mirror of
https://github.com/gatieme/LDD-LinuxDeviceDrivers.git
synced 2026-09-20 19:33:40 +08:00
description/memory: vmalloc
This commit is contained in:
Cheng Jian
@@ -664,20 +664,31 @@ https://lore.kernel.org/patchwork/cover/668967
|
||||
|
||||
| 时间 | 作者 | 特性 | 描述 | 是否合入主线 | 链接 |
|
||||
|:----:|:----:|:---:|:----:|:---------:|:----:|
|
||||
| NA | Nick Piggin <npiggin@suse.de> | [Import 0.99.13k](https://elixir.bootlin.com/linux/0.99.13k/source/mm/vmalloc.c) | 引入 vmalloc 分配器. | ☑ 0.99.13k | [HISTORY commit](https://github.com/gatieme/linux-history/commit/537b6ff02ce317a747658a9f5dff96ed2733b28a) |
|
||||
| 2002/08/06 | Nick Piggin <npiggin@suse.de> | [VM: Rework vmalloc code to support mapping of arbitray pages](https://github.com/gatieme/linux-history/commit/d24919a7fbc635bea6ecc267058dcdbadf03f565) | 引入 vmap/vunmap. 将 vmalloc 操作分为两部分: 分配支持物理地址不连续的页面(vmalloc)以及将这些页面映射到内核页表中, 以便进行实际的临时访问. 因此引入一组新的接口vmap/vunmap 允许将任意页映射到内核虚拟内存中. | ☑ 2.5.32~39 | [HISTORY commit](https://github.com/gatieme/linux-history/commit/d24919a7fbc635bea6ecc267058dcdbadf03f565) |
|
||||
| 2008/07/28 | Nick Piggin <npiggin@suse.de> | [mm: vmap rewrite](https://lwn.net/Articles/304188) | 重写 vmap 分配器 | RFC ☑ 2.6.28-rc1 | [PatchWork](https://lore.kernel.org/patchwork/cover/118352))<br>*-*-*-*-*-*-*-* <br>[PatchWork](https://lore.kernel.org/patchwork/patch/124065), [commit](https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=db64fe02258f1507e13fe5212a989922323685ce) |
|
||||
| 2013/03/13 | Joonsoo Kim <iamjoonsoo.kim@lge.com> | [remove vm_struct list management](https://lore.kernel.org/patchwork/cover/365284) | 删除了原本使用链表管理方式. | v2 ☑ 3.10-rc1 | [PatchWork v2,0/8](https://lore.kernel.org/patchwork/cover/365284)) |
|
||||
| 2021/03/17 | Nicholas Piggin <npiggin@gmail.com> | [huge vmalloc mappings](https://lore.kernel.org/patchwork/cover/1397495) | vmalloc 支持透明大页 | v13 ☑ [5.13-rc1](https://kernelnewbies.org/Linux_5.13#Memory_management) | [PatchWork v13,00/14](https://patchwork.kernel.org/project/linux-mm/cover/20210317062402.533919-1-npiggin@gmail.com/) |
|
||||
| 2019/01/03 | "Uladzislau Rezki (Sony)" <urezki@gmail.com> | [test driver to analyse vmalloc allocator](https://lore.kernel.org/patchwork/cover/1028793) | 实现一个驱动来帮助分析和测试 vmalloc | RFC v4 ☐ | [PatchWork RFC v4](https://patchwork.kernel.org/project/linux-mm/cover/20190103142108.20744-1-urezki@gmail.com) |
|
||||
| 2019/10/31 | Daniel Axtens <dja@axtens.net> | [kasan: support backing vmalloc space with real shadow memory](https://lore.kernel.org/patchwork/cover/1146684) | NA | v11 ☑ 5.5-rc1 | [PatchWork v11,0/4](https://lore.kernel.org/patchwork/cover/1146684) |
|
||||
