文件的预读操作分析
预读的两个优势:
1 )每个 IO 一般涉及一组较大相邻的扇区,预读使磁盘控制器处理较少的命令。
2 )提高系统的响应能力,顺序读取的文件一般不需要等待请求的数据
文件的顺序访问:如果请求的第一页紧跟上次访问所请求的最后一页,就认为是顺序的。
1. 主要的数据结构
/*
* Track a single file's readahead state
*/
struct file_ra_state {
unsigned long start; /* Current window */
unsigned long size;
unsigned long flags; /* ra flags RA_FLAG_xxx*/
unsigned long cache_hit; /* cache hit count*/
unsigned long prev_page; /* Cache last read() position */
unsigned long ahead_start; /* Ahead window */
unsigned long ahead_size;
unsigned long ra_pages; /* Maximum readahead window */
unsigned long mmap_hit; /* Cache hit stat for mmap accesses */
unsigned long mmap_miss; /* Cache miss stat for mmap accesses */
};
#define RA_FLAG_MISS 0x01 /* a cache miss occured against this file */
#define RA_FLAG_INCACHE 0x02 /* file is already in cache */
2. 主要流程
2.1. 系统调用:设置预取策略
asmlinkage long sys_fadvise64(int fd, loff_t offset, size_t len, int advice)
{
return sys_fadvise64_64(fd, offset, len, advice);
}
/*
* POSIX_FADV_WILLNEED could set PG_Referenced, and POSIX_FADV_NOREUSE could
* deactivate the pages and clear PG_Referenced.
*/
asmlinkage long sys_fadvise64_64(int fd, loff_t offset, loff_t len, int advice)
{
struct file *file = fget(fd);
struct address_space *mapping;
struct backing_dev_info *bdi;
loff_t endbyte; /* inclusive */
pgoff_t start_index;
pgoff_t end_index;
unsigned long nrpages;
int ret = 0;
if (!file)
return -EBADF;
if (S_ISFIFO(file->f_dentry->d_inode->i_mode)) {
ret = -ESPIPE;
goto out;
}
mapping = file->f_mapping;
if (!mapping || len < 0) {
ret = -EINVAL;
goto out;
}
if (mapping->a_ops->get_xip_page)
/* no bad return value, but ignore advice */
goto out;
/* Careful about overflows. Len == 0 means "as much as possible" */
endbyte = offset + len;
if (!len || endbyte < len)
endbyte = -1;
else
endbyte--; /* inclusive */
bdi = mapping->backing_dev_info;
switch (advice) {
case POSIX_FADV_NORMAL:
file->f_ra.ra_pages = bdi->ra_pages;
break;
case POSIX_FADV_RANDOM:
file->f_ra.ra_pages = 0;
break;
case POSIX_FADV_SEQUENTIAL:
file->f_ra.ra_pages = bdi->ra_pages * 2;
break;
case POSIX_FADV_WILLNEED:
if (!mapping->a_ops->readpage) {
ret = -EINVAL;
break;
}
/* First and last PARTIAL page! */
start_index = offset >> PAGE_CACHE_SHIFT;
end_index = endbyte >> PAGE_CACHE_SHIFT;
/* Careful about overflow on the "+1" */
nrpages = end_index - start_index + 1;
if (!nrpages)
nrpages = ~0UL;
// 强制预读某部分的页
ret = force_page_cache_readahead(mapping, file,
start_index,
max_sane_readahead(nrpages));
if (ret > 0)
ret = 0;
break;
case POSIX_FADV_NOREUSE:
break;
case POSIX_FADV_DONTNEED:
if (!bdi_write_congested(mapping->backing_dev_info))
filemap_flush(mapping);
/* First and last FULL page! */
start_index = (offset+(PAGE_CACHE_SIZE-1)) >> PAGE_CACHE_SHIFT;
end_index = (endbyte >> PAGE_CACHE_SHIFT);
if (end_index >= start_index)
invalidate_mapping_pages(mapping, start_index,
end_index);
break;
default:
ret = -EINVAL;
}
out:
fput(file);
return ret;
}
2.2. 初始化
当一个文件被打开时,在它的 file_ra_state 描述符中,仅设置两个参数:
prev_page 为进程上一次读操作中的最后一页索引,初值为 -1
ra_pages 为当前窗的最大页数,即对文件允许的最大预读量。存放在该文件所在设备的 backing_dev_info->ra_pages 中,应用程序通过系统调用 fadvise 中来设置策略
POSIX_FADV_NORMAL 最大预读量为缺失值 32
POSIX_FADV_SEQUENTIAL 最大预读量为缺失值两倍
POSIX_FADV_RANDOM 最大预读量为 0
/*
* Initialise a struct file's readahead state. Assumes that the caller has
* memset *ra to zero.
