From patchwork Thu Nov 10 23:35:34 2011 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: John Stultz X-Patchwork-Id: 5034 Return-Path: X-Original-To: patchwork@peony.canonical.com Delivered-To: patchwork@peony.canonical.com Received: from fiordland.canonical.com (fiordland.canonical.com [91.189.94.145]) by peony.canonical.com (Postfix) with ESMTP id 861B723E08 for ; Thu, 10 Nov 2011 23:35:52 +0000 (UTC) Received: from mail-fx0-f52.google.com (mail-fx0-f52.google.com [209.85.161.52]) by fiordland.canonical.com (Postfix) with ESMTP id 6F5DDA1831E for ; Thu, 10 Nov 2011 23:35:52 +0000 (UTC) Received: by faan26 with SMTP id n26so4980473faa.11 for ; Thu, 10 Nov 2011 15:35:52 -0800 (PST) Received: by 10.152.134.116 with SMTP id pj20mr5713851lab.5.1320968152193; Thu, 10 Nov 2011 15:35:52 -0800 (PST) X-Forwarded-To: linaro-patchwork@canonical.com X-Forwarded-For: patch@linaro.org linaro-patchwork@canonical.com Delivered-To: patches@linaro.org Received: by 10.152.40.7 with SMTP id t7cs28590lak; Thu, 10 Nov 2011 15:35:51 -0800 (PST) Received: by 10.42.161.132 with SMTP id t4mr10065383icx.16.1320968149146; Thu, 10 Nov 2011 15:35:49 -0800 (PST) Received: from e5.ny.us.ibm.com (e5.ny.us.ibm.com. [32.97.182.145]) by mx.google.com with ESMTPS id a6si5125647icw.57.2011.11.10.15.35.48 (version=TLSv1/SSLv3 cipher=OTHER); Thu, 10 Nov 2011 15:35:49 -0800 (PST) Received-SPF: pass (google.com: domain of jstultz@us.ibm.com designates 32.97.182.145 as permitted sender) client-ip=32.97.182.145; Authentication-Results: mx.google.com; spf=pass (google.com: domain of jstultz@us.ibm.com designates 32.97.182.145 as permitted sender) smtp.mail=jstultz@us.ibm.com Received: from /spool/local by e5.ny.us.ibm.com with IBM ESMTP SMTP Gateway: Authorized Use Only! Violators will be prosecuted for from ; Thu, 10 Nov 2011 18:35:47 -0500 Received: from d01relay06.pok.ibm.com ([9.56.227.116]) by e5.ny.us.ibm.com ([192.168.1.105]) with IBM ESMTP SMTP Gateway: Authorized Use Only! Violators will be prosecuted; Thu, 10 Nov 2011 18:35:45 -0500 Received: from d01av01.pok.ibm.com (d01av01.pok.ibm.com [9.56.224.215]) by d01relay06.pok.ibm.com (8.13.8/8.13.8/NCO v10.0) with ESMTP id pAANZibs3346618; Thu, 10 Nov 2011 18:35:44 -0500 Received: from d01av01.pok.ibm.com (loopback [127.0.0.1]) by d01av01.pok.ibm.com (8.14.4/8.13.1/NCO v10.0 AVout) with ESMTP id pAANZi42019604; Thu, 10 Nov 2011 18:35:44 -0500 Received: from kernel.beaverton.ibm.com ([9.47.67.96]) by d01av01.pok.ibm.com (8.14.4/8.13.1/NCO v10.0 AVin) with ESMTP id pAANZhkW019585; Thu, 10 Nov 2011 18:35:43 -0500 Received: by kernel.beaverton.ibm.com (Postfix, from userid 1056) id CE3801E74FB; Thu, 10 Nov 2011 15:35:42 -0800 (PST) From: John Stultz To: Dave Hansen Cc: John Stultz Subject: [PATCH] madvise: Add _VOLATILE,_ISVOLATILE, and _NONVOLATILE flags Date: Thu, 10 Nov 2011 15:35:34 -0800 Message-Id: <1320968134-26770-1-git-send-email-john.stultz@linaro.org> X-Mailer: git-send-email 1.7.3.2.146.gca209 x-cbid: 11111023-5930-0000-0000-00000156A882 This patch provides new madvise flags that can be used to mark memory as volatile, which will allow it to be discarded if the kernel wants to reclaim memory. This is very much influenced by the Android Ashmem interface by Robert Love so credits to him and the Android developers. In many cases the code & logic come directly from the ashmem patch. The intent of this patch is to allow for ashmem-like behavior, but embeds the idea a little deeper into the VM code, instead of isolating it into a specific driver. Also many thanks to Dave Hansen who helped design and develop the initial version of this patch, and has provided continued review and mentoring in the VM code. Signed-off-by: John Stultz --- fs/inode.c | 1 + include/asm-generic/mman-common.h | 3 + include/linux/fs.h | 28 ++++++ mm/madvise.c | 171 +++++++++++++++++++++++++++++++++++++ mm/shmem.c | 20 +++++ 5 files changed, 223 insertions(+), 0 deletions(-) diff --git a/fs/inode.c b/fs/inode.c index ee4e66b..c1f55f4 100644 --- a/fs/inode.c +++ b/fs/inode.c @@ -278,6 +278,7 @@ void address_space_init_once(struct address_space *mapping) spin_lock_init(&mapping->private_lock); INIT_RAW_PRIO_TREE_ROOT(&mapping->i_mmap); INIT_LIST_HEAD(&mapping->i_mmap_nonlinear); + INIT_LIST_HEAD(&mapping->unpinned_list); } EXPORT_SYMBOL(address_space_init_once); diff --git a/include/asm-generic/mman-common.h b/include/asm-generic/mman-common.h index 787abbb..c633d13 100644 --- a/include/asm-generic/mman-common.h +++ b/include/asm-generic/mman-common.h @@ -47,6 +47,9 @@ #define MADV_HUGEPAGE 14 /* Worth backing with hugepages */ #define MADV_NOHUGEPAGE 15 /* Not worth backing with hugepages */ +#define MADV_VOLATILE 16 /* _can_ toss, but don't toss now */ +#define MADV_ISVOLATILE 17 /* Check if page is marked volatile or not */ +#define MADV_NONVOLATILE 18 /* Remove VOLATILE flag */ /* compatibility flags */ #define MAP_FILE 0 diff --git a/include/linux/fs.h b/include/linux/fs.h index 0c4df26..a4e281f 100644 --- a/include/linux/fs.h +++ b/include/linux/fs.h @@ -632,6 +632,33 @@ int pagecache_write_end(struct file *, struct address_space *mapping, loff_t pos, unsigned len, unsigned copied, struct page *page, void *fsdata); + + +/* upinned_mem_range & range macros from Robert Love's Ashmem patch */ +struct unpinned_mem_range { + struct address_space *addrsp; /* associated address_space */ + struct list_head unpinned; /* Next unpinned range */ + size_t start; + size_t end; + unsigned int purged; +}; + +#define range_size(range) \ + ((range)->end - (range)->start +1) +#define page_range_subsumes_range(range, start_addr, end_addr) \ + (((range)->start >= (start_addr)) && ((range)->end <= (end_addr))) +#define page_range_subsumed_by_range(range, start_addr, end_addr) \ + (((range)->start <= (start_addr)) && ((range)->end >= (end_addr))) +#define page_in_range(range, page) \ + (((range)->start <= (page)) && ((range)->end >= (page))) +#define page_range_in_range(range, start_addr, end_addr) \ + (page_in_range(range, start_addr) || page_in_range(range, end_addr) || \ + page_range_subsumes_range(range, start_addr, end_addr)) +#define range_before_page(range, page) \ + ((range)->end < (page)) + + + struct backing_dev_info; struct address_space { struct inode *host; /* owner: inode, block_device */ @@ -650,6 +677,7 @@ struct address_space { spinlock_t private_lock; /* for use by the address_space */ struct list_head private_list; /* ditto */ struct address_space *assoc_mapping; /* ditto */ + struct list_head unpinned_list; /* unpinned area list */ } __attribute__((aligned(sizeof(long)))); /* * On most architectures that alignment is already the case; but diff --git a/mm/madvise.c b/mm/madvise.c index 74bf193..fdc98e5 100644 --- a/mm/madvise.c +++ b/mm/madvise.c @@ -225,6 +225,168 @@ static long madvise_remove(struct vm_area_struct *vma, return error; } + + +static int unpinned_range_alloc(struct address_space *addrsp, + struct unpinned_mem_range *prev_range, + unsigned int purged, size_t start, size_t end) +{ + struct unpinned_mem_range *range; + + range = kzalloc(sizeof(struct unpinned_mem_range), GFP_KERNEL); + if (unlikely(!range)) + return -ENOMEM; + + range->addrsp = addrsp; + range->start = start; + range->end = end; + range->purged = purged; + + list_add_tail(&range->unpinned, &prev_range->unpinned); + + return 0; +} + +static void unpinned_range_del(struct unpinned_mem_range *range) +{ + list_del(&range->unpinned); + kfree(range); +} + +static inline void unpinned_range_shrink(struct unpinned_mem_range *range, + size_t start, size_t end) +{ + range->start = start; + range->end = end; +} + + +static long madvise_volatile(struct vm_area_struct * vma, + unsigned long start, unsigned long end) +{ + struct unpinned_mem_range *range, *next; + unsigned int purged = 0; + int ret; + struct address_space *addrsp; + + /* XXX - check start->end is within mmapped shm region */ + if(!vma->vm_file) + return -1; + + addrsp=vma->vm_file->f_mapping; + + + start -= vma->vm_start; + end -= vma->vm_start; + + printk("Madvise_volatile: start: 0x%lx end: 0x%lx vma start: 0x%lx\n", start, end, vma->vm_start); + +restart: + list_for_each_entry_safe(range, next, &addrsp->unpinned_list, + unpinned) { + if (range_before_page(range, start)) + break; + + if(page_range_subsumed_by_range(range, start, end)) + return 0; + if (page_range_in_range(range, start,end)) { + start = min_t(size_t, range->start, start); + end = max_t(size_t, range->end, end); + purged |= range->purged; + unpinned_range_del(range); + goto restart; + } + + } + ret = unpinned_range_alloc(addrsp, range, purged, start, end); + return ret; +} + +static long madvise_nonvolatile(struct vm_area_struct * vma, + unsigned long start, unsigned long end) +{ + struct unpinned_mem_range *range, *next; + int ret = 0; + struct address_space *addrsp; + if(!vma->vm_file) + return -1; + addrsp=vma->vm_file->f_mapping; + + + start -= vma->vm_start; + end -= vma->vm_start; + + list_for_each_entry_safe(range, next, &addrsp->unpinned_list, + unpinned) { + if (range_before_page(range, start)) + break; + + if (page_range_in_range(range, start, end)) { + ret |= range->purged; + /* Case #1: Easy. Just nuke the whole thing. */ + if (page_range_subsumes_range(range, start, end)) { + unpinned_range_del(range); + continue; + } + + /* Case #2: We overlap from the start, so adjust it */ + if (range->start >= start) { + unpinned_range_shrink(range, end + 1, + range->end); + continue; + } + + /* Case #3: We overlap from the rear, so adjust it */ + if (range->end <= end) { + unpinned_range_shrink(range, range->start, + start-1); + continue; + } + + /* + * Case #4: We eat a chunk out of the middle. A bit + * more complicated, we allocate a new range for the + * second half and adjust the first chunk's endpoint. + */ + unpinned_range_alloc(addrsp, range, + range->purged, end + 1, + range->end); + unpinned_range_shrink(range, range->start, start - 1); + + } + + } + + printk("Madvise_nonvolatile: start: 0x%lx end: 0x%lx vma start: 0x%lx ret: %i\n", start, end, vma->vm_start, ret); + return ret; + + +} + +static long madvise_isvolatile(struct vm_area_struct * vma, + unsigned long start, unsigned long end) +{ + struct unpinned_mem_range *range; + long ret = 0; + struct address_space *addrsp; + if(!vma->vm_file) + return -1; + addrsp=vma->vm_file->f_mapping; + + start -= vma->vm_start; + end -= vma->vm_start; + + list_for_each_entry(range, &addrsp->unpinned_list, unpinned) { + if (range_before_page(range, start)) + break; + if (page_range_in_range(range, start, end)) { + ret = 1; + break; + } + } + return ret; +} + #ifdef CONFIG_MEMORY_FAILURE /* * Error injection support for memory error handling. @@ -268,6 +430,12 @@ madvise_vma(struct vm_area_struct *vma, struct vm_area_struct **prev, return madvise_willneed(vma, prev, start, end); case MADV_DONTNEED: return madvise_dontneed(vma, prev, start, end); + case MADV_VOLATILE: + return madvise_volatile(vma, start, end); + case MADV_ISVOLATILE: + return madvise_isvolatile(vma, start, end); + case MADV_NONVOLATILE: + return madvise_nonvolatile(vma, start, end); default: return madvise_behavior(vma, prev, start, end, behavior); } @@ -293,6 +461,9 @@ madvise_behavior_valid(int behavior) case MADV_HUGEPAGE: case MADV_NOHUGEPAGE: #endif + case MADV_VOLATILE: + case MADV_ISVOLATILE: + case MADV_NONVOLATILE: return 1; default: diff --git a/mm/shmem.c b/mm/shmem.c index d672250..85506e0 100644 --- a/mm/shmem.c +++ b/mm/shmem.c @@ -679,6 +679,26 @@ static int shmem_writepage(struct page *page, struct writeback_control *wbc) index = page->index; inode = mapping->host; info = SHMEM_I(inode); + + if (!list_empty(&mapping->unpinned_list)) { + struct unpinned_mem_range *range, *next; + printk("shmem_writepage:\n"); + list_for_each_entry_safe(range, next, &mapping->unpinned_list, + unpinned) { + printk(" range: 0x%lx - 0x%lx vs 0x%lx\n", + range->start, range->end, + (long)page->index << PAGE_SHIFT); + + if (page_in_range(range, + (long)page->index << PAGE_SHIFT)) { + range->purged = 1; + unlock_page(page); + printk(" Purged page!\n"); + return 0; + } + } + } + if (info->flags & VM_LOCKED) goto redirty; if (!total_swap_pages)