| 2021/03/24 | "Matthew Wilcox (Oracle)" <willy@infradead.org> | [vmalloc: Improve vmalloc(4MB) performance](https://lore.kernel.org/patchwork/cover/1401688) | 加速 4MB vmalloc 分配. | v2 ☑ 5.13-rc1 | [PatchWork v2](https://lore.kernel.org/patchwork/cover/1401688) |
|
||||
|
||||
|
||||
### 2.3.2 VMALLOC 性能优化
|
||||
### 2.3.2 VMALLOC 中的 RBTREE 和 LIST 以及保护它们的锁
|
||||
-------
|
||||
|
||||
| 时间 | 作者 | 特性 | 描述 | 是否合入主线 | 链接 |
|
||||
|:----:|:----:|:---:|:----:|:---------:|:----:|
|
||||
| 2018/10/19 | "Uladzislau Rezki (Sony)" <urezki@gmail.com> | [improve vmalloc allocation](https://patchwork.kernel.org/project/linux-mm/cover/20181019173538.590-1-urezki@gmail.com) | 优化 VMALLOC | v1 ☑ [5.2-rc1](https://kernelnewbies.org/Linux_5.2#Memory_management) | [PatchWork](https://patchwork.kernel.org/project/linux-mm/cover/20181019173538.590-1-urezki@gmail.com) |
|
||||
| 2019/04/06 | "Uladzislau Rezki (Sony)" <urezki@gmail.com> | [improve vmap allocation](https://lore.kernel.org/patchwork/cover/1059021) | 优化 alloc_vmap_area(), 优化查找复杂度从 O(N) 下降为 O(logN) | v2 ☑ [5.2-rc7](https://kernelnewbies.org/Linux_5.2#Memory_management) | [PatchWork RFC](https://lore.kernel.org/patchwork/cover/1002038)<br>*-*-*-*-*-*-*-* <br>[PatchWork v4](https://patchwork.kernel.org/project/linux-mm/patch/20190406183508.25273-2-urezki@gmail.com), [PatchWork v2](https://lore.kernel.org/patchwork/cover/1059021) |
|
||||
| 2019/01/03 | "Uladzislau Rezki (Sony)" <urezki@gmail.com> | [test driver to analyse vmalloc allocator](https://lore.kernel.org/patchwork/cover/1028793) | 实现一个驱动来帮助分析和测试 vmalloc | RFC v4 ☐ | [PatchWork RFC v4](https://patchwork.kernel.org/project/linux-mm/cover/20190103142108.20744-1-urezki@gmail.com) |
|
||||
| 2021/03/17 | Nicholas Piggin <npiggin@gmail.com> | [huge vmalloc mappings](https://lore.kernel.org/patchwork/cover/1397495) | vmalloc 支持透明大页 | v13 ☑ [5.13-rc1](https://kernelnewbies.org/Linux_5.13#Memory_management) | [PatchWork v13,00/14](https://patchwork.kernel.org/project/linux-mm/cover/20210317062402.533919-1-npiggin@gmail.com/) |
|
||||
| 2021/03/24 | "Matthew Wilcox (Oracle)" <willy@infradead.org> | [vmalloc: Improve vmalloc(4MB) performance](https://lore.kernel.org/patchwork/cover/1401688) | 加速 4MB vmalloc 分配. | v2 ☑ 5.13-rc1 | [PatchWork v2](https://lore.kernel.org/patchwork/cover/1401688) |
|
||||
| 2008/07/28 | Nick Piggin <npiggin@suse.de> | [mm: vmap rewrite](https://lwn.net/Articles/304188) | 重写 vmap 分配器, 为 vmap_area 首次引入了红黑树组织(使用 vmap_area_lock 来保护). | RFC ☑ 2.6.28-rc1 | [PatchWork](https://lore.kernel.org/patchwork/cover/118352))<br>*-*-*-*-*-*-*-* <br>[PatchWork](https://lore.kernel.org/patchwork/patch/124065), [commit](https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=db64fe02258f1507e13fe5212a989922323685ce) |
|
||||
