XEN_GRANT_DEV_ALLOC) += xen-gntalloc.o diff --git a/drivers/xen/balloon.c b/drivers/xen/balloon.c index 77e6ad38c0b9..7497041d0631 100644 --- a/drivers/xen/balloon.c +++ b/drivers/xen/balloon.c @@ -1,6 +1,4 @@ /****************************************************************************** - * balloon.c - * * Xen balloon driver - enables returning/claiming memory to/from Xen. * * Copyright (c) 2003, B Dragovic @@ -33,7 +31,6 @@ */ #include -#include #include #include #include @@ -42,13 +39,11 @@ #include #include #include -#include #include #include #include #include -#include #include #include @@ -58,14 +53,10 @@ #include #include #include -#include +#include #include #include -#define PAGES2KB(_p) ((_p)<<(PAGE_SHIFT-10)) - -#define BALLOON_CLASS_NAME "xen_memory" - /* * balloon_process() state: * @@ -80,28 +71,11 @@ enum bp_state { BP_ECANCELED }; -#define RETRY_UNLIMITED 0 - -struct balloon_stats { - /* We aim for 'current allocation' == 'target allocation'. */ - unsigned long current_pages; - unsigned long target_pages; - /* Number of pages in high- and low-memory balloons. */ - unsigned long balloon_low; - unsigned long balloon_high; - unsigned long schedule_delay; - unsigned long max_schedule_delay; - unsigned long retry_count; - unsigned long max_retry_count; -}; static DEFINE_MUTEX(balloon_mutex); -static struct sys_device balloon_sysdev; - -static int register_balloon(struct sys_device *sysdev); - -static struct balloon_stats balloon_stats; +struct balloon_stats balloon_stats; +EXPORT_SYMBOL_GPL(balloon_stats); /* We increase/decrease in batches which fit in a page */ static unsigned long frame_list[PAGE_SIZE / sizeof(unsigned long)]; @@ -384,51 +358,13 @@ static void balloon_process(struct work_struct *work) } /* Resets the Xen limit, sets new target, and kicks off processing. */ -static void balloon_set_new_target(unsigned long target) +void balloon_set_new_target(unsigned long target) { /* No need for lock. Not read-modify-write updates. */ balloon_stats.target_pages = target; schedule_delayed_work(&balloon_worker, 0); } - -static struct xenbus_watch target_watch = -{ - .node = "memory/target" -}; - -/* React to a change in the target key */ -static void watch_target(struct xenbus_watch *watch, - const char **vec, unsigned int len) -{ - unsigned long long new_target; - int err; - - err = xenbus_scanf(XBT_NIL, "memory", "target", "%llu", &new_target); - if (err != 1) { - /* This is ok (for domain0 at least) - so just return */ - return; - } - - /* The given memory/target value is in KiB, so it needs converting to - * pages. PAGE_SHIFT converts bytes to pages, hence PAGE_SHIFT - 10. - */ - balloon_set_new_target(new_target >> (PAGE_SHIFT - 10)); -} - -static int balloon_init_watcher(struct notifier_block *notifier, - unsigned long event, - void *data) -{ - int err; - - err = register_xenbus_watch(&target_watch); - if (err) - printk(KERN_ERR "Failed to set balloon watcher\n"); - - return NOTIFY_DONE; -} - -static struct notifier_block xenstore_notifier; +EXPORT_SYMBOL_GPL(balloon_set_new_target); static int __init balloon_init(void) { @@ -438,7 +374,7 @@ static int __init balloon_init(void) if (!xen_pv_domain()) return -ENODEV; - pr_info("xen_balloon: Initialising balloon driver.\n"); + pr_info("xen/balloon: Initialising balloon driver.