linux-kernel - Re: [PATCH] usb-storage: Optimize scan delay more precisely

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date: Wed, 27 Mar 2024 09:15:28 +0100
From: Greg KH <gregkh@...uxfoundation.org>
To: Norihiko Hama <norihiko.hama@...salpine.com>
Cc: "stern@...land.harvard.edu" <stern@...land.harvard.edu>,
	"linux-usb@...r.kernel.org" <linux-usb@...r.kernel.org>,
	"usb-storage@...ts.one-eyed-alien.net" <usb-storage@...ts.one-eyed-alien.net>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] usb-storage: Optimize scan delay more precisely

On Wed, Mar 27, 2024 at 07:57:52AM +0000, Norihiko Hama wrote:
> On Wed, Mar 27, 2024 at 07:39:55AM +0000, Norihiko Hama wrote:
> > > > Sorry, but module parameters are from the 1990's, we will not go back to that if at all possible as it's not easy to maintain and will not work properly for multiple devices.
> > > >
> > > > I can understand wanting something between 1 and 0 seconds, but adding yet-another-option isn't probably the best way, sorry.
> > > 1 second does not meet with performance requirement.
> >
> > Who is requiring such a performance requirement?  The USB specification?
> > Or something else?
> This is our customer requirement.

If it is impossible to do, why are they making you do it?  :)

> > > I have no good idea except module parameter so that we can maintain backward compatibility but be configurable out of module.
> > > Do you have any other better solution?
> > How long do you exactly need to wait?  Why not figure out how long the device takes and if it fails, slowly back off until the full time delay happens and then you can abort?
> It's IOP issue and difficult to figure out because it depends on device itself.

Agreed.

> I know we have multiple devices with delay_use=0, but then it's recovered and detected by reset after 30s timeout, that is too long than 1 sec.

So how do you know that making this smaller will help?  There are many
many odd and broken devices out there that take a long time to spin up
before they are able to be accessed.  That's what that option is there
for, if you "know" you don't need to wait, you don't have to wait.
Otherwise you HAVE to wait as you do not know how long things take.

sorry,

greg k-h