hotplugging SATA harddisks in Linux

7 Apr

That’s just a few records on my personal experience regarding the hot-swapping/hot-plugging capabilities of SATA

First, make sure the controller is set to AHCI in the BIOS so it will issue the required hotplug-messages on status changes (which seems quite likely to be suppressed in legacy-mode, since PATA doesn’t know anything about hotplug)

Next, check if nothing on the harddisk is mounted, preferrably do the whole testing from a live-cd like grml or similar. Then try to talk to the suspected device to ensure it is working correctly, and you’re addressing the right one. I like to check the S.M.A.R.T values for this:

smartctl -A /dev/sdb

In theory, you could just disconnect the drive’s SATA- and power-cables now, I’ve tested this several times without any problems. Just after detaching the cables, the kernel writes some statusmessages to syslog. Should be something like this here:

kernel: ata1: exception Emask 0×10 SAct 0×0 SErr 0×4010000 action 0xe frozen
kernel: ata1: irq_stat 0×00400040, connection status changed
kernel: ata1: SError: { PHYRdyChg DevExch }
kernel: ata1: hard resetting link
kernel: ata1: SATA link down (SStatus 0 SControl 300)
kernel: ata1: EH complete
kernel: ata1.00: detaching (SCSI 0:0:0:0)

If you want to be nicer to your friendly kernel (or safer, depending on how you look at this…), check for the supported SCSI-commands of your drive via sdparm:

sdparm -e -C x /dev/sdb

In my case, this yields the following output (coming from a 500GB Seagate drive):

Available commands: capacity eject load ready sense start stop sync unlock

Now you could try to detach the drive using those commands:

sdparm -C stop /dev/sdb
sdparm -C eject /dev/sdb

Another possibility is the tool “scsiadd”, first print the status using “-p”, then eject the appropriate device using “-r”:

scsiadd -p
scsiadd -r 1

4 Replies to “hotplugging SATA harddisks in Linux

  1. Hello, I am looking for information on proper SATA hot-swap / hot-plug on Ubuntu for about a month and there doesn’t seem to be any consensus. Your blog post caught my attention because I see the exact same kernel messages on my syslog when I pull one of my sata drives out.

    My motherboard bios says nothing about AHCI, nor does its documentation, so I am guessing it is not AHCI compliant.

    I wanted to ask you: Do you issue any command prior to inserting a new drive? I mean, suppose you want to add a new drive to your array, not replace a drive. If so, does your system recognize it? Mine does, but in wrong PIO/udma modes.

  2. Hi!
    If you can’t find anything like AHCI in your BIOS settings, look for things like “legacy” mode or similar in the controller config.

    To answer your question: no, I don’t issue any command prior to inserting drives – but I’ve never checked the transfer modes on the new drives. The system just recognized them and I was happy 🙂 (to be more verbose, I didn’t do this on any production server but just on my stupid desktop machine at work that does not have enough SATA-ports – so I used the hotplugging feature just to avoid the unnecessary reboot when changing drives).

    Ah, here’s the link my posting is based on:

Leave a Reply

Your email address will not be published. Required fields are marked *