\n"); balloon_stats.current_pages = min(xen_start_info->nr_pages, max_pfn); balloon_stats.target_pages = balloon_stats.current_pages; @@ -450,8 +386,6 @@ static int __init balloon_init(void) balloon_stats.retry_count = 1; balloon_stats.max_retry_count = RETRY_UNLIMITED; - register_balloon(&balloon_sysdev); - /* * Initialise the balloon with excess memory space. We need * to make sure we don't add memory which doesn't exist or @@ -472,160 +406,9 @@ static int __init balloon_init(void) __balloon_append(page); } - target_watch.callback = watch_target; - xenstore_notifier.notifier_call = balloon_init_watcher; - - register_xenstore_notifier(&xenstore_notifier); - return 0; } subsys_initcall(balloon_init); -static void balloon_exit(void) -{ - /* XXX - release balloon here */ - return; -} - -module_exit(balloon_exit); - -#define BALLOON_SHOW(name, format, args...) \ - static ssize_t show_##name(struct sys_device *dev, \ - struct sysdev_attribute *attr, \ - char *buf) \ - { \ - return sprintf(buf, format, ##args); \ - } \ - static SYSDEV_ATTR(name, S_IRUGO, show_##name, NULL) - -BALLOON_SHOW(current_kb, "%lu\n", PAGES2KB(balloon_stats.current_pages)); -BALLOON_SHOW(low_kb, "%lu\n", PAGES2KB(balloon_stats.balloon_low)); -BALLOON_SHOW(high_kb, "%lu\n", PAGES2KB(balloon_stats.balloon_high)); - -static SYSDEV_ULONG_ATTR(schedule_delay, 0444, balloon_stats.schedule_delay); -static SYSDEV_ULONG_ATTR(max_schedule_delay, 0644, balloon_stats.max_schedule_delay); -static SYSDEV_ULONG_ATTR(retry_count, 0444, balloon_stats.retry_count); -static SYSDEV_ULONG_ATTR(max_retry_count, 0644, balloon_stats.max_retry_count); - -static ssize_t show_target_kb(struct sys_device *dev, struct sysdev_attribute *attr, - char *buf) -{ - return sprintf(buf, "%lu\n", PAGES2KB(balloon_stats.target_pages)); -} - -static ssize_t store_target_kb(struct sys_device *dev, - struct sysdev_attribute *attr, - const char *buf, - size_t count) -{ - char *endchar; - unsigned long long target_bytes; - - if (!capable(CAP_SYS_ADMIN)) - return -EPERM; - - target_bytes = simple_strtoull(buf, &endchar, 0) * 1024; - - balloon_set_new_target(target_bytes >> PAGE_SHIFT); - - return count; -} - -static SYSDEV_ATTR(target_kb, S_IRUGO | S_IWUSR, - show_target_kb, store_target_kb); - - -static ssize_t show_target(struct sys_device *dev, struct sysdev_attribute *attr, - char *buf) -{ - return sprintf(buf, "%llu\n", - (unsigned long long)balloon_stats.target_pages - << PAGE_SHIFT); -} - -static ssize_t store_target(struct sys_device *dev, - struct sysdev_attribute *attr, - const char *buf, - size_t count) -{ - char *endchar; - unsigned long long target_bytes; - - if (!capable(CAP_SYS_ADMIN)) - return -EPERM; - - target_bytes = memparse(buf, &endchar); - - balloon_set_new_target(target_bytes >> PAGE_SHIFT); - - return count; -} - -static SYSDEV_ATTR(target, S_IRUGO | S_IWUSR, - show_target, store_target); - - -static struct sysdev_attribute *balloon_attrs[] = { - &attr_target_kb, - &attr_target, - &attr_schedule_delay.attr, - &attr_max_schedule_delay.attr, - &attr_retry_count.attr, - &attr_max_retry_count.attr -}; - -static struct attribute *balloon_info_attrs[] = { - &attr_current_kb.attr, - &attr_low_kb.attr, - &attr_high_kb.attr, - NULL -}; - -static struct attribute_group balloon_info_group = { - .name = "info", - .attrs = balloon_info_attrs, -}; - -static struct sysdev_class balloon_sysdev_class = { - .name = BALLOON_CLASS_NAME, -}; - -static int register_balloon(struct sys_device *sysdev) -{ - int i, error; - - error = sysdev_class_register(&balloon_sysdev_class); - if (error) - return error; - - sysdev->id = 0; - sysdev->cls = &balloon_sysdev_class; - - error = sysdev_register(sysdev); - if (error) { - sysdev_class_unregister(&balloon_sysdev_class); - return error; - } - - for (i = 0; i < ARRAY_SIZE(balloon_attrs); i++) { - error = sysdev_create_file(sysdev, balloon_attrs[i]); - if (error) - goto fail; - } - - error = sysfs_create_group(&sysdev->kobj, &balloon_info_group); - if (error) - goto fail; - - return 0; - - fail: - while (--i >= 0) - sysdev_remove_file(sysdev, balloon_attrs[i]); - sysdev_unregister(sysdev); - sysdev_class_unregister(&balloon_sysdev_class); - return error; -} - MODULE_LICENSE("GPL"); diff --git a/drivers/xen/xen-balloon.c b/drivers/xen/xen-balloon.c new file mode 100644 index 000000000000..a4ff225ee868 --- /dev/null +++ b/drivers/xen/xen-balloon.c @@ -0,0 +1,256 @@ +/****************************************************************************** + * Xen balloon driver - enables returning/claiming memory to/from Xen. + * + * Copyright (c) 2003, B Dragovic + * Copyright (c) 2003-2004, M Williamson, K Fraser + * Copyright (c) 2005 Dan M. Smith, IBM Corporation + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License version 2 + * as published by the Free Software Foundation; or, when distributed + * separately from the Linux kernel or incorporated into other + * software packages, subject to the following license: + * + * Permission is hereby granted, free of charge, to any person obtaining a copy + * of this source file (the "Software"), to deal in the Software without + * restriction, including without limitation the rights to use, copy, modify, + * merge, publish, distribute, sublicense, and/or sell copies of the Software, + * and to permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE + * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS + * IN THE SOFTWARE. + */ + +#include +#include +#include +#include + +#include +#include +#include +#include +#include +#include + +#define PAGES2KB(_p) ((_p)<<(PAGE_SHIFT-10)) + +#define BALLOON_CLASS_NAME "xen_memory" + +static struct sys_device balloon_sysdev; + +static int register_balloon(struct sys_device *sysdev); + +static struct xenbus_watch target_watch = +{ + .node = "memory/target" +}; + +/* React to a change in the target key */ +static void watch_target(struct xenbus_watch *watch, + const char **vec, unsigned int len) +{ + unsigned long long new_target; + int err; + + err = xenbus_scanf(XBT_NIL, "memory", "target", "%llu", &new_target); + if (err != 1) { + /* This is ok (for domain0 at least) - so just return */ + return; + } + + /* The given memory/target value is in KiB, so it needs converting to + * pages. PAGE_SHIFT converts bytes to pages, hence PAGE_SHIFT - 10. + */ + balloon_set_new_target(new_target >> (PAGE_SHIFT - 10)); +} + +static int balloon_init_watcher(struct notifier_block *notifier, + unsigned long event, + void *data) +{ + int err; + + err = register_xenbus_watch(&target_watch); + if (err) + printk(KERN_ERR "Failed to set balloon watcher\n"); + + return NOTIFY_DONE; +} + +static struct notifier_block xenstore_notifier; + +static int __init balloon_init(void) +{ + if (!xen_domain()) + return -ENODEV; + + pr_info("xen-balloon: Initialising balloon driver.\n"); + + register_balloon(&balloon_sysdev); + + target_watch.callback = watch_target; + xenstore_notifier.notifier_call = balloon_init_watcher; + + register_xenstore_notifier(&xenstore_notifier); + + return 0; +} +subsys_initcall(balloon_init); + +static void balloon_exit(void) +{ + /* XXX - release balloon here */ + return; +} + +module_exit(balloon_exit); + +#define BALLOON_SHOW(name, format, args...) \ + static ssize_t show_##name(struct sys_device *dev, \ + struct sysdev_attribute *attr, \ + char *buf) \ + { \ + return sprintf(buf, format, ##args); \ + } \ + static SYSDEV_ATTR(name, S_IRUGO, show_##name, NULL) + +BALLOON_SHOW(current_kb, "%lu\n", PAGES2KB(balloon_stats.current_pages)); +BALLOON_SHOW(low_kb, "%lu\n", PAGES2KB(balloon_stats.balloon_low)); +BALLOON_SHOW(high_kb, "%lu\n", PAGES2KB(balloon_stats.balloon_high)); + +static SYSDEV_ULONG_ATTR(schedule_delay, 0444, balloon_stats.schedule_delay); +static SYSDEV_ULONG_ATTR(max_schedule_delay, 0644, balloon_stats.max_schedule_delay); +static SYSDEV_ULONG_ATTR(retry_count, 0444, balloon_stats.retry_count); +static SYSDEV_ULONG_ATTR(max_retry_count, 0644, balloon_stats.max_retry_count); + +static ssize_t show_target_kb(struct sys_device *dev, struct sysdev_attribute *attr, + char *buf) +{ + return sprintf(buf, "%lu\n", PAGES2KB(balloon_stats.target_pages)); +} + +static ssize_t store_target_kb(struct sys_device *dev, + struct sysdev_attribute *attr, + const char *buf, + size_t count) +{ + char *endchar; + unsigned long long target_bytes; + + if (!capable(CAP_SYS_ADMIN)) + return -EPERM; + + target_bytes = simple_strtoull(buf, &endchar, 0) * 1024; + + balloon_set_new_target(target_bytes >> PAGE_SHIFT); + + return count; +} + +static SYSDEV_ATTR(target_kb, S_IRUGO | S_IWUSR, + show_target_kb, store_target_kb); + + +static ssize_t show_target(struct sys_device *dev, struct sysdev_attribute *attr, + char *buf) +{ + return sprintf(buf, "%llu\n", + (unsigned long long)balloon_stats.target_pages + << PAGE_SHIFT); +} + +static ssize_t store_target(struct sys_device *dev, + struct sysdev_attribute *attr, + const char *buf, + size_t count) +{ + char *endchar; + unsigned long long target_bytes; + + if (!capable(CAP_SYS_ADMIN)) + return -EPERM; + + target_bytes = memparse(buf, &endchar); + + balloon_set_new_target(target_bytes >> PAGE_SHIFT); + + return count; +} + +static SYSDEV_ATTR(target, S_IRUGO | S_IWUSR, + show_target, store_target); + + +static struct sysdev_attribute *balloon_attrs[] = { + &attr_target_kb, + &attr_target, + &attr_schedule_delay.attr, + &attr_max_schedule_delay.attr, + &attr_retry_count.attr, + &attr_max_retry_count.attr +}; + +static struct attribute *balloon_info_attrs[] = { + &attr_current_kb.attr, + &attr_low_kb.attr, + &attr_high_kb.attr, + NULL +}; + +static struct attribute_group balloon_info_group = { + .name = "info", + .attrs = balloon_info_attrs +}; + +static struct sysdev_class balloon_sysdev_class = { + .name = BALLOON_CLASS_NAME +}; + +static int register_balloon(struct sys_device *sysdev) +{ + int i, error; + + error = sysdev_class_register(&balloon_sysdev_class); + if (error) + return error; + + sysdev->id = 0; + sysdev->cls = &balloon_sysdev_class; + + error = sysdev_register(sysdev); + if (error) { + sysdev_class_unregister(&balloon_sysdev_class); + return error; + } + + for (i = 0; i < ARRAY_SIZE(balloon_attrs); i++) { + error = sysdev_create_file(sysdev, balloon_attrs[i]); + if (error) + goto fail; + } + + error = sysfs_create_group(&sysdev->kobj, &balloon_info_group); + if (error) + goto fail; + + return 0; + + fail: + while (--i >= 0) + sysdev_remove_file(sysdev, balloon_attrs[i]); + sysdev_unregister(sysdev); + sysdev_class_unregister(&balloon_sysdev_class); + return error; +} + +MODULE_LICENSE("GPL"); diff --git a/include/xen/balloon.h b/include/xen/balloon.h new file mode 100644 index 000000000000..f72e4794ec77 --- /dev/null +++ b/include/xen/balloon.h @@ -0,0 +1,22 @@ +/****************************************************************************** + * Xen balloon functionality + */ + +#define RETRY_UNLIMITED 0 + +struct balloon_stats { + /* We aim for 'current allocation' == 'target allocation'. */ + unsigned long current_pages; + unsigned long target_pages; + /* Number of pages in high- and low-memory balloons. */ + unsigned long balloon_low; + unsigned long balloon_high; + unsigned long schedule_delay; + unsigned long max_schedule_delay; + unsigned long retry_count; + unsigned long max_retry_count; +}; + +extern struct balloon_stats balloon_stats; + +void balloon_set_new_target(unsigned long target); -- cgit v1.2.3 From b6f3067985f12d514187059fb10fe3c877f87cb2 Mon Sep 17 00:00:00 2001 From: Konrad Rzeszutek Wilk Date: Tue, 15 Mar 2011 10:23:57 -0400 Subject: xen-balloon: Add interface to retrieve ballooned pages Pages that have been ballooned are useful for other Xen drivers doing grant table actions, because these pages have valid struct page/PFNs but have no valid MFN so are available for remapping. Acked-by: Ian Campbell Signed-off-by: Daniel De Graaf [v2: Deal with rebasing on top of modified balloon code] Signed-off-by: Konrad Rzeszutek Wilk --- drivers/xen/balloon.c | 73 ++++++++++++++++++++++++++++++++++++++++++++++----- include/xen/balloon.h | 3 +++ 2 files changed, 70 insertions(+), 6 deletions(-) (limited to 'include/xen') diff --git