| 2011/03/22 | Nick Piggin <npiggin@suse.de> | [mm: vmap area cache](https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=89699605fe7cfd8611900346f61cb6cbf179b10a) | 为 vmalloc 分配器提供一个空闲区域缓存 free_vmap_cache, 它缓存了上次搜索和分配的空闲区域的节点, 下次分配和释放时, 都将从这里开始搜索, 避免了每次从 VMALLOC_START 搜索. 之前的实现中, 如果 vmalloc/vfree 频繁, 将从 VMALLOC_START 地址建立一个区段链, 每次都必须对其进行迭代, 这将是 O(n) 的开销. 在这个补丁之后, 搜索将从它结束的地方开始, 给出更接近于 O(1) 的平摊结果. 这一定程度上减少了rbtree操作的数量. 这解决了在 [db64fe02 mm: rewrite vmap layer](https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=db64fe02258f1507e13fe5212a989922323685ce) 中引入的惰性 vunmap TLB 清除导致的回归. 在 vunmapped 区段之后, 该补丁将在 vmap 分配器中保留区段, 直到可以在单个批处理中刷新的大量区段积累为止. | v2 ☑ 3.10-rc1 | [commit](https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=89699605fe7cfd8611900346f61cb6cbf179b10a) |
|
||||
| 2013/03/13 | Joonsoo Kim <iamjoonsoo.kim@lge.com> | [remove vm_struct list management](https://lore.kernel.org/patchwork/cover/365284) | 在 vmalloc 初始化后, 不再使用旧的 vmlist 链表管理方式. 使用由 vmap_area 组织成的 vmap_area_list 管理. | v2 ☑ 3.10-rc1 | [PatchWork v2,0/8](https://lore.kernel.org/patchwork/cover/365284)<br>*-*-*-*-*-*-*-* <br>[关注 commit](https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=4341fa454796b8a37efd5db98112524e85e7114e) |
|
||||
| 2016/04/15 | "Uladzislau Rezki (Sony)" <urezki@gmail.com> | [mm/vmalloc: Keep a separate lazy-free list](https://lore.kernel.org/patchwork/cover/669083) | 将惰性释放的 vmap_area 添加到一个单独的无锁空闲 vmap_purge_list, 这样我们就不必在每次清除时遍历整个列表. | v2 ☑ 4.7-rc1 | [PatchWork v1](https://lore.kernel.org/patchwork/cover/667471)<br>*-*-*-*-*-*-*-* <br>[PatchWork v2](https://lore.kernel.org/patchwork/cover/669083)<br>*-*-*-*-*-*-*-* <br>[commit](https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=80c4bd7a5e4368b680e0aeb57050a1b06eb573d8) |
|
||||
| 2019/04/06 | "Uladzislau Rezki (Sony)" <urezki@gmail.com> | [improve vmap allocation](https://lore.kernel.org/patchwork/cover/1059021) | [improve vmalloc allocation](https://patchwork.kernel.org/project/linux-mm/cover/20181019173538.590-1-urezki@gmail.com) 的 RESEND 和再版. 优化 alloc_vmap_area(), 优化查找复杂度从 O(N) 下降为 O(logN). 引入了 free_vmap_area 的 list 和 rbtree. | v3 ☑ [5.2-rc7](https://kernelnewbies.org/Linux_5.2#Memory_management) | [PatchWork RFC](https://lore.kernel.org/patchwork/cover/1002038)<br>*-*-*-*-*-*-*-* <br>[PatchWork v4](https://patchwork.kernel.org/project/linux-mm/patch/20190406183508.25273-2-urezki@gmail.com)<br>*-*-*-*-*-*-*-* <br>[PatchWork v3](https://lore.kernel.org/patchwork/cover/1057418) |
|
||||
| 2020/11/16 | "Uladzislau Rezki (Sony)" <urezki@gmail.com> | [mm/vmalloc: rework the drain logic](https://lore.kernel.org/patchwork/cover/1339347) | 当前的"惰性释放”模式至少存在两个问题: <br>1. 由于只有 vmap_purge_list 链表组织, 因此 vmap 区域的未排序的, 因此为了识别要耗尽的区域的 [min:max] 范围, 它需要扫描整个 vmap_purge_list 链表. 可能会耗费较多时间.<br>2. 作为下一步是关于合并所有片段与自由空间.这也是一种耗时的操作, 因为它必须遍历包含突出惰性区域的整个列表.
