Re: Is fdisk, disklabel and newfs enough to reset an SSD
Jan Stary wrote: If so, how does one reset a used SSD for optimal operation with a fresh install? Just treat it as any other disk - which it is. This is wrong, unfortunately. From the OS perspective, sure, sort of. But there is more to the story. There is overprovising, garbage collection, entire page write restrictions/design, TRIM, wear leveling, non-linear aspects, write amplification concerns, etc. Anyway, Ted and Christer nailed it with the ATA Secure Erase command. That's exactly what I was looking for. It's a hardware level reset that reconstitutes the over provisioned areas, unsets core storage (not a 1 or 0), and reinitializes wear leveling, etc. Thanks, Clint
Re: Is fdisk, disklabel and newfs enough to reset an SSD
On Tue, May 14, 2013 at 7:50 AM, Clint Pachl pa...@ecentryx.com wrote: Would dd'ing to the drive all 1s then all 0s be effective? Yes, and a complete waste of time. 'atactl drive secerase' will do the job for you. hdparm in linux has a similar command. But dd-ing twice is just idiotic. If you must use dd, one time is enough. -- chs,
Re: Is fdisk, disklabel and newfs enough to reset an SSD
On May 13 21:04:00, pa...@ecentryx.com wrote: I would like to reinstall a fresh system on an SSD that contains an existing installation. From my limited knowledge of SSDs, I wonder if the drive controller may retain data from the old filesystem, unaware that there is a new filesystem put in place. The drive controller knows nothing about filesystems. It just passes bytes to and from a device. The OpenBSD installer will just use the disk portion that you allocate to it. It will not waste time inspectioing what was on the disk before. Will there be pieces of the disk that still contain untouched data blocks of files that existed on some previous filesystem? Yes, they might. Is this a concern? No. If so, how does one reset a used SSD for optimal operation with a fresh install? Just treat it as any other disk - which it is. On May 13 22:50:08, pa...@ecentryx.com wrote: Scott McEachern wrote: 2) Do you mean there could still be data residing on unused parts of the SSD? Yes, it can happen. Yes, this is what I'm referring to. I was hoping there was some way to instruct the drive controller that the entire drive space is free? It is, if you say so in the installer. Making sure there are no datablocks left over is another thing. The first question would be why would you concern yoursewlf with this.
Re: Is fdisk, disklabel and newfs enough to reset an SSD
On Tue, May 14, 2013 at 15:01, Jan Stary wrote: If so, how does one reset a used SSD for optimal operation with a fresh install? Just treat it as any other disk - which it is. Almost, but not entirely. On May 13 22:50:08, pa...@ecentryx.com wrote: Scott McEachern wrote: 2) Do you mean there could still be data residing on unused parts of the SSD? Yes, it can happen. Yes, this is what I'm referring to. I was hoping there was some way to instruct the drive controller that the entire drive space is free? It is, if you say so in the installer. No. OpenBSD knows that space is free, the drive does not. Yes, the drive cares.
Is fdisk, disklabel and newfs enough to reset an SSD
I would like to reinstall a fresh system on an SSD that contains an existing installation. From my limited knowledge of SSDs, I wonder if the drive controller may retain data from the old filesystem, unaware that there is a new filesystem put in place. Is this a concern? If so, how does one reset a used SSD for optimal operation with a fresh install?
Re: Is fdisk, disklabel and newfs enough to reset an SSD
On 05/14/13 00:04, Clint Pachl wrote: I would like to reinstall a fresh system on an SSD that contains an existing installation. From my limited knowledge of SSDs, I wonder if the drive controller may retain data from the old filesystem, unaware that there is a new filesystem put in place. Is this a concern? If so, how does one reset a used SSD for optimal operation with a fresh install? I've done a fresh install of OpenBSD over top of OpenBSD (and other OSes) many times across many SSDs and I've never had a problem. But I'm not entirely sure what you mean... 1) Do you mean your new installation will see files left over from a previous install? No, it won't. 2) Do you mean there could still be data residing on unused parts of the SSD? Yes, it can happen. SSDs have their own way of wear-leveling. What the filesystem considers to be cylinder X, head Y and sector Z will probably not be the same *physical* cells on the SSD twice in a row. That's not a function of the OS, but the SSD itself. Do a little googling and you'll see what I mean: There's no guaranteed way to erase an SSD. I've read stories of people that have had SSDs crap out on them and instead of sending them back to the manufacturer for warranty repair/replacement, they just chuck them out and buy new ones. Why? Because there's no way to guarantee your private data has actually been erased. -- Scott McEachern https://www.blackstaff.ca Those who would give up essential liberty to purchase a little temporary safety deserve neither liberty nor safety. -- Benjamin Franklin
Re: Is fdisk, disklabel and newfs enough to reset an SSD
On Mon, May 13, 2013 at 21:04, Clint Pachl wrote: I would like to reinstall a fresh system on an SSD that contains an existing installation. From my limited knowledge of SSDs, I wonder if the drive controller may retain data from the old filesystem, unaware that there is a new filesystem put in place. Is this a concern? If so, how does one reset a used SSD for optimal operation with a fresh install? 'atactl drive secerase' will tell the drive the old contents are no longer needed. None of the other tools currently do that. I don't think it's a concern.
Re: Is fdisk, disklabel and newfs enough to reset an SSD
Scott McEachern wrote: 2) Do you mean there could still be data residing on unused parts of the SSD? Yes, it can happen. Yes, this is what I'm referring to. I was hoping there was some way to instruct the drive controller that the entire drive space is free? SSDs have their own way of wear-leveling. What the filesystem considers to be cylinder X, head Y and sector Z will probably not be the same *physical* cells on the SSD twice in a row. That's not a function of the OS, but the SSD itself. I understand. That's why I'm concerned about #2 above. Would dd'ing to the drive all 1s then all 0s be effective? I see Intel has an SSD Toolbox that does secure erase. It requires windows so I am unable to check it out. www.intel.com/go/ssdtoolbox I wonder what this utility does to achieve secure erase? Can we replicate its functionality with standard Unix tools?
