Re: [PATCH 5 of 8] sd: Detect non-rotational devices

[Jeff Garzik <jeff@garzik.org>]

Jens Axboe wrote:
> On Thu, Apr 23 2009, Jeff Garzik wrote:
>> Jens Axboe wrote:
>>>> +static void sd_read_block_characteristics(struct scsi_disk *sdkp)
>>>> +{
>>>> +	char *buffer;
>>>> +	u16 rot;
>>>> +
>>>> +	/* Block Device Characteristics VPD */
>>>> +	buffer = scsi_get_vpd_page(sdkp->device, 0xb1);
>>>> +
>>>> +	if (buffer == NULL)
>>>> +		return;
>>>> +
>>>> +	rot = get_unaligned_be16(&buffer[4]);
>>>> +
>>>> +	if (rot == 1)
>>>> +		queue_flag_set_unlocked(QUEUE_FLAG_NONROT, sdkp->disk->queue);
>>>> +
>>>> +	kfree(buffer);
>>>> +}
>>>> +
>>>> +/**
>>>>   *	sd_revalidate_disk - called the first time a new disk is seen,
>>>>   *	performs disk spin up, read_capacity, etc.
>>>>   *	@disk: struct gendisk we care about
>>>> @@ -1836,6 +1860,7 @@ static int sd_revalidate_disk(struct gen
>>>>  	 */
>>>>  	if (sdkp->media_present) {
>>>>  		sd_read_capacity(sdkp, buffer);
>>>> +		sd_read_block_characteristics(sdkp);
>>>>  		sd_read_write_protect_flag(sdkp, buffer);
>>>>  		sd_read_cache_type(sdkp, buffer);
>>>>  		sd_read_app_tag_own(sdkp, buffer);
>>>> @@ -1976,6 +2001,8 @@ static void sd_probe_async(void *data, a
>>>>  	add_disk(gd);
>>>>  	sd_dif_config_host(sdkp);
>>>>  +	sd_revalidate_disk(gd);
>>>> +
>>>>  	sd_printk(KERN_NOTICE, sdkp, "Attached SCSI %sdisk\n",
>>>>  		  sdp->removable ? "removable " : "");
>>> Make sure this works for libata as well, and then kill the rotational
>>> check in there instead.
>> Yep.  libata-scsi.c would need to simulate that VPD page.
> 
> Exactly
> 
>> Also (to mkp or whoever does the work) -- note Linus's comment, and my  
>> provisional patch[1], about libata potentially wanting to detect NONROT  
>> by looking for "*SSD" from IDENTIFY DEVICE'S model string.
> 
> I think that's an entirely orthogonal issue.

Yes.

If somebody is going to be essentially rewriting the libata SSD 
detection logic, that area will get stirred _anyway_.

So I hope it is not out of bounds to point out additional areas where a 
developer could serve a legitimate need of Linux users, particularly 
when doing so is easy and a provisional patch already exists.

	Jeff