|
||||
|
||||
|
||||
|
||||
参见下面的“preemptirqsoff”跟踪程序, 它说明了高延迟.它是~ 24676 us.我们的音频和视频等工作负载受到了如此长的延迟的影响: | v1 ☑ 5.11-rc1 | [PatchWork](https://patchwork.kernel.org/project/linux-mm/patch/20201116220033.1837-2-urezki@gmail.com/) |
|
||||
| 2019/10/22 | Uladzislau Rezki (Sony) <urezki@gmail.com> | [mm/vmalloc: rework vmap_area_lock](https://lore.kernel.org/patchwork/cover/1143029) | 随着 5.2 版本 [improve vmap allocation](https://lore.kernel.org/patchwork/cover/1059021) 的合入, 全局的 vmap_area_lock 可以被拆分成两个锁: 一个用于分配部分(vmap_area_lock), 另一个用于回收(free_vmap_area_lock), 因为有两个不同的实体: "空闲数据结构"和"繁忙数据结构". 这和那后的减少锁争用, 允许在不同的 CPU 上并行执行"空闲"和"忙碌"树的操作. 但是需要注意的是, 分配/释放操作仍然存在依赖. 如果在不同的 CPU 上同时运行, 分配/释放操作仍然会相互干扰. | v1 ☑ 5.5-rc1 | [PatchWork v2,0/8](https://patchwork.kernel.org/project/linux-mm/patch/20191022155800.20468-1-urezki@gmail.com)<br>*-*-*-*-*-*-*-* <br>[commit](https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=e36176be1c3920a487681e37158849b9f50189c4) |
|
||||
|
||||
|
||||
### 2.3.1.3 lazy drain
|
||||
@@ -694,23 +705,12 @@ vmalloc 分配的内存并不是内核预先分配好的, 而是需要动态分
|
||||
|
||||
| 时间 | 作者 | 特性 | 描述 | 是否合入主线 | 链接 |
|
||||
|:----:|:----:|:---:|:----:|:---------:|:----:|
|
||||
| 2008/07/28 | Nick Piggin <npiggin@suse.de> | [mm: vmap rewrite](https://lwn.net/Articles/304188) | 重写 vmap 分配器 | RFC ☑ 2.6.28-rc1 | [PatchWork](https://lore.kernel.org/patchwork/cover/118352))<br>*-*-*-*-*-*-*-* <br>[PatchWork](https://lore.kernel.org/patchwork/patch/124065), [commit](https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=db64fe02258f1507e13fe5212a989922323685ce) |
|
||||
| 2020/11/16 | "Uladzislau Rezki (Sony)" <urezki@gmail.com> | [mm/vmalloc: rework the drain logic](https://lore.kernel.org/patchwork/cover/1339347) | NA | v1 ☑ 5.11-rc1 | [PatchWork](https://patchwork.kernel.org/project/linux-mm/patch/20201116220033.1837-2-urezki@gmail.com/) |
|
||||
| 2008/07/28 | Nick Piggin <npiggin@suse.de> | [mm: vmap rewrite](https://lwn.net/Articles/304188) | 重写 vmap 分配器 | RFC ☑ 2.6.28-rc1 | [PatchWork](https://lore.kernel.org/patchwork/cover/118352)<br>*-*-*-*-*-*-*-* <br>[PatchWork](https://lore.kernel.org/patchwork/patch/124065), [commit](https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=db64fe02258f1507e13fe5212a989922323685ce) |
|
||||
