From: Russell Currey <ruscur@russell.cc> To: Christophe Leroy <christophe.leroy@c-s.fr>, linuxppc-dev@lists.ozlabs.org Cc: mikey@neuling.org, kernel-hardening@lists.openwall.com, npiggin@gmail.com Subject: Re: [PATCH v2 2/3] powerpc/lib: Refactor __patch_instruction() to use __put_user_asm() Date: Wed, 20 Feb 2019 22:57:59 +1100 [thread overview] Message-ID: <f4fc28e2f6807ef0669d1de8a72fcb5f5b5cfc43.camel@russell.cc> (raw) In-Reply-To: <5ff8b24e-a748-19d3-8651-b626dd676ea4@c-s.fr> On Fri, 2019-01-25 at 12:45 +0100, Christophe Leroy wrote: > Hi Russel, > > Le 17/12/2018 à 08:09, Christophe Leroy a écrit : > > Hi Russel, > > > > Le 10/12/2018 à 08:00, Russell Currey a écrit : > > > __patch_instruction() is called in early boot, and uses > > > __put_user_size(), which includes the locks and unlocks for KUAP, > > > which could either be called too early, or in the Radix case, > > > forced to > > > use "early_" versions of functions just to safely handle this one > > > case. > > > > Looking at x86, I see that __put_user_size() doesn't includes the > > locks. > > The lock/unlock is do by callers. I'll do the same. > > > > > > > __put_user_asm() does not do this, and thus is safe to use both > > > in early > > > boot, and later on since in this case it should only ever be > > > touching > > > kernel memory. > > > > > > __patch_instruction() was previously refactored to use > > > __put_user_size() > > > in order to be able to return -EFAULT, which would allow the > > > kernel to > > > patch instructions in userspace, which should never happen. This > > > has > > > the functional change of causing faults on userspace addresses if > > > KUAP > > > is turned on, which should never happen in practice. > > > > > > A future enhancement could be to double check the patch address > > > is > > > definitely allowed to be tampered with by the kernel. > > > > This makes me realise that we are calling lock_user_access() with > > kernel > > addresses. That most likely breaks protection on kernel addresses > > for > > book3s/32. I'll have to work around it. > > > > Another thing I realised also is that get_user() at least is called > > in > > some exceptions/trap handlers. Which means it can be called nested > > with > > an ongoing user access. It means that get_paca()- > > >user_access_allowed > > might be modified during those exceptions/traps. > > Any comment about that ? Isn't it a problem ? Yes, I think so. I wonder why I haven't hit this issue, though. Which handlers is this an issue with? Maybe we could do something like... unlock_user_access() checks if user access is already unlocked (== 1), if so sets user_access_allowed to 2 lock_user_access() sees that user_access_allowed is 2, and knows it's nested and sets user_access_allowed back to 1 instead of its usual 0. ...that's pretty gross, though. It also means that every implementation has to figure out how to cope with that. I've done a lot of testing where a) user access hasn't been left unlocked and b) faults haven't happened where they shouldn't, so I wonder how I could try and hit such a case. Could also have a get_user() without locking that's only allowed to be used by exception handlers... I dunno. Open to better ideas. - Russell > > Christophe > > > Christophe > > > > > Signed-off-by: Russell Currey <ruscur@russell.cc> > > > --- > > > arch/powerpc/lib/code-patching.c | 4 ++-- > > > 1 file changed, 2 insertions(+), 2 deletions(-) > > > > > > diff --git a/arch/powerpc/lib/code-patching.c > > > b/arch/powerpc/lib/code-patching.c > > > index 89502cbccb1b..15e8c6339960 100644 > > > --- a/arch/powerpc/lib/code-patching.c > > > +++ b/arch/powerpc/lib/code-patching.c > > > @@ -26,9 +26,9 @@ > > > static int __patch_instruction(unsigned int *exec_addr, > > > unsigned int > > > instr, > > > unsigned int *patch_addr) > > > { > > > - int err; > > > + int err = 0; > > > - __put_user_size(instr, patch_addr, 4, err); > > > + __put_user_asm(instr, patch_addr, err, "stw"); > > > if (err) > > > return err; > > >
WARNING: multiple messages have this Message-ID (diff)
