On Mon, Mar 5, 2018 at 11:44 AM, Joerg Roedel <j...@8bytes.org> wrote: > On Mon, Mar 05, 2018 at 09:51:29AM -0500, Brian Gerst wrote: >> For the IRET fault case you will still need to catch it in the >> exception code. See the 64-bit code (.Lerror_bad_iret) for example. >> For 32-bit, you could just expand that check to cover the whole exit >> prologue after the CR3 switch, including the data segment loads. > > I had a look at the 64 bit code and the exception-in-kernel case seems > to be handled differently than on 32 bit. The 64 bit entry code has > checks for certain kinds of errors like iret exceptions. > > On 32 bit this is implemented via the standard exception tables which > get an entry for every EIP that might fault (usually segment loading > operations, but also iret). > > So, unless I am missing something, all the exception entry code has to > do is to remember the stack and the cr3 with which it was entered (if > entered from kernel mode) and restore those before iret. And this is > what I implemented in v3 of this patch-set.
I also noticed that 32-bit will raise SIGILL for all IRET faults, while 64-bit will raise SIGBUS (#NP/#SS) or SIGSEGV (#GP). The 64-bit code is better since it doesn't lose the original fault type, whereas SIGILL is wrong for this case (illegal opcode). -- Brian Gerst
