On Mon, 8 Mar 2010, Oliver Neukum wrote: > Am Sonntag, 7. März 2010 21:45:50 schrieb Alan Stern: > > > len and sg_page(sg) aren't particularly useful, as far as I can see. > > > > sg and src need to be compared with the original values when the > > scatterlist was created. In invoke_transport() print out the values of > > scsi_sglist(srb), srb->sense_buffer, &ses->sense_sgl, and > > sg_virt(&ses->sense_sgl). Do this after the call to > > scsi_eh_prep_cmnd(). > Mar 8 19:06:10 vanamonde kernel: usb-storage: Issuing auto-REQUEST_SENSE > Mar 8 19:06:10 vanamonde kernel: usb_stor_invoke_transport - sglist ffff880135b37dc8 sense buffer ffff8801363db940 sense sgl ffff8801363db940 sg_virt ffffea00043dd7ea page link 0000000087654321 dmaaddr ffffea0 0043dd7ea > Mar 8 19:06:10 vanamonde kernel: usb-storage: Bulk Command S 0x43425355 T 0x118 L 18 F 128 Trg 0 LUN 0 CL 6 > Mar 8 19:06:10 vanamonde kernel: usb-storage: usb_stor_bulk_transfer_buf: xfer 31 bytes > Mar 8 19:06:10 vanamonde kernel: mon_text_get_data: num_sgs == 0: src ffff880037807000 There are several problems here. Are you sure your printk line really matches the right value with each label? To begin with, scsi_sglist(srb) is defined as srb->sdb.table.sgl, and in scsi_eh_prep_cmnd() that field is set to &ses->sense_sgl. Hence the sglist value and the sense sgl values above should be the same, but they aren't. Instead, &ses->sense_sgl is the same as srb->sense_buffer! I can't imagine how that would happen. (Also, sg_virt(&ses->sense_sgl) should be the same as srb->sense_buffer; the value is assigned when scsi_eh_prep_cmnd() calls sg_init_one(). Instead its value looks random -- it's the same as the DMA address, with no relation to the actual buffer address. Are you using some sort of iotlb?) You might want to look at the assembly code generated by the compiler for scsi_eh_prep_cmnd(), just to see if this could be a compiler bug. It certainly looks very suspicious. Alan Stern -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