a/drivers/xen/balloon.c b/drivers/xen/balloon.c index 7497041d0631..8c81cd24ed83 100644 --- a/drivers/xen/balloon.c +++ b/drivers/xen/balloon.c @@ -128,14 +128,17 @@ static void balloon_append(struct page *page) } /* balloon_retrieve: rescue a page from the balloon, if it is not empty. */ -static struct page *balloon_retrieve(void) +static struct page *balloon_retrieve(bool prefer_highmem) { struct page *page; if (list_empty(&ballooned_pages)) return NULL; - page = list_entry(ballooned_pages.next, struct page, lru); + if (prefer_highmem) + page = list_entry(ballooned_pages.prev, struct page, lru); + else + page = list_entry(ballooned_pages.next, struct page, lru); list_del(&page->lru); if (PageHighMem(page)) { @@ -233,7 +236,7 @@ static enum bp_state increase_reservation(unsigned long nr_pages) return BP_EAGAIN; for (i = 0; i < rc; i++) { - page = balloon_retrieve(); + page = balloon_retrieve(false); BUG_ON(page == NULL); pfn = page_to_pfn(page); @@ -263,7 +266,7 @@ static enum bp_state increase_reservation(unsigned long nr_pages) return BP_DONE; } -static enum bp_state decrease_reservation(unsigned long nr_pages) +static enum bp_state decrease_reservation(unsigned long nr_pages, gfp_t gfp) { enum bp_state state = BP_DONE; unsigned long pfn, i; @@ -279,7 +282,7 @@ static enum bp_state decrease_reservation(unsigned long nr_pages) nr_pages = ARRAY_SIZE(frame_list); for (i = 0; i < nr_pages; i++) { - if ((page = alloc_page(GFP_BALLOON)) == NULL) { + if ((page = alloc_page(gfp)) == NULL) { nr_pages = i; state = BP_EAGAIN; break; @@ -340,7 +343,7 @@ static void balloon_process(struct work_struct *work) state = increase_reservation(credit); if (credit < 0) - state = decrease_reservation(-credit); + state = decrease_reservation(-credit, GFP_BALLOON); state = update_schedule(state); @@ -366,6 +369,64 @@ void balloon_set_new_target(unsigned long target) } EXPORT_SYMBOL_GPL(balloon_set_new_target); +/** + * alloc_xenballooned_pages - get pages that have been ballooned out + * @nr_pages: Number of pages to get + * @pages: pages returned + * @return 0 on success, error otherwise + */ +int alloc_xenballooned_pages(int nr_pages, struct page** pages) +{ + int pgno = 0; + struct page* page; + mutex_lock(&balloon_mutex); + while (pgno < nr_pages) { + page = balloon_retrieve(true); + if (page) { + pages[pgno++] = page; + } else { + enum bp_state st; + st = decrease_reservation(nr_pages - pgno, GFP_HIGHUSER); + if (st != BP_DONE) + goto out_undo; + } + } + mutex_unlock(&balloon_mutex); + return 0; + out_undo: + while (pgno) + balloon_append(pages[--pgno]); + /* Free the memory back to the kernel soon */ + schedule_delayed_work(&balloon_worker, 0); + mutex_unlock(&balloon_mutex); + return -ENOMEM; +} +EXPORT_SYMBOL(alloc_xenballooned_pages); + +/** + * free_xenballooned_pages - return pages retrieved with get_ballooned_pages + * @nr_pages: Number of pages + * @pages: pages to return + */ +void free_xenballooned_pages(int nr_pages, struct page** pages) +{ + int i; + + mutex_lock(&balloon_mutex); + + for (i = 0; i < nr_pages; i++) { + if (pages[i]) + balloon_append(pages[i]); + } + + /* The balloon may be too large now. Shrink it if needed. */ + if (current_target() != balloon_stats.current_pages) + schedule_delayed_work(&balloon_worker, 0); + + mutex_unlock(&balloon_mutex); +} +EXPORT_SYMBOL(free_xenballooned_pages); + static int __init balloon_init(void) { unsigned long pfn, extra_pfn_end; diff --git a/include/xen/balloon.h b/include/xen/balloon.h index f72e4794ec77..a2b22f01a51d 100644 --- a/include/xen/balloon.h +++ b/include/xen/balloon.h @@ -20,3 +20,6 @@ struct balloon_stats { extern struct balloon_stats balloon_stats; void balloon_set_new_target(unsigned long target); + +int alloc_xenballooned_pages(int nr_pages, struct page** pages); +void free_xenballooned_pages(int nr_pages, struct page** pages); -- cgit v1.2.3