*/
void file_ra_state_init(struct file_ra_state *ra, struct address_space *mapping)
{
ra->ra_pages = mapping->backing_dev_info->ra_pages;
ra->prev_page = -1;
}
2.3. page_cache_read_ahead
本函数是预读的主要流程,主要设置当前窗和预读窗两个窗口。
/**
* page_cache_readahead - generic adaptive readahead
* @mapping: address_space which holds the pagecache and I/O vectors
* @ra: file_ra_state which holds the readahead state
* @filp: passed on to ->readpage() and ->readpages()
* @offset: start offset into @mapping, in PAGE_CACHE_SIZE units
* @req_size: hint: total size of the read which the caller is performing in
* PAGE_CACHE_SIZE units
*
* page_cache_readahead() is the main function. If performs the adaptive
* readahead window size management and submits the readahead I/O.
*
* Note that @filp is purely used for passing on to the ->readpage[s]()
* handler: it may refer to a different file from @mapping (so we may not use
* @filp->f_mapping or @filp->f_dentry->d_inode here).
* Also, @ra may not be equal to &@filp->f_ra.
*
*/
unsigned long
page_cache_readahead(struct address_space *mapping, struct file_ra_state *ra,
struct file *filp, pgoff_t offset, unsigned long req_size)
{
unsigned long max, newsize;
int sequential;
/*
* We avoid doing extra work and bogusly perturbing the readahead
* window expansion logic.
*/
if (offset == ra->prev_page && --req_size)
++offset;
/* Note that prev_page == -1 if it is a first read */
// 如果要读的页 offset 正好等于上次进程请求的最后一页,标记为顺序
// 初始化化时 ra->prev_page 等于 -1 ,也就是开始 offset 等于 0 时才有预读?
sequential = (offset == ra->prev_page + 1);
ra->prev_page = offset;
max = get_max_readahead(ra);
newsize = min(req_size, max);
/* No readahead or sub-page sized read or file already in cache */
// 没有文件预读或者文件已经在高速缓存内
if (newsize == 0 || (ra->flags & RA_FLAG_INCACHE))
goto out;
ra->prev_page += newsize - 1;
/*
* Special case - first read at start of file. We'll assume it's
* a whole-file read and grow the window fast. Or detect first
* sequential access
*/
/* 顺序读且当前窗为空(第一次读) */
if (sequential && ra->size == 0) {
ra->size = get_init_ra_size(newsize, max); // 获得初始化的当前窗的大小
ra->start = offset; // 设置当前窗的开始
// 对当前窗进行 I/O 操作
if (!blockable_page_cache_readahead(mapping, filp, offset,
ra->size, ra, 1))
goto out;
/*
* If the request size is larger than our max readahead, we
* at least want to be sure that we get 2 IOs in flight and
* we know that we will definitly need the new I/O.
* once we do this, subsequent calls should be able to overlap
* IOs,* thus preventing stalls. so issue the ahead window
* immediately.
*/
/* 如果请求的页数大于当前窗的最大值,创建一个预读窗,并对其进行 IO 操作 */
if (req_size >= max)
make_ahead_window(mapping, filp, ra, 1);
goto out;
}
/*
* Now handle the random case:
* partial page reads and first access were handled above,
* so this must be the next page otherwise it is random
*/
如果不是顺序的,则关闭预读
if (!sequential) {
ra_off(ra);
blockable_page_cache_readahead(mapping, filp, offset,
newsize, ra, 1);
goto out;
}
/*
* If we get here we are doing sequential IO and this was not the first
* occurence (ie we have an existing window)
*/
顺序读且不是一个首次读,没有预读窗
if (ra->ahead_start == 0) { /* no ahead window yet */
if (!make_ahead_window( mapping, filp, ra, 0))
goto recheck;
}
/*
* Already have an ahead window, check if we crossed into it.
* If so, shift windows and issue a new ahead window.
* Only return the #pages that are in the current window, so that
* we get called back on the first page of the ahead window which
* will allow us to submit more IO.
