On Tue, May 5, 2009 at 5:12 PM, FUJITA Tomonori <fujita.tomonori@xxxxxxxxxxxxx> wrote: > On Tue, 5 May 2009 11:45:37 -0700 > adam radford <aradford@xxxxxxxxx> wrote: > >> This patch for the 3w-9xxx driver is for 2.6.30-rc5. >> >> This patch fixes the following regression the occurred during the >> scsi_dma_map()/unmap() changes: > > Yeah, sorry about that. > >> 3w-9xxx 0001:45:00.0: DMA-API: device driver tries to free DMA memory >> it has not allocated [device address=0x0000000000000000] [size=36 >> bytes] > > Might be better to put the pointer to report this bug? > > http://marc.info/?l=linux-kernel&m=124141488730478&w=2 > > >> James, could you please apply this? >> >> Thanks, >> >> -Adam >> >> Signed-off-by: Adam Radford <aradford@xxxxxxxxx> > > The patch looks correct. Let's add 'Cc: stable@xxxxxxxxxx' here so > that this patch will be sent to stable trees. > > > btw, can we simplify the dma buffer handling by removing the usage of > the generi_buffer_virt (always use scatter gather buffers)? > > diff --git a/drivers/scsi/3w-9xxx.c b/drivers/scsi/3w-9xxx.c > index 8b7983a..4c7e80e 100644 > --- a/drivers/scsi/3w-9xxx.c > +++ b/drivers/scsi/3w-9xxx.c > @@ -1847,29 +1847,19 @@ static int twa_scsiop_execute_scsi(TW_Device_Extension *tw_dev, int request_id, > /* Map sglist from scsi layer to cmd packet */ > > if (scsi_sg_count(srb)) { > - if ((scsi_sg_count(srb) == 1) && > - (scsi_bufflen(srb) < TW_MIN_SGL_LENGTH)) { > - if (srb->sc_data_direction == DMA_TO_DEVICE || > - srb->sc_data_direction == DMA_BIDIRECTIONAL) > - scsi_sg_copy_to_buffer(srb, > - tw_dev->generic_buffer_virt[request_id], > - TW_SECTOR_SIZE); > - command_packet->sg_list[0].address = TW_CPU_TO_SGL(tw_dev->generic_buffer_phys[request_id]); > - command_packet->sg_list[0].length = cpu_to_le32(TW_MIN_SGL_LENGTH); > - } else { > - sg_count = twa_map_scsi_sg_data(tw_dev, request_id); > - if (sg_count == 0) > - goto out; > + sg_count = twa_map_scsi_sg_data(tw_dev, request_id); > + if (sg_count == 0) > + goto out; > > - scsi_for_each_sg(srb, sg, sg_count, i) { > - command_packet->sg_list[i].address = TW_CPU_TO_SGL(sg_dma_address(sg)); > - command_packet->sg_list[i].length = cpu_to_le32(sg_dma_len(sg)); > - if (command_packet->sg_list[i].address & TW_CPU_TO_SGL(TW_ALIGNMENT_9000_SGL)) { > - TW_PRINTK(tw_dev->host, TW_DRIVER, 0x2e, "Found unaligned sgl address during execute scsi"); > - goto out; > - } > + scsi_for_each_sg(srb, sg, sg_count, i) { > + command_packet->sg_list[i].address = TW_CPU_TO_SGL(sg_dma_address(sg)); > + command_packet->sg_list[i].length = cpu_to_le32(sg_dma_len(sg)); > + if (command_packet->sg_list[i].address & TW_CPU_TO_SGL(TW_ALIGNMENT_9000_SGL)) { > + TW_PRINTK(tw_dev->host, TW_DRIVER, 0x2e, "Found unaligned sgl address during execute scsi"); > + goto out; > } > } > + > command_packet->sgl_entries__lunh = cpu_to_le16(TW_REQ_LUN_IN((srb->device->lun >> 4), scsi_sg_count(tw_dev->srb[request_id]))); > } > } else { > I wouldn't do that. The generic_buffer_virt stuff is there because the hardware can't DMA < 512 bytes, so we sg_copy_to/from_buffer() there, then dma from generic_buffer_virt. -Adam -- 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
