|
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] Re: [Xen-devel] [PATCH] GSoC 2010 - Memory hotplug support for Xen guests - third fully working version
Hi, On Fri, Aug 13, 2010 at 09:49:52AM +0400, Vasiliy G Tolstov wrote: [...] > Thank You for try. > link ftp://ftp.suse.com/pub/projects/kernel/kotd/SLE11-SP1/src/ Here is the patch which applies to Linux Kernel Ver. 2.6.32.19. (SLES kernel is based on 2.6.32 and patch applies with one warning). It compiles however I did not have time to test it. Daniel Signed-off-by: Daniel Kiper <dkiper@xxxxxxxxxxxx> --- diff -Npru linux-2.6.32.19.orig/drivers/xen/Kconfig linux-2.6.32.19/drivers/xen/Kconfig --- linux-2.6.32.19.orig/drivers/xen/Kconfig 2010-08-13 22:24:37.000000000 +0200 +++ linux-2.6.32.19/drivers/xen/Kconfig 2010-08-16 02:01:35.000000000 +0200 @@ -7,6 +7,16 @@ config XEN_BALLOON the system to expand the domain's memory allocation, or alternatively return unneeded memory to the system. +config XEN_BALLOON_MEMORY_HOTPLUG + bool "Xen memory balloon driver with memory hotplug support" + default n + depends on XEN_BALLOON && MEMORY_HOTPLUG + help + Xen memory balloon driver with memory hotplug support allows expanding + memory available for the system above limit declared at system startup. + It is very useful on critical systems which require long run without + rebooting. + config XEN_SCRUB_PAGES bool "Scrub pages before returning them to system" depends on XEN_BALLOON diff -Npru linux-2.6.32.19.orig/drivers/xen/balloon.c linux-2.6.32.19/drivers/xen/balloon.c --- linux-2.6.32.19.orig/drivers/xen/balloon.c 2010-08-13 22:24:37.000000000 +0200 +++ linux-2.6.32.19/drivers/xen/balloon.c 2010-08-16 02:13:12.000000000 +0200 @@ -6,6 +6,7 @@ * Copyright (c) 2003, B Dragovic * Copyright (c) 2003-2004, M Williamson, K Fraser * Copyright (c) 2005 Dan M. Smith, IBM Corporation + * Copyright (c) 2010 Daniel Kiper * * This program is free software; you can redistribute it and/or * modify it under the terms of the GNU General Public License version 2 @@ -43,6 +44,7 @@ #include <linux/mutex.h> #include <linux/list.h> #include <linux/sysdev.h> +#include <linux/memory.h> #include <asm/page.h> #include <asm/pgalloc.h> @@ -74,6 +76,11 @@ struct balloon_stats { /* Number of pages in high- and low-memory balloons. */ unsigned long balloon_low; unsigned long balloon_high; +#ifdef CONFIG_XEN_BALLOON_MEMORY_HOTPLUG + unsigned long boot_max_pfn; + u64 hotplug_start_paddr; + u64 hotplug_size; +#endif }; static DEFINE_MUTEX(balloon_mutex); @@ -181,17 +188,173 @@ static void balloon_alarm(unsigned long schedule_work(&balloon_worker); } -static unsigned long current_target(void) +#ifdef CONFIG_XEN_BALLOON_MEMORY_HOTPLUG +static inline u64 is_memory_resource_reserved(void) +{ + return balloon_stats.hotplug_start_paddr; +} + +static int allocate_additional_memory(unsigned long nr_pages) +{ + long rc; + resource_size_t r_min, r_size; + struct resource *r; + struct xen_memory_reservation reservation = { + .address_bits = 0, + .extent_order = 0, + .domid = DOMID_SELF + }; + unsigned long flags, i, pfn; + + if (nr_pages > ARRAY_SIZE(frame_list)) + nr_pages = ARRAY_SIZE(frame_list); + + if (!is_memory_resource_reserved()) { + + /* + * Look for first unused memory region starting at page + * boundary. Skip last memory section created at boot time + * becuase it may contains unused memory pages with PG_reserved + * bit not set (online_pages require PG_reserved bit set). + */ + + r = kzalloc(sizeof(struct resource), GFP_KERNEL); + + if (!r) { + rc = -ENOMEM; + goto out_0; + } + + r->name = "System RAM"; + r->flags = IORESOURCE_MEM | IORESOURCE_BUSY; + r_min = PFN_PHYS(section_nr_to_pfn(pfn_to_section_nr(balloon_stats.boot_max_pfn) + 1)); + r_size = (balloon_stats.target_pages - balloon_stats.current_pages) << PAGE_SHIFT; + + rc = allocate_resource(&iomem_resource, r, r_size, r_min, + ULONG_MAX, PAGE_SIZE, NULL, NULL); + + if (rc < 0) { + kfree(r); + goto out_0; + } + + balloon_stats.hotplug_start_paddr = r->start; + } + + spin_lock_irqsave(&balloon_lock, flags); + + pfn = PFN_DOWN(balloon_stats.hotplug_start_paddr + balloon_stats.hotplug_size); + + for (i = 0; i < nr_pages; ++i, ++pfn) + frame_list[i] = pfn; + + set_xen_guest_handle(reservation.extent_start, frame_list); + reservation.nr_extents = nr_pages; + + rc = HYPERVISOR_memory_op(XENMEM_populate_physmap, &reservation); + + if (rc < 0) + goto out_1; + + pfn = PFN_DOWN(balloon_stats.hotplug_start_paddr + balloon_stats.hotplug_size); + + for (i = 0; i < rc; ++i, ++pfn) { + BUG_ON(!xen_feature(XENFEAT_auto_translated_physmap) && + phys_to_machine_mapping_valid(pfn)); + set_phys_to_machine(pfn, frame_list[i]); + } + + balloon_stats.hotplug_size += rc << PAGE_SHIFT; + balloon_stats.current_pages += rc; + +out_1: + spin_unlock_irqrestore(&balloon_lock, flags); + +out_0: + return rc < 0 ? rc : rc != nr_pages; +} + +static void hotplug_allocated_memory(void) +{ + int nid, ret; + struct memory_block *mem; + unsigned long pfn, pfn_limit; + + nid = memory_add_physaddr_to_nid(balloon_stats.hotplug_start_paddr); + + ret = add_registered_memory(nid, balloon_stats.hotplug_start_paddr, + balloon_stats.hotplug_size); + + if (ret) { + pr_err("%s: add_registered_memory: Memory hotplug failed: %i\n", + __func__, ret); + goto error; + } + + if (xen_pv_domain()) { + pfn = PFN_DOWN(balloon_stats.hotplug_start_paddr); + pfn_limit = pfn + (balloon_stats.hotplug_size >> PAGE_SHIFT); + + for (; pfn < pfn_limit; ++pfn) + if (!PageHighMem(pfn_to_page(pfn))) + BUG_ON(HYPERVISOR_update_va_mapping( + (unsigned long)__va(pfn << PAGE_SHIFT), + mfn_pte(pfn_to_mfn(pfn), PAGE_KERNEL), 0)); + } + + ret = online_pages(PFN_DOWN(balloon_stats.hotplug_start_paddr), + balloon_stats.hotplug_size >> PAGE_SHIFT); + + if (ret) { + pr_err("%s: online_pages: Failed: %i\n", __func__, ret); + goto error; + } + + pfn = PFN_DOWN(balloon_stats.hotplug_start_paddr); + pfn_limit = pfn + (balloon_stats.hotplug_size >> PAGE_SHIFT); + + for (; pfn < pfn_limit; pfn += PAGES_PER_SECTION) { + mem = find_memory_block(__pfn_to_section(pfn)); + BUG_ON(!mem); + BUG_ON(!present_section_nr(mem->phys_index)); + mutex_lock(&mem->state_mutex); + mem->state = MEM_ONLINE; + mutex_unlock(&mem->state_mutex); + } + + goto out; + +error: + balloon_stats.current_pages -= balloon_stats.hotplug_size >> PAGE_SHIFT; + balloon_stats.target_pages -= balloon_stats.hotplug_size >> PAGE_SHIFT; + +out: + balloon_stats.hotplug_start_paddr = 0; + balloon_stats.hotplug_size = 0; +} +#else +static inline u64 is_memory_resource_reserved(void) +{ + return 0; +} + +static inline int allocate_additional_memory(unsigned long nr_pages) { - unsigned long target = balloon_stats.target_pages; + /* + * CONFIG_XEN_BALLOON_MEMORY_HOTPLUG is not set. + * balloon_stats.target_pages could not be bigger + * than balloon_stats.current_pages because additional + * memory allocation is not possible. + */ + balloon_stats.target_pages = balloon_stats.current_pages; - target = min(target, - balloon_stats.current_pages + - balloon_stats.balloon_low + - balloon_stats.balloon_high); + return 0; +} - return target; +static inline void hotplug_allocated_memory(void) +{ } +#endif /* CONFIG_XEN_BALLOON_MEMORY_HOTPLUG */ static int increase_reservation(unsigned long nr_pages) { @@ -233,7 +396,7 @@ static int increase_reservation(unsigned set_phys_to_machine(pfn, frame_list[i]); /* Link back into the page tables if not highmem. */ - if (pfn < max_low_pfn) { + if (xen_pv_domain() && !PageHighMem(page)) { int ret; ret = HYPERVISOR_update_va_mapping( (unsigned long)__va(pfn << PAGE_SHIFT), @@ -283,7 +446,7 @@ static int decrease_reservation(unsigned scrub_page(page); - if (!PageHighMem(page)) { + if (xen_pv_domain() && !PageHighMem(page)) { ret = HYPERVISOR_update_va_mapping( (unsigned long)__va(pfn << PAGE_SHIFT), __pte_ma(0), 0); @@ -331,9 +494,15 @@ static void balloon_process(struct work_ mutex_lock(&balloon_mutex); do { - credit = current_target() - balloon_stats.current_pages; - if (credit > 0) - need_sleep = (increase_reservation(credit) != 0); + credit = balloon_stats.target_pages - balloon_stats.current_pages; + + if (credit > 0) { + if (balloon_stats.balloon_low || balloon_stats.balloon_high) + need_sleep = (increase_reservation(credit) != 0); + else + need_sleep = (allocate_additional_memory(credit) != 0); + } + if (credit < 0) need_sleep = (decrease_reservation(-credit) != 0); @@ -344,8 +513,10 @@ static void balloon_process(struct work_ } while ((credit != 0) && !need_sleep); /* Schedule more work if there is some still to be done. */ - if (current_target() != balloon_stats.current_pages) + if (balloon_stats.target_pages != balloon_stats.current_pages) mod_timer(&balloon_timer, jiffies + HZ); + else if (is_memory_resource_reserved()) + hotplug_allocated_memory(); mutex_unlock(&balloon_mutex); } @@ -402,17 +573,27 @@ static int __init balloon_init(void) unsigned long pfn; struct page *page; - if (!xen_pv_domain()) + if (!xen_domain()) return -ENODEV; pr_info("xen_balloon: Initialising balloon driver.\n"); - balloon_stats.current_pages = min(xen_start_info->nr_pages, max_pfn); + if (xen_pv_domain()) + balloon_stats.current_pages = min(xen_start_info->nr_pages, max_pfn); + else + balloon_stats.current_pages = max_pfn; + balloon_stats.target_pages = balloon_stats.current_pages; balloon_stats.balloon_low = 0; balloon_stats.balloon_high = 0; balloon_stats.driver_pages = 0UL; +#ifdef CONFIG_XEN_BALLOON_MEMORY_HOTPLUG + balloon_stats.boot_max_pfn = max_pfn; + balloon_stats.hotplug_start_paddr = 0; + balloon_stats.hotplug_size = 0; +#endif + init_timer(&balloon_timer); balloon_timer.data = 0; balloon_timer.function = balloon_alarm; @@ -420,11 +601,12 @@ static int __init balloon_init(void) register_balloon(&balloon_sysdev); /* Initialise the balloon with excess memory space. */ - for (pfn = xen_start_info->nr_pages; pfn < max_pfn; pfn++) { - page = pfn_to_page(pfn); - if (!PageReserved(page)) - balloon_append(page); - } + if (xen_pv_domain()) + for (pfn = xen_start_info->nr_pages; pfn < max_pfn; pfn++) { + page = pfn_to_page(pfn); + if (!PageReserved(page)) + balloon_append(page); + } target_watch.callback = watch_target; xenstore_notifier.notifier_call = balloon_init_watcher; diff -Npru linux-2.6.32.19.orig/include/linux/memory_hotplug.h linux-2.6.32.19/include/linux/memory_hotplug.h --- linux-2.6.32.19.orig/include/linux/memory_hotplug.h 2010-08-13 22:24:37.000000000 +0200 +++ linux-2.6.32.19/include/linux/memory_hotplug.h 2010-08-16 02:01:35.000000000 +0200 @@ -203,6 +203,7 @@ static inline int is_mem_section_removab } #endif /* CONFIG_MEMORY_HOTREMOVE */ +extern int add_registered_memory(int nid, u64 start, u64 size); extern int add_memory(int nid, u64 start, u64 size); extern int arch_add_memory(int nid, u64 start, u64 size); extern int remove_memory(u64 start, u64 size); diff -Npru linux-2.6.32.19.orig/mm/memory_hotplug.c linux-2.6.32.19/mm/memory_hotplug.c --- linux-2.6.32.19.orig/mm/memory_hotplug.c 2010-08-13 22:24:37.000000000 +0200 +++ linux-2.6.32.19/mm/memory_hotplug.c 2010-08-16 02:20:13.000000000 +0200 @@ -477,22 +477,13 @@ static void rollback_node_hotadd(int nid return; } - /* we are OK calling __meminit stuff here - we have CONFIG_MEMORY_HOTPLUG */ -int __ref add_memory(int nid, u64 start, u64 size) +static int __ref __add_memory(int nid, u64 start, u64 size) { pg_data_t *pgdat = NULL; int new_pgdat = 0; - struct resource *res; int ret; - lock_system_sleep(); - - res = register_memory_resource(start, size); - ret = -EEXIST; - if (!res) - goto out; - if (!node_online(nid)) { pgdat = hotadd_new_pgdat(nid, start); ret = -ENOMEM; @@ -523,14 +514,48 @@ int __ref add_memory(int nid, u64 start, goto out; error: - /* rollback pgdat allocation and others */ + /* rollback pgdat allocation */ if (new_pgdat) rollback_node_hotadd(nid, pgdat); - if (res) - release_memory_resource(res); out: + return ret; +} + +int add_registered_memory(int nid, u64 start, u64 size) +{ + int ret; + + lock_system_sleep(); + ret = __add_memory(nid, start, size); unlock_system_sleep(); + + return ret; +} +EXPORT_SYMBOL_GPL(add_registered_memory); + +int add_memory(int nid, u64 start, u64 size) +{ + int ret = -EEXIST; + struct resource *res; + + lock_system_sleep(); + + res = register_memory_resource(start, size); + + if (!res) + goto out; + + ret = __add_memory(nid, start, size); + + if (!ret) + goto out; + + release_memory_resource(res); + +out: + unlock_system_sleep(); + return ret; } EXPORT_SYMBOL_GPL(add_memory); _______________________________________________ Xen-devel mailing list Xen-devel@xxxxxxxxxxxxxxxxxxx http://lists.xensource.com/xen-devel
|
 |
Lists.xenproject.org is hosted with RackSpace, monitoring our |