*/
顺序读且不是一个首次读,有预读窗,且当前窗和预读窗有重叠
则预读窗为当前窗,重新构建预读窗
if (ra->prev_page >= ra->ahead_start) {
ra->start = ra->ahead_start;
ra->size = ra->ahead_size;
make_ahead_window(mapping, filp, ra, 0);
recheck:
/* prev_page shouldn't overrun the ahead window */
// 进程最后一次访问的页号必须小于预读窗
ra->prev_page = min(ra->prev_page,
ra->ahead_start + ra->ahead_size - 1);
}
out:
return ra->prev_page + 1;
}
预读操作的主要流程图如下:
get_init_ra_size
/*
* Set the initial window size, round to next power of 2 and square
* for small size, x 4 for medium, and x 2 for large
* for 128k (32 page) max ra
* 1-8 page = 32k initial, > 8 page = 128k initial
*/
/* 设置初始窗口大小,首先对齐到 2 的倍数 */
对于比较小,即小于最大值的 32 分之一,设置 4 倍
对于小于 4 分之一,设置为 2 倍
否则,直接设为最大值
static unsigned long get_init_ra_size(unsigned long size, unsigned long max)
{
unsigned long newsize = roundup_pow_of_two(size);
if (newsize <= max / 32)
newsize = newsize * 4;
else if (newsize <= max / 4)
newsize = newsize * 2;
else
newsize = max;
return newsize;
}
make_ahead_window
static int make_ahead_window(struct address_space *mapping, struct file *filp,
struct file_ra_state *ra, int force)
{
int block, ret;
// 获得下一个预读窗的大小
ra->ahead_size = get_next_ra_size (ra);
// 预读窗的开始为当前窗的结尾
ra->ahead_start = ra->start + ra->size;
// 强制执行或者进程最后请求的页在预读窗内
block = force || (ra->prev_page >= ra->ahead_start);
ret = blockable_page_cache_readahead(mapping, filp,
ra->ahead_start, ra->ahead_size, ra, block);
// 如果没有成功,且不是强制执行,则复位预读窗
if (!ret && !force) {
/* A read failure in blocking mode, implies pages are
* all cached. So we can safely assume we have taken
* care of all the pages requested in this call.
* A read failure in non-blocking mode, implies we are
* reading more pages than requested in this call. So
* we safely assume we have taken care of all the pages
* requested in this call.
*
* Just reset the ahead window in case we failed due to
* congestion. The ahead window will any way be closed
* in case we failed due to excessive page cache hits.
*/
reset_ahead_window(ra);
}
return ret;
}
get_next_ra_size
/*
* Set the new window size, this is called only when I/O is to be submitted,
* not for each call to readahead. If a cache miss occured, reduce next I/O
* size, else increase depending on how close to max we are.
*/
static inline unsigned long get_next_ra_size(struct file_ra_state *ra)
{
unsigned long max = get_max_readahead(ra); // 默认最大 32 页
unsigned long min = get_min_readahead(ra); // 最小 4 页
unsigned long cur = ra->size;
unsigned long newsize;
if (ra->flags & RA_FLAG_MISS) { // 如果有没有命中的标志
ra->flags &= ~RA_FLAG_MISS; // 取消该标志
newsize = max((cur - 2), min); // 当前窗的值减 2 ,预读减少
} else if (cur < max / 16) {
newsize = 4 * cur; // 当前窗的 4 倍
} else {
newsize = 2 * cur; // 当前窗的 2 倍
}
return min(newsize, max);
}
handle_ra_miss
/*
* handle_ra_miss() is called when it is known that a page which should have
* been present in the pagecache (we just did some readahead there) was in fact
* not found. This will happen if it was evicted by the VM (readahead
* thrashing)
*
* Turn on the cache miss flag in the RA struct, this will cause the RA code
* to reduce the RA size on the next read.
*/
void handle_ra_miss(struct address_space *mapping,
struct file_ra_state *ra, pgoff_t offset)
{
ra->flags |= RA_FLAG_MISS;
ra->flags &= ~RA_FLAG_INCACHE;
ra->cache_hit = 0;
}
blockable_page_cache_readahead
/*
* Read 'nr_to_read' pages starting at page 'offset'. If the flag 'block'
* is set wait till the read completes. Otherwise attempt to read without
* blocking.
* Returns 1 meaning 'success' if read is successful without switching off
* readahead mode. Otherwise return failure.
*/
static int
blockable_page_cache_readahead(struct address_space *mapping, struct file *filp,
pgoff_t offset, unsigned long nr_to_read,
struct file_ra_state *ra, int block)
{
int actual;
if (!block && bdi_read_congested(mapping->backing_dev_info))
return 0;
actual = __do_page_cache_readahead(mapping, filp, offset, nr_to_read);
return check_ra_success(ra, nr_to_read, actual);
}
check_ra_success
/*
* Check how effective readahead is being. If the amount of started IO is
* less than expected then the file is partly or fully in pagecache and
* readahead isn't helping.
*
*/
static inline int check_ra_success(struct file_ra_state *ra,
unsigned long nr_to_read, unsigned long actual)
{
// 如果预读成功,且如请求的页大于 256 页,则关闭预读,设置全部在缓存的标志。
if (actual == 0) {
ra->cache_hit += nr_to_read;
if (ra->cache_hit >= VM_MAX_CACHE_HIT) {
ra_off(ra);
ra->flags |= RA_FLAG_INCACHE;
return 0;
}
} else {
ra->cache_hit=0;
}
return 1;
}
__do_page_cache_readahead
/*
* Readahead design.
*
* The fields in struct file_ra_state represent the most-recently-executed
* readahead attempt:
*
* start: Page index at which we started the readahead
* size: Number of pages in that read
* Together, these form the "current window".