From: Russell Currey <ruscur@russell.cc> To: Christophe Leroy <christophe.leroy@c-s.fr>, linuxppc-dev@lists.ozlabs.org Cc: mikey@neuling.org, mpe@ellerman.id.au, benh@kernel.crashing.org, npiggin@gmail.com, kernel-hardening@lists.openwall.com Subject: Re: [PATCH v2 2/3] powerpc/lib: Refactor __patch_instruction() to use __put_user_asm() Date: Wed, 20 Feb 2019 22:57:59 +1100 [thread overview] Message-ID: <f4fc28e2f6807ef0669d1de8a72fcb5f5b5cfc43.camel@russell.cc> (raw) In-Reply-To: <5ff8b24e-a748-19d3-8651-b626dd676ea4@c-s.fr> On Fri, 2019-01-25 at 12:45 +0100, Christophe Leroy wrote: > Hi Russel, > > Le 17/12/2018 à 08:09, Christophe Leroy a écrit : > > Hi Russel, > > > > Le 10/12/2018 à 08:00, Russell Currey a écrit : > > > __patch_instruction() is called in early boot, and uses > > > __put_user_size(), which includes the locks and unlocks for KUAP, > > > which could either be called too early, or in the Radix case, > > > forced to > > > use "early_" versions of functions just to safely handle this one > > > case. > > > > Looking at x86, I see that __put_user_size() doesn't includes the > > locks. > > The lock/unlock is do by callers. I'll do the same. > > > > > > > __put_user_asm() does not do this, and thus is safe to use both > > > in early > > > boot, and later on since in this case it should only ever be > > > touching > > > kernel memory. > > > > > > __patch_instruction() was previously refactored to use > > > __put_user_size() > > > in order to be able to return -EFAULT, which would allow the > > > kernel to > > > patch instructions in userspace, which should never happen. This > > > has > > > the functional change of causing faults on userspace addresses if > > > KUAP > > > is turned on, which should never happen in practice. > > > > > > A future enhancement could be to double check the patch address > > > is > > > definitely allowed to be tampered with by the kernel. > > > > This makes me realise that we are calling lock_user_access() with > > kernel > > addresses. That most likely breaks protection on kernel addresses > > for > > book3s/32. I'll have to work around it. > > > > Another thing I realised also is that get_user() at least is called > > in > > some exceptions/trap handlers. Which means it can be called nested > > with > > an ongoing user access. It means that get_paca()- > > >user_access_allowed > > might be modified during those exceptions/traps. > > Any comment about that ? Isn't it a problem ? Yes, I think so. I wonder why I haven't hit this issue, though. Which handlers is this an issue with? Maybe we could do something like... unlock_user_access() checks if user access is already unlocked (== 1), if so sets user_access_allowed to 2 lock_user_access() sees that user_access_allowed is 2, and knows it's nested and sets user_access_allowed back to 1 instead of its usual 0. ...that's pretty gross, though. It also means that every implementation has to figure out how to cope with that. I've done a lot of testing where a) user access hasn't been left unlocked and b) faults haven't happened where they shouldn't, so I wonder how I could try and hit such a case. Could also have a get_user() without locking that's only allowed to be used by exception handlers... I dunno. Open to better ideas. - Russell > > Christophe > > > Christophe > > > > > Signed-off-by: Russell Currey <ruscur@russell.cc> > > > --- > > > arch/powerpc/lib/code-patching.c | 4 ++-- > > > 1 file changed, 2 insertions(+), 2 deletions(-) > > > > > > diff --git a/arch/powerpc/lib/code-patching.c > > > b/arch/powerpc/lib/code-patching.c > > > index 89502cbccb1b..15e8c6339960 100644 > > > --- a/arch/powerpc/lib/code-patching.c > > > +++ b/arch/powerpc/lib/code-patching.c > > > @@ -26,9 +26,9 @@ > > > static int __patch_instruction(unsigned int *exec_addr, > > > unsigned int > > > instr, > > > unsigned int *patch_addr) > > > { > > > - int err; > > > + int err = 0; > > > - __put_user_size(instr, patch_addr, 4, err); > > > + __put_user_asm(instr, patch_addr, err, "stw"); > > > if (err) > > > return err; > > >
next prev parent reply other threads:[~2019-02-20 11:59 UTC|newest] Thread overview: 14+ messages / expand[flat|nested] mbox.gz Atom feed top 2018-12-10 7:00 [PATCH v2 0/3] [PATCH v2 0/3] Kernel Userspace Protection for Radix MMU Russell Currey 2018-12-10 7:00 ` Russell Currey 2018-12-10 7:00 ` [PATCH v2 1/3] powerpc/mm/radix: Use KUEP API " Russell Currey 2018-12-10 7:00 ` Russell Currey 2018-12-10 7:00 ` [PATCH v2 2/3] powerpc/lib: Refactor __patch_instruction() to use __put_user_asm() Russell Currey 2018-12-10 7:00 ` Russell Currey 2018-12-17 7:09 ` Christophe Leroy 2018-12-17 7:09 ` Christophe Leroy 2019-01-25 11:45 ` Christophe Leroy 2019-01-25 11:45 ` Christophe Leroy 2019-02-20 11:57 ` Russell Currey [this message] 2019-02-20 11:57 ` Russell Currey 2018-12-10 7:00 ` [PATCH v2 3/3] powerpc/64s: Implement KUAP for Radix MMU Russell Currey 2018-12-10 7:00 ` Russell Currey
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=f4fc28e2f6807ef0669d1de8a72fcb5f5b5cfc43.camel@russell.cc \ --to=ruscur@russell.cc \ --cc=christophe.leroy@c-s.fr \ --cc=kernel-hardening@lists.openwall.com \ --cc=linuxppc-dev@lists.ozlabs.org \ --cc=mikey@neuling.org \ --cc=npiggin@gmail.com \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.