| 2016/04/15 | "Uladzislau Rezki (Sony)" <urezki@gmail.com> | [mm/vmalloc: Keep a separate lazy-free list](https://lore.kernel.org/patchwork/cover/669083) | 之前惰性释放的 vmap_area 也是放到 vmap_area_list 中, 但是使用 LAZY_FREE 标记. 每次处理时, 需要遍历整个 vmap_area_list, 然后将 LAZY_FREE 的节点添加到一个临时的 purge_list 中进行释放. 当混合使用大量 vmalloc 和 set_memory_*()(它调用vm_unmap_aliases())时, 触发了由于每次调用都要遍历整个 vmap_area_list 而导致性能严重下降的情况. 因此将惰性释放的 vmap_area 添加到一个单独的无锁空闲列表, 这样我们就不必在每次清除时遍历整个列表. | v2 ☑ 4.7-rc1 | [PatchWork v1](https://lore.kernel.org/patchwork/cover/667471)<br>*-*-*-*-*-*-*-* <br>[PatchWork v2](https://lore.kernel.org/patchwork/cover/669083)<br>*-*-*-*-*-*-*-* <br>[commit](https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=80c4bd7a5e4368b680e0aeb57050a1b06eb573d8) |
|
||||
| 2018/08/23 | "Uladzislau Rezki (Sony)" <urezki@gmail.com> | [minor mmu_gather patches](https://lore.kernel.org/patchwork/cover/976960) | NA | v1 ☑ 5.11-rc1 | [PatchWork](https://lore.kernel.org/patchwork/cover/976960) |
|
||||
|
||||
|
||||
|
||||
### 2.3.1.4 vmalloc 锁优化
|
||||
-------
|
||||
|
||||
| 时间 | 作者 | 特性 | 描述 | 是否合入主线 | 链接 |
|
||||
|:----:|:----:|:---:|:----:|:---------:|:----:|
|
||||
| 2008/07/28 | Nick Piggin <npiggin@suse.de> | [mm: vmap rewrite](https://lwn.net/Articles/304188) | 重写 vmap 分配器, 当前关注的是引入了红黑树组织的 vmap_area 后, 使用 vmap_area_lock 来保护. | RFC ☑ 2.6.28-rc1 | [PatchWork](https://lore.kernel.org/patchwork/cover/118352)<br>*-*-*-*-*-*-*-* <br>[PatchWork](https://lore.kernel.org/patchwork/patch/124065), [commit](https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=db64fe02258f1507e13fe5212a989922323685ce) |
|
||||
| 2013/03/13 | Joonsoo Kim <iamjoonsoo.kim@lge.com> | [remove vm_struct list management](https://lore.kernel.org/patchwork/cover/365284) | 在 vmalloc 初始化后, 不再使用旧的 vmlist 链表管理方式. 使用由 vmap_area 组织成的 vmap_area_list 管理. | v2 ☑ 3.10-rc1 | [PatchWork v2,0/8](https://lore.kernel.org/patchwork/cover/365284)<br>*-*-*-*-*-*-*-* <br>[关注 commit](https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=4341fa454796b8a37efd5db98112524e85e7114e) |
|
||||
| 2019/10/22 | Uladzislau Rezki (Sony) <urezki@gmail.com> | [remove vm_struct list management](https://lore.kernel.org/patchwork/cover/1143029) | 随着 5.2 版本 [improve vmap allocation](https://lore.kernel.org/patchwork/cover/1059021) 的合入, 全局的 vmap_area_lock 可以被拆分成两个锁: 一个用于分配部分(vmap_area_lock), 另一个用于回收(free_vmap_area_lock), 因为有两个不同的实体: "空闲数据结构"和"繁忙数据结构". 这和那后的减少锁争用, 允许在不同的 CPU 上并行执行"空闲"和"忙碌"树的操作. 但是需要注意的是, 分配/释放操作仍然存在依赖. 如果在不同的 CPU 上同时运行, 分配/释放操作仍然会相互干扰. | v1 ☑ 5.5-rc1 | [PatchWork v2,0/8](https://patchwork.kernel.org/project/linux-mm/patch/20191022155800.20468-1-urezki@gmail.com)<br>*-*-*-*-*-*-*-* <br>[commit](https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=e36176be1c3920a487681e37158849b9f50189c4) |
