I was not referring to the page cache. The block devices still go through VFS, and the blocks get cached by the VFS layer.
I can't say I have read the latest kernel source, so I could be wrong if this changed in 2.6. -----Original Message----- From: Linux on 390 Port [mailto:[EMAIL PROTECTED] Behalf Of Carsten Otte Sent: Friday, August 22, 2008 3:18 AM To: [email protected] Subject: Re: Root filesystem error switches to ReadOnly Fargusson.Alan wrote: > Doing a dd on a device goes through the cache, so in this case the cache > can't be the problem. This is not the case. In fact, the page cache is indexed by address spaces and each file as well as each block device has its own address space. Consequently, Linux cannot tell that you're reading/writing the same block on disk via the device node that is already in the cache because it belongs to a file of a mounted filesystem that relies on there. The only safe way of doing an online snapshot of a mounted file system is the dm-snapshot target. Flashcopy and DD don't do the trick. ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 __________________________________________________________________________________________________ CONFIDENTIALITY NOTICE: This email from the State of California is for the sole use of the intended recipient and may contain confidential and privileged information. Any unauthorized review or use, including disclosure or distribution, is prohibited. If you are not the intended recipient, please contact the sender and destroy all copies of this email. ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
