> -----Original Message----- > From: Mimi Zohar [mailto:zohar@xxxxxxxxxxxxx] > Sent: Monday, March 30, 2020 8:37 PM > To: Roberto Sassu <roberto.sassu@xxxxxxxxxx>; > matthewgarrett@xxxxxxxxxx > Cc: linux-integrity@xxxxxxxxxxxxxxx; Silviu Vlasceanu > <Silviu.Vlasceanu@xxxxxxxxxx> > Subject: Re: Immutable metadata > > On Mon, 2020-03-30 at 15:56 +0000, Roberto Sassu wrote: > > > -----Original Message----- > > > From: Mimi Zohar [mailto:zohar@xxxxxxxxxxxxx] > > > Sent: Monday, March 30, 2020 5:47 PM > > > To: Roberto Sassu <roberto.sassu@xxxxxxxxxx>; > > > matthewgarrett@xxxxxxxxxx > > > Cc: linux-integrity@xxxxxxxxxxxxxxx; Silviu Vlasceanu > > > <Silviu.Vlasceanu@xxxxxxxxxx> > > > Subject: Re: Immutable metadata > > > > > > On Sun, 2020-03-29 at 22:10 -0400, Mimi Zohar wrote: > > > > Hi Roberto, > > > > > > > > On Sat, 2020-03-28 at 11:18 +0000, Roberto Sassu wrote: > > > > > Hi Matthew, Mimi > > > > > > > > > > I have a question about portable signatures. Is there any particular > > > reason > > > > > why a write to a file is not denied by IMA if metadata are immutable? > > > > > > > > As much as possible, IMA and EVM should be independent of each > other. > > > > EVM is responsible for the integrity of file metadata, so it needs to > > > > read other security xattrs, but IMA shouldn't be looking at the EVM > > > > xattr. > > > > > > > > Like any other security xattr, responsibility for maintaining the > > > > xattr is left up to the particular LSM. In this case, EVM would need > > > > to prevent the file from being opened rw. Should that be hard coded > > > > or based on an EVM policy? > > > > > > Thinking about this a bit more, evm_verifyxattr() is already returning > > > INTEGRITY_PASS_IMMUTABLE. I guess IMA could make decisions based > on > > > it. > > > > Yes, this was the idea. > > > > I would say also that files with portable signatures fulfill the > appraise_type=imasig > > requirement. I would set the IMA_DIGSIG bit in iint->atomic_flags. Is it ok? > > Ok, so locking doesn't seem to be an issue here. I'm not sure about > re-using the existing bit. EVM_XATTR_PORTABLE_DIGSIG is dependent on > the existence of a file hash. The existing bit prevents calculating > and writing the file hash as an xattr. Would this affect installing > new files? Since it is a portable signature, security.ima is known in advance. IMA ability to update security.ima is not necessary in this case. security.ima will be set by user space tools. Currently, there is still the issue of adding all xattrs to a file when EVM is enforcing (it refuses to copy all xattrs, as the integrity of the partial set cannot be verified). I will address this later. Roberto HUAWEI TECHNOLOGIES Duesseldorf GmbH, HRB 56063 Managing Director: Li Peng, Li Jian, Shi Yanli
