Re: ahci 0000:00:1f.2: DMA-API: device driver maps memory from stack

[Robert Hancock <hancockrwd@gmail.com>]

FUJITA Tomonori wrote:
>> One could print out where buf points in 
>> this function and see if that is the case. Sense buffer gets allocated 
>> by the SCSI layer..
> 
> I guess that SCSI-ml doesn't use a buffer on the stack for
> scmd->sense_buffer... atapi_eh_clear_ua() looks guilty?
> 
> 
> static int atapi_eh_clear_ua(struct ata_device *dev)
> {
> 	int i;
> 
> 	for (i = 0; i < ATA_EH_UA_TRIES; i++) {
> 		u8 sense_buffer[SCSI_SENSE_BUFFERSIZE];
> 		u8 sense_key = 0;
> 		unsigned int err_mask;
> 
> 		err_mask = atapi_eh_tur(dev, &sense_key);
> 		if (err_mask != 0 && err_mask != AC_ERR_DEV) {
> 			ata_dev_printk(dev, KERN_WARNING, "TEST_UNIT_READY "
> 				"failed (err_mask=0x%x)\n", err_mask);
> 			return -EIO;
> 		}
> 
> 		if (!err_mask || sense_key != UNIT_ATTENTION)
> 			return 0;
> 
> 		err_mask = atapi_eh_request_sense(dev, sense_buffer, sense_key);
> 

I don't think this is the path that Chuck's warning came through, that 
one looks like atapi_eh_request_sense. But yeah, this code does look 
bad, we DMA into that sense buffer so it can't be on the stack.