|
||||
|
||||
|
||||
|
||||
|
||||
### 2.3.2 连续内存分配器(CMA)
|
||||
-------
|
||||
|
||||
@@ -661,7 +661,7 @@ NUMA 机器一个重要特性就是不同 node 之间的内存访问速度有差
|
||||
| 2019/10/18 | Mel Gorman | [Reconcile NUMA balancing decisions with the load balancer](https://linuxplumbersconf.org/event/4/contributions/480) | NUMA Balancing 和 Load Blance 经常会做出一些相互冲突的决策(任务放置和页面迁移), 导致不必要的迁移, 这组补丁减少了冲突 | v4 ☑ | [LWN](https://lwn.net/Articles/793427), [PatchWork](https://lore.kernel.org/patchwork/cover/1199507), [lkml](https://lkml.org/lkml/2019/10/18/676) |
|
||||
| 2020/2/21 | Vincent Guittot | [remove runnable_load_avg and improve group_classify](https://lore.kernel.org/patchwork/cover/1198654) | 重构组调度的 PELT 跟踪, 在每次更新平均负载的时候, 更新整个 CFS_RQ 的平均负载| V10 ☑ 5.7 | [PatchWork](https://lore.kernel.org/patchwork/cover/1198654), [lkml](https://lkml.org/lkml/2020/2/21/1386) |
|
||||
| 2020/02/03 | Mel Gorman | [Accumulated fixes for Load/NUMA Balancing reconcilation series](https://lore.kernel.org/patchwork/cover/1203922) | 解决一个负载平衡问题 | | [PatchWork](https://lore.kernel.org/patchwork/cover/1203922) |
|
||||
| 2020/02/03 | Mel Gorman | [Revisit NUMA imbalance tolerance and fork balancing](https://lore.kernel.org/patchwork/cover/1340184) | 解决一个负载平衡问题 | 之前解决 NUMA Balancing 和 Load Balancing 被调和时, 尝试允许在一定程度的不平衡, 但是也引入了诸多问题. 因此当时限制不平衡只允许在几乎空闲的 NUMA 域. 现在大多数问题都已经解决掉了, 现在允许不平衡扩大一定的范围. 同时该补丁还解决了 fork 时候 balance 的问题 | [PatchWork](https://lore.kernel.org/patchwork/cover/1340184) |
|
||||
| 2020/02/03 | Mel Gorman | [Revisit NUMA imbalance tolerance and fork balancing](https://lore.kernel.org/patchwork/cover/1340184) | 解决一个负载平衡问题<br>之前解决 NUMA Balancing 和 Load Balancing 被调和时, 尝试允许在一定程度的不平衡, 但是也引入了诸多问题. 因此当时限制不平衡只允许在几乎空闲的 NUMA 域. 现在大多数问题都已经解决掉了, 现在允许不平衡扩大一定的范围. 同时该补丁还解决了 fork 时候 balance 的问题 | RFC ☑ 5.11-rc1 | [PatchWork](https://lore.kernel.org/patchwork/cover/1340184) |
|
||||
|
||||
|
||||
此外 load_balance 上还有一些接口层次的变化
|
||||
@@ -689,7 +689,7 @@ Vincent Guittot 深耕与解决 load_balance 各种疑难杂症和不均衡状
|
||||
| 时间 | 作者 | 特性 | 描述 | 是否合入主线 | 链接 |
|
||||
|:----:|:----:|:---:|:---:|:----------:|:----:|
|
||||
| 2017/12/18 | Vincent Guittot | [sched/fair: Improve fairness between cfs tasks](https://lore.kernel.org/patchwork/cover/1308748) | 当系统没有足够的周期用于所有任务时, 调度器必须确保在CFS任务之间公平地分配这些cpu周期. 某些用例的公平性不能通过在系统上静态分配任务来解决, 需要对系统进行周期性的再平衡但是, 这种动态行为并不总是最优的, 也不总是能够确保CPU绑定的公平分配. <br>这组补丁通过减少选择可迁移任务的限制来提高公平性. 这个更改可以降低不平衡阈值, 因为 1st LB将尝试迁移完全匹配不平衡的任务. | v1 ☑ [5.10-rc1](https://kernelnewbies.org/Linux_5.10#Memory_management) | [PatchWork](https://lore.kernel.org/patchwork/cover/1308748) |
