From mboxrd@z Thu Jan 1 00:00:00 1970 From: Pat LaVarre Subject: Re: SATA ATAPI work in progress Date: 14 May 2004 17:47:33 -0600 Sender: linux-ide-owner@vger.kernel.org Message-ID: <1084578453.5343.26.camel@patibmrh9> References: <1084393233.3999.2.camel@patibmrh9> <40A28BB6.7090204@pobox.com> <1084403654.3196.31.camel@patibmrh9> <40A3E595.8000003@pobox.com><108455 9 000.4017.52.camel@patibmrh9> <40A51604.8070208@pobox.com> Mime-Version: 1.0 Content-Type: text/plain Content-Transfer-Encoding: 7bit Return-path: Received: from email-out1.iomega.com ([147.178.1.82]:26240 "EHLO email.iomega.com") by vger.kernel.org with ESMTP id S264571AbUENXr6 (ORCPT ); Fri, 14 May 2004 19:47:58 -0400 In-Reply-To: <40A51604.8070208@pobox.com> List-Id: linux-ide@vger.kernel.org To: Jeff Garzik Cc: linux-ide@vger.kernel.org Applying patch.[12345] in order to linux-2.6.6-bk1/ here yielded: kernel: ata_scsi_dump_cdb: CDB (1:1,15,0) 12 00 00 00 24 00 00 00 46 kernel: ata_scsi_dump_cdb: CDB (2:0,0,0) 12 00 00 00 24 00 00 00 46 kernel: ata_scsi_translate: ENTER kernel: ata_sg_setup_one: mapped buffer of 36 bytes for read kernel: ata_fill_sg: PRD[0] = (0x469E54, 0x24) kernel: ata_dev_select: ENTER, ata2: device 0, wait 1 kernel: ata_tf_load_pio: feat 0x1 nsect 0x0 lba 0x0 0x0 0x0 kernel: ata_tf_load_pio: device 0xA0 kernel: ata_exec_command_pio: ata2: cmd 0xA0 kernel: ata_scsi_translate: EXIT kernel: atapi_packet_task: busy wait kernel: atapi_packet_task: send cdb kernel: ata_scsi_error: ENTER kernel: ata_eng_timeout: ENTER kernel: ata2: unknown timeout, cmd 0xa0 stat 0xd0 kernel: ata_sg_clean: unmapping 1 sg elements kernel: ata_eng_timeout: EXIT kernel: ata_scsi_error: EXIT kernel: ata_scsi_dump_cdb: CDB (2:0,1,0) 12 00 00 00 24 00 00 00 46 We now get only one ata_exec_command_pio, good. Next I will try o x1F1 Features = x05 Dma In and x01 Dma Out because: > Subject: Re: SATA ATAPI UDMA - first in Linux > ... > Silicon Image has introduced an addition to ATA called the "DMADIR" bit. > Their bridges require this bit in order to use DMA. > > If your OS driver does not know about DMADIR (read: most OS drivers), > then it will fall back _unconditionally_ to PIO. http://www.google.com/search?q=silicon+image+dmadir yields the proposal: Packet DMA Issue ... http://www.t13.org/docs2003/e03132r3.pdf Seemingly impractical on various grounds, such as requiring different device firmware for an old or new host, but including the detail: --- Features ... [x04] DMADIR ... Support for this bit is determined by reading bit 15 of word 62 in the device configuration data. If bit 15 of word 62 is set to 1, the device requires the use of the DMADIR bit for Packet DMA commands. If the device requires the DMADIR bit to be set for Packet DMA operations and the current operations is DMA (i.e. bit 0, the DMA bit, is set), this bit indicates the direction of data transfer (0 = transfer to the device; 1 = transfer to the host). --- Pat LaVarre