On Sat, Jun 13, 2009 at 1:10 PM, Joerg Roedel<joro@xxxxxxxxxx> wrote: > On Fri, Jun 12, 2009 at 04:51:01PM +0200, Torsten Kaiser wrote: >> On Fri, Jun 12, 2009 at 4:13 PM, Joerg Roedel<joerg.roedel@xxxxxxx> wrote: >> > On Thu, Jun 11, 2009 at 07:38:47PM +0200, Torsten Kaiser wrote: >> >> DMA-API: device driver tries to free DM >> >> A memory it has not allocated [device address=0x000000011e4c3000] >> >> [size=4096 bytes] >> > >> > Hmm, looking again over the code I've seen that the ref >> > dma_debug_entries are not alway filled with all necessary information >> > for best-fit. Can you please try if you still get false warnings when >> > you apply the two patches attached instead of the one I sent yesterday? >> >> Hmm... There isn't a get_nr_mapped_entries() in 2.6.30. >> Are these patches again the current linus tree? > > Indeed. These patches are against tip/core/iommu but should also apply > against current linus/master too. OK, I will try 2.6.30-git6 with and without these patches. >> For unfilled information I wanted to complain that for non-sg mapping >> sg_call_ents never gets filled. But as dma_entry_alloc() does a >> memset() to clean the entrys it should not matter. >> And what about dma_debug_entry.paddr? Should this also be compared? > > paddr shouldn't matter. We also can't compare against it because it is > not a parameter we get in the unmap path (not always). check_unmap() checks it and sg_call_ents are also optional. Torsten -- To unsubscribe from this list: send the line "unsubscribe linux-scsi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
