Differences
This shows you the differences between two versions of the page.
| Both sides previous revision Previous revision Next revision | Previous revision | ||
| tmpfs [2007/07/05 15:52] – HowardSu | tmpfs [2022/09/19 01:33] (current) – Remove ACL: not 'devs only' for a long time, all:read is already default. Add Intro & move ToC KubilayKocak | ||
|---|---|---|---|
| Line 1: | Line 1: | ||
| - | & | + | ====== TMPFS (Temporary Filesystem) ====== |
| - | ---- | + | ====== Introduction ====== |
| - | TMPFS is now in -Current tree under /src/sys/fs/tmpfs. | + | %%TmpFS%% was first developed by Julio M. Merino Vidal for %%NetBSD%% as a %%SoC%% project: [[http://netbsd-soc.sourceforge.net/projects/tmpfs/|Efficient memory file-system]]. Rohit Jalan ported it from %%NetBSD%% to %%FreeBSD%%. Now HowardSu and Glen Leeder (< |
| - | ====== | + | ====== |
| - | %%TmpFS%% | + | |
| + | tmpfs is included in Release since 7-Release. A typical usage is use tmpfs as /tmp mount point. | ||
| + | - Change the /tmp mount point in /etc/fstab to: **tmpfs /tmp tmpfs rw, | ||
| + | | ||
| + | - **mount /tmp** | ||
| + | |||
| + | ====== TODO ====== | ||
| + | |||
| + | * Refine | ||
| + | * Support unionfs, tmpfs can be used as an upper filesystem. | ||
| + | * Extra attribute support | ||
| + | |||
| + | ====== Testing Status ====== | ||
| + | |||
| + | ===== Regression Testing ===== | ||
| - | ==== Regression Testing ==== | ||
| We port %%NetBSD%% regression test cases. | We port %%NetBSD%% regression test cases. | ||
| - | All tests sucessful. | + | All tests successful. |
| + | |||
| + | ===== POSIX Correctness Testing ===== | ||
| - | ==== POSIX Correctness Testing ==== | ||
| All tests successful. | All tests successful. | ||
| Files=184, Tests=3353, 823 wallclock secs (29.94 cusr + 368.38 csys = 398.31 CPU) | Files=184, Tests=3353, 823 wallclock secs (29.94 cusr + 368.38 csys = 398.31 CPU) | ||
| - | ==== Stress Testing ==== | + | ===== Stress Testing ===== |
| Survive after fsstress -n 10000 -p 50. Tested under AMD64/i386. | Survive after fsstress -n 10000 -p 50. Tested under AMD64/i386. | ||
| - | Any presure | + | Any pressure |
| - | ==== Buildworld testing ==== | + | ===== Buildworld testing ===== |
| - | We sucessfully build world to a tmpfs in 4-CPU box with -j32. | + | |
| - | ==== fsx Testing ==== | + | We successfully build world to a tmpfs in 4-CPU box with -j32. |
| - | fsx(tools/ | + | |
| - | 1. The hole create by truncate or extend write is not zeroed correctly. | + | |
| - | 2. Some write data will lost due to unknow issue. | + | |
| - | I suppose all these bugs are related | + | ===== fsx Testing ===== |
| + | |||
| + | fsx(tools/ | ||
| + | |||
| + | We passed fsx testing now. | ||
| + | |||
| + | ===== Performance Testing ===== | ||
| - | ==== Performance Testing ==== | ||
| The vacache code is removed. sf_buf is used for data I/O. The performance will depends on sf_buf implementation in different platform. In Amd64, sf_buf is zero copy however i386 is not. So you will notice a performance difference. | The vacache code is removed. sf_buf is used for data I/O. The performance will depends on sf_buf implementation in different platform. In Amd64, sf_buf is zero copy however i386 is not. So you will notice a performance difference. | ||
| + | |||
| + | ---- | ||
| + | |||
| + | {{tag> | ||