On Fri, Sep 24, 2021 at 04:06:10PM +0300, Reco wrote: > Hi. > > On Fri, Sep 24, 2021 at 01:59:58PM +0200, tomas@tuxteam.de wrote: > > On Fri, Sep 24, 2021 at 12:27:56PM +0300, Reco wrote: > > > > [...] > > > > > FUSE = slow + CPU wastage > > > > > > Using a filesystem the way it was intended is much cleaner solution. > > > > On the flip side, using an in-kernel file system is running code > > in kernel space which was conceived and written in happier times. > > I cannot see what's exactly wrong with ext4 these days. > Unless you have something against IBM/RH that is. About IBM, I'll shut up. Too much Smeagol for me ;-) But this is unrelated. RedHat is, in my eyes, a cool company, whithin the constraints they are subjected to. But this, too, is unrelated. > And by using FUSE one does not get a magical safeguard against kernel > panics and processes in D-state. > > > > Back then you could more or less safely assume that a file system > > image wasn't out to kill you. These days, though... > > Oh. Citation needed. Curious minds want to know. > How exactly one can produce a filesystem image that tries to get you? > Just in case, I'm asking out of mere curiosity, not with an intent on > using said image on somebody ;) I'll leave the word to Dave Chinner [1], who should know a thing or two more about file systems than we both do. Thing is nowadays I [2] can engineer an ext4 file system image in an USB stick exploiting some vulnerability in the file system code and presto, I'm in your kernel code whenever you mount the thing. FUSE provides (at least) some mitigation to that, by providing a much narrower interface to the kernel. Cheers [1] https://lwn.net/ml/linux-kernel/20180524004931.GB23861@dastard/ [2] A "generic" "I": in reality, it might take me a couple of years to learn what I have to to pull that off. - t
Attachment:
signature.asc
Description: Digital signature