* Together, start and size represent the `readahead window'.
* prev_page: The page which the readahead algorithm most-recently inspected.
* It is mainly used to detect sequential file reading.
* If page_cache_readahead sees that it is again being called for
* a page which it just looked at, it can return immediately without
* making any state changes.
* ahead_start,
* ahead_size: Together, these form the "ahead window".
* ra_pages: The externally controlled max readahead for this fd.
*
* When readahead is in the off state (size == 0), readahead is disabled.
* In this state, prev_page is used to detect the resumption of sequential I/O.
*
* The readahead code manages two windows - the "current" and the "ahead"
* windows. The intent is that while the application is walking the pages
* in the current window, I/O is underway on the ahead window. When the
* current window is fully traversed, it is replaced by the ahead window
* and the ahead window is invalidated. When this copying happens, the
* new current window's pages are probably still locked. So
* we submit a new batch of I/O immediately, creating a new ahead window.
*
* So:
*
* ----|----------------|----------------|-----
* ^start ^start+size
* ^ahead_start ^ahead_start+ahead_size
*
* ^ When this page is read, we submit I/O for the
* ahead window.
*
* A `readahead hit' occurs when a read request is made against a page which is
* the next sequential page. Ahead window calculations are done only when it
* is time to submit a new IO. The code ramps up the size agressively at first,
* but slow down as it approaches max_readhead.
*
* Any seek/ramdom IO will result in readahead being turned off. It will resume
* at the first sequential access.
*
* There is a special-case: if the first page which the application tries to
* read happens to be the first page of the file, it is assumed that a linear
* read is about to happen and the window is immediately set to the initial size
* based on I/O request size and the max_readahead.
*
* This function is to be called for every read request, rather than when
* it is time to perform readahead. It is called only once for the entire I/O
* regardless of size unless readahead is unable to start enough I/O to satisfy
* the request (I/O request > max_readahead).
*/
/*
* do_page_cache_readahead actually reads a chunk of disk. It allocates all
* the pages first, then submits them all for I/O. This avoids the very bad
* behaviour which would occur if page allocations are causing VM writeback.
* We really don't want to intermingle reads and writes like that.
*
* Returns the number of pages requested, or the maximum amount of I/O allowed.
*
* do_page_cache_readahead() returns -1 if it encountered request queue
* congestion.
*/
static int
__do_page_cache_readahead(struct address_space *mapping, struct file *filp,
pgoff_t offset, unsigned long nr_to_read)
{
struct inode *inode = mapping->host;
struct page *page;
unsigned long end_index; /* The last page we want to read */
LIST_HEAD(page_pool);
int page_idx;
int ret = 0;
loff_t isize = i_size_read(inode);
if (isize == 0)
goto out;
end_index = ((isize - 1) >> PAGE_CACHE_SHIFT);
/*
* Preallocate as many pages as we will need.
*/
read_lock_irq(&mapping->tree_lock);
for (page_idx = 0; page_idx < nr_to_read; page_idx++) {
pgoff_t page_offset = offset + page_idx;
if (page_offset > end_index)
break;
// 如果已在缓存中,跳过
page = radix_tree_lookup(&mapping->page_tree, page_offset);
if (page)
continue;
read_unlock_irq(&mapping->tree_lock);
// 分配新的页框
page = page_cache_alloc_cold(mapping);
read_lock_irq(&mapping->tree_lock);
if (!page)
break;
page->index = page_offset;
list_add(&page->lru, &page_pool);
ret++;
}
read_unlock_irq(&mapping->tree_lock);
/*
* Now start the IO. We ignore I/O errors - if the page is not
* uptodate then the caller will launch readpage again, and
* will then handle the error.
*/
if (ret)
read_pages(mapping, filp, &page_pool, ret);
BUG_ON(!list_empty(&page_pool));
out:
return ret;
}
read_pages
static int read_pages(struct address_space *mapping, struct file *filp,
struct list_head *pages, unsigned nr_pages)
{
unsigned page_idx;
struct pagevec lru_pvec;
int ret;
if (mapping->a_ops->readpages) {
ret = mapping->a_ops->readpages(filp, mapping, pages, nr_pages);
goto out;
}
pagevec_init(&lru_pvec, 0);
for (page_idx = 0; page_idx < nr_pages; page_idx++) {
struct page *page = list_to_page(pages);
list_del(&page->lru);
if (!add_to_page_cache(page, mapping,
page->index, GFP_KERNEL)) {
mapping->a_ops->readpage(filp, page);
if (!pagevec_add(&lru_pvec, page))
__pagevec_lru_add(&lru_pvec);
} else
page_cache_release(page);
}
pagevec_lru_add(&lru_pvec);
ret = 0;
out:
return ret;
}