Folio

似乎需要首先了解一下 transparent 的内容

阅读一下各种资料

An introduction to compound pages
- Compound pages can serve as anonymous memory or be used as buffers within the kernel; they cannot, however, appear in the page cache, which is only prepared to deal with singleton pages.

A compound page is simply a grouping of two or more physically contiguous pages into a unit that can, in many ways, be treated as a single, larger page. They are most commonly used to create huge pages, used within hugetlbfs or the transparent huge pages subsystem, but they show up in other contexts as well. Compound pages can serve as anonymous memory or be used as buffers within the kernel; they cannot, however, appear in the page cache, which is only prepared to deal with singleton pages.

Minimizing the use of tail pages
Cramming more into struct page
https://lore.kernel.org/linux-mm/ZNVaEOmUUM5rR4CA@casper.infradead.org/T/#t
https://lwn.net/Articles/619333/
Transparent huge page reference counting

Compound page 的构造过程

prep_compound_page : 初始化和 page 相关的各种参数

asm_exc_page_fault
- handle_page_fault
  - do_user_addr_fault
    - handle_mm_fault
      - __handle_mm_fault
        
        create_huge_pmd
        
        do_huge_pmd_anonymous_page
        
        vma_alloc_folio
        
        __folio_alloc_node
        
        __folio_alloc
        
        __alloc_pages
        
        get_page_from_freelist
        
        prep_new_page
        
        prep_compound_page

使用 readv 读一个大文件，可以得到如下结果，看来 xfs 对于 folio 的支持的确还可以:

kfunc:page_cache_ra_order{ @order = hist(args->new_order); }

结果为

@order:
[0]                12813 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@|
[1]                    0 |                                                    |
[2, 4)                 2 |                                                    |
[4, 8)              4096 |@@@@@@@@@@@@@@@@                                    |

整个的执行路径为:

@[
    folio_add_lru+5
    filemap_add_folio+90
    page_cache_ra_order+413
    filemap_get_pages+1246
    filemap_read+223
    xfs_file_buffered_read+79
    xfs_file_read_iter+110
    vfs_read+499
    ksys_read+111
    do_syscall_64+59
    entry_SYSCALL_64_after_hwframe+114
]: 420691

各种报道

https://news.ycombinator.com/item?id=27509944

看看评测

解读 Linux 内存管理新特性 Memory folios

struct folio *folio_alloc(gfp_t gfp, unsigned int order)
{
	return page_rmappable_folio(alloc_pages(gfp | __GFP_COMP, order));
}

__GFP_COMP 的影响在于， prep_new_page -> prep_compound_page : 也就是说，

关于 folio_alloc 和 alloc_pages 的关系
- 增加了 __GFP_COMP
- 设置了释放的代码
  - https://lore.kernel.org/all/20230816151201.3655946-9-willy@infradead.org/T/#rb3e69a3ef6042dfdd423ae83426f02d8f42f3609

我不得不去学习 LWN: A discussion on folios（https://lwn.net/Articles/869942/）， LPC 2021 - File Systems MC（https://www.youtube.com/watch?v=U6HYrd85hQ8&t=1475s）大佬关于 folio 的讨论。然后发现 Matthew Wilcox 的主题不是《The folio》，而是《Efficient buffered I/O》。事情并不简单。

https://openanolis.cn/sig/Cloud-Kernel/doc/475049355931222178

https://mp.weixin.qq.com/s/CfiUSsYGv8kCPiBqQydhDg

https://mp.weixin.qq.com/s/Jp0TgD2p91m6mUczRdTh0Q

https://mp.weixin.qq.com/s/4XnyOCSQwf6NGXY8RIAI0A

回答这个疑惑

关于 include/linux/page_idle.h 中各种 folio 的 reference ，让人迷茫啊，为什么感觉明显有些 page 是单个 page 的，但是还是使用 folio 接口。

https://mp.weixin.qq.com/s/Jp0TgD2p91m6mUczRdTh0Q

Linux 社区在文件页方面，发展出多个文件系统支持 large folio。这类文件系统会通过 mapping_set_large_folios() 告诉 page cache 这层，它支持 large folio：

等等，这个总结的很好的

为什么文件系统需要特殊的支持 large folio ?

https://mp.weixin.qq.com/s/FeQfJ3P59f01_9UkRoKpjw

永恒的问题

memory 的 folio 的问题，直接

本站所有文章转发 CSDN 将按侵权追究法律责任，其它情况随意。