Pavel Emelyanov | 0f8975e | 2013-07-03 15:01:20 -0700 | [diff] [blame] | 1 | SOFT-DIRTY PTEs |
| 2 | |
| 3 | The soft-dirty is a bit on a PTE which helps to track which pages a task |
| 4 | writes to. In order to do this tracking one should |
| 5 | |
| 6 | 1. Clear soft-dirty bits from the task's PTEs. |
| 7 | |
| 8 | This is done by writing "4" into the /proc/PID/clear_refs file of the |
| 9 | task in question. |
| 10 | |
| 11 | 2. Wait some time. |
| 12 | |
| 13 | 3. Read soft-dirty bits from the PTEs. |
| 14 | |
| 15 | This is done by reading from the /proc/PID/pagemap. The bit 55 of the |
| 16 | 64-bit qword is the soft-dirty one. If set, the respective PTE was |
| 17 | written to since step 1. |
| 18 | |
| 19 | |
| 20 | Internally, to do this tracking, the writable bit is cleared from PTEs |
| 21 | when the soft-dirty bit is cleared. So, after this, when the task tries to |
| 22 | modify a page at some virtual address the #PF occurs and the kernel sets |
| 23 | the soft-dirty bit on the respective PTE. |
| 24 | |
| 25 | Note, that although all the task's address space is marked as r/o after the |
| 26 | soft-dirty bits clear, the #PF-s that occur after that are processed fast. |
| 27 | This is so, since the pages are still mapped to physical memory, and thus all |
| 28 | the kernel does is finds this fact out and puts both writable and soft-dirty |
| 29 | bits on the PTE. |
| 30 | |
Cyrill Gorcunov | d9104d1 | 2013-09-11 14:22:24 -0700 | [diff] [blame] | 31 | While in most cases tracking memory changes by #PF-s is more than enough |
| 32 | there is still a scenario when we can lose soft dirty bits -- a task |
| 33 | unmaps a previously mapped memory region and then maps a new one at exactly |
| 34 | the same place. When unmap is called, the kernel internally clears PTE values |
| 35 | including soft dirty bits. To notify user space application about such |
| 36 | memory region renewal the kernel always marks new memory regions (and |
| 37 | expanded regions) as soft dirty. |
Pavel Emelyanov | 0f8975e | 2013-07-03 15:01:20 -0700 | [diff] [blame] | 38 | |
| 39 | This feature is actively used by the checkpoint-restore project. You |
| 40 | can find more details about it on http://criu.org |
| 41 | |
| 42 | |
| 43 | -- Pavel Emelyanov, Apr 9, 2013 |