|
||||
| 2021/01/06 | Vincent Guittot | [Reduce number of active LB](https://lore.kernel.org/patchwork/cover/1361676) | 减少 ACTIVE LOAD_BALANCE 的次数 | v2 ☑ 5.10-rc1 | [PatchWork](https://lore.kernel.org/patchwork/cover/1361676) |
|
||||
| 2021/01/06 | Vincent Guittot | [Reduce number of active LB](https://lore.kernel.org/patchwork/cover/1361676) | 减少 ACTIVE LOAD_BALANCE 的次数 | v2 ☑ 5.12-rc1 | [PatchWork](https://lore.kernel.org/patchwork/cover/1361676) |
|
||||
| 2021/01/06 | Vincent Guittot | [move update blocked load outside newidle_balance](https://lore.kernel.org/patchwork/cover/1383963) |oel报告了newidle_balance中的抢占和irq关闭序列很长, 因为大量的CPU cgroup正在使用, 并且需要更新. 这个补丁集将更新移动到newidle_imblance之外. 这使我们能够在更新期间提前中止. 在选择并进入空闲状态之前, 触发CPU空闲线程中的统计数据更新, 而不是踢一个正常的ILB来唤醒已经处于空闲状态的CPU. | v2 ☑ 5.10-rc1 | [PatchWork](https://lore.kernel.org/patchwork/cover/1361676) |
|
||||
|
||||
此外 update_blocked_averages 也是内核调度器一个非常耗时的操作.
|
||||
@@ -806,7 +806,7 @@ TencentOS-kernel 回合了主线 wake_affine 中几个优化迁移的补丁, 可
|
||||
| 2019/07/08 | Parth Shah | [Optimize the idle CPU search](https://lore.kernel.org/patchwork/patch/1098092) | 通过标记 idle_cpu 来降低 select_idle_sibling/select_idle_cpu 的搜索开销 | | |
|
||||
| 2020/04/15 | Valentin Schneider | [sched: Streamline select_task_rq() & select_task_rq_fair()](https://lore.kernel.org/patchwork/patch/1224733) | 选核流程上的重构和优化, 当然除此之外还做了其他操作, 比如清理了 sd->flags 信息, 甚至 sysfs 接口都变成只读了 | v3 ☑ 5.8-rc1 | [PatchWork](https://lore.kernel.org/patchwork/cover/1224733) |
|
||||
| 2021/03/26 | Rik van Riel | [Throttle select_idle_sibling when a target domain is overloaded](https://lore.kernel.org/patchwork/patch/1213189) | 这是 CPU/NUMA 负载平衡器协调的后续工作. 补丁中做了如下的优化<br>1. 补丁1-2 添加了 schedstats 来跟踪 select_idle_sibling() 的效率. 默认是禁用的<br>2. 补丁3 是一个细微的优化, 如果一个 CPU 已经找到, 避免清除部分 cpumask.<br>3. 补丁 4 跟踪在 select_idle_cpu() 期间域是否出现重载, 以便将来的扫描可以在必要时尽早终止. 这减少了在域超载时无用扫描的运行队列数量. | v2 ☐ | [PatchWork](https://lore.kernel.org/patchwork/patch/1213189) |
|
||||
| 2020/12/03 | Mel Gorman | [Reduce time complexity of select_idle_sibling](https://lore.kernel.org/patchwork/cover/1348877) | 通过自己完善的 schedstat 的统计信息, 发现 select_idle_XXX 中不合理的地方(提高了p->recent_used_cpu的命中率. 