On Oct. 25, 2007, 17:40 +0200, Linus Torvalds <torvalds@linux-foundation.org> wrote:There are a number of indicators that need to be kept in sync, depending on the usage. The number of entries is set when it is allocated and is currently needed to free it up (note that in the sgtable "sketch" James proposed we saved the sg pool index in the sgtable header and used it to free each chunk to the right pool). The number of entries is then used in many places to scan the list, however, after the sg list is dma mapped, the dma mapping may be shorter than the original sg list when multiple pages are coalesced and there we need to defer to use the dma_length (plus the number of entries) to determine the end of the list. IMO I think that the byte count can be used authoritatively to scan the contents of the sg list either before or after dma mapping while the number of entries is relevant to walking the list in a context free manner (i.e. to go over all the entries that were allocated) -
| Tarkan Erimer | Re: Dual-Licensing Linux Kernel with GPL V2 and GPL V3 |
| Greg Kroah-Hartman | [PATCH 006/196] Chinese: add translation of oops-tracing.txt |
| Eric Sandeen | Re: [RFC] Heads up on sys_fallocate() |
| YOSHIFUJI Hideaki / | request_module: runaway loop modprobe net-pf-1 (is Re: Linux 2.6.21-rc1) |
git: | |
| Gerrit Renker | [PATCH 0/37] dccp: Feature negotiation - last call for comments |
| Jarek Poplawski | Re: [PATCH] pkt_sched: Destroy gen estimators under rtnl_lock(). |
| Ben Greear | Re: MACVLANs really best solution? How about a bridge with multiple bridge virtual... |
| Rafael J. Wysocki | 2.6.29-rc8: Reported regressions from 2.6.28 |
