On Fri, Oct 05, 2007 at 12:51:56PM -0400, der Mouse wrote:To really prevent this, you need an MMU between the expansion bus and the main memory. While some systems have been built that way for a long time (e.g. sparc64) others have not: this feature only really began to appear on i386 systems as the AGP GART, which isn't really flexible enough to offer the protection needed in this case, and has only really been present on any common i386 or amd64 system within the past couple of years -- and AMD and Intel do it differently, and we don't support either; nor do most other operating systems. In sum, PCI DMA on most systems with PCI is direct to host memory addresses and there's no protection mechanism in between. A malicious device can write whatever part of host memory it cares to.
| Greg KH | [GIT PATCH] driver core patches against 2.6.24 |
| James Bottomley | Re: Announce: Linux-next (Or Andrew's dream :-)) |
| Andrew Morton | echo mem > /sys/power/state |
| Peter Zijlstra | [PATCH 00/23] per device dirty throttling -v8 |
git: | |
| David Miller | Re: [PATCH] pkt_sched: Destroy gen estimators under rtnl_lock(). |
| David Miller | [GIT]: Networking |
| Gerrit Renker | [PATCH 18/37] dccp: Support for Mandatory options |
| Michael S. Tsirkin | Re: [RFC PATCH v2 03/19] vbus: add connection-client helper infrastructure |
| NeilBrown | [PATCH 00/18] Assorted md patches headed for 2.6.30 |
| Justin Piszcz | General question (scheduler) with SSDs? |
| Neil Brown | Re: Any hope for a 27 disk RAID6+1HS array with four disks reporting "No md superb... |
| Ryan Wagoner | High IO Wait with RAID 1 |