以减少扫描开销, 同时如果在扫描a时发现了一个候选, 那么补丁4将返回一个空闲的候选免费的核心等), 降低搜索开销 | RFC v3 | 这组补丁其实有很多名字, 作者发了几版本之后, 不断重构, 也改了名字<br>*-*-*-*-*-*-*-* <br>2020/12/03 v1[Reduce time complexity of select_idle_sibling](https://lore.kernel.org/patchwork/cover/1348877)<br>*-*-*-*-*-*-*-* <br>2020/12/07 v2[Reduce scanning of runqueues in select_idle_sibling](https://lore.kernel.org/patchwork/cover/1350876)<br>*-*-*-*-*-*-*-* <br>2020/12/07 v3[Reduce worst-case scanning of runqueues in select_idle_sibling](https://lore.kernel.org/patchwork/patch/1350248)<br>*-*-*-*-*-*-*-* <br>2020/12/08 v4[Reduce scanning of runqueues in select_idle_sibling](https://lore.kernel.org/patchwork/patch/1350877) |
|
||||
| 2020/12/03 | Mel Gorman | [Reduce time complexity of select_idle_sibling](https://lore.kernel.org/patchwork/cover/1348877) | 通过自己完善的 schedstat 的统计信息, 发现 select_idle_XXX 中不合理的地方(提高了p->recent_used_cpu的命中率. 以减少扫描开销, 同时如果在扫描a时发现了一个候选, 那么补丁4将返回一个空闲的候选免费的核心等), 降低搜索开销 | RFC v3 ☐ | 这组补丁其实有很多名字, 作者发了几版本之后, 不断重构, 也改了名字<br>*-*-*-*-*-*-*-* <br>2020/12/03 [Reduce time complexity of select_idle_sibling RFC,00/10](https://lore.kernel.org/patchwork/cover/1348877)<br>*-*-*-*-*-*-*-* <br>2020/12/07 RFC [Reduce worst-case scanning of runqueues in select_idle_sibling 0/4](https://lore.kernel.org/patchwork/cover/1350248)<br>*-*-*-*-*-*-*-* <br>2020/12/08 [Reduce scanning of runqueues in select_idle_sibling 0/4](https://lore.kernel.org/patchwork/patch/1350876)<br>*-*-*-*-*-*-*-* <br>后面换标题名重发 [Scan for an idle sibling in a single pass](https://lore.kernel.org/patchwork/cover/1371921) |
|
||||
| 2020/12/14 | Peter Zijlstra | [select_idle_sibling() wreckage](https://lore.kernel.org/patchwork/cover/1353496) | 重构 SIS_PROP 的逻辑, 重新计算 CPU 的扫描成本, 同时归一 select_idle_XXX 中对 CPU 的遍历, 统一选核的搜索逻辑来降低开销, 提升性能 | RFC | [PatchWork](https://lore.kernel.org/patchwork/cover/1353496), [LKML](https://lkml.org/lkml/2020/12/14/560) |
|
||||
| 2020/12/08 | Mel Gorman | [Scan for an idle sibling in a single pass](https://lore.kernel.org/patchwork/cover/1371921) | 基于上面 Peter 的补丁的思路进行了重构, 减少 select_idle_XXX 的开销<br>1. 优化了 IDLE_CPU 扫描深度的计算方法<br>2. 减少了 CPU 的遍历次数, 重构了 sched_idle_XXX 函数<br>3. 将空闲的核心扫描调节机制转换为SIS_PROP, 删除了 SIS_PROP_CPU<br>3. | v5 ☑ 5.12-rc1 | [PatchWork](https://lore.kernel.org/patchwork/cover/1371921) |
|
||||
| 2021/03/26 | Rik van Riel | [sched/fair: bring back select_idle_smt, but differently](https://lore.kernel.org/patchwork/patch/1402916) | Mel Gorman 上面的补丁在 9fe1f127b913("sched/fair: Merge select_idle_core/cpu()") 中做了一些出色的工作, 从而提高了内核查找空闲cpu的效率, 可以有效地降低任务等待运行的时间. 但是较多的均衡和迁移, 减少的局部性和相应的 L2 缓存丢失的增加会带来一些负面的效应. 这个补丁重新将 `select_idle_smt` 引入回来, 并做了优化, 修复了性能回归, 在搜索所有其他 CPU 之前, 检查 prev 的兄弟 CPU 是否空闲. | v2 ☐ | [PatchWork](https://lore.kernel.org/patchwork/patch/1402916) |
|
||||
|
||||
Reference in New Issue
Block a user