Re: [PATCH v3] KVM: arm64: Add memory length checks and remove inline in do_ffa_mem_xfer

[Marc Zyngier]

On Tue, 10 Sep 2024 19:18:41 +0100,
Oliver Upton <oliver.upton@xxxxxxxxx> wrote:
> 
> On Wed, Sep 11, 2024 at 12:32:29AM +0800, Wei-Lin Chang wrote:
> > Hi everyone,
> > 
> > On Mon, Sep 09, 2024 at 06:01:54PM GMT, Snehal Koukuntla wrote:
> > > When we share memory through FF-A and the description of the buffers
> > > exceeds the size of the mapped buffer, the fragmentation API is used.
> > > The fragmentation API allows specifying chunks of descriptors in subsequent
> > > FF-A fragment calls and no upper limit has been established for this.
> > > The entire memory region transferred is identified by a handle which can be
> > > used to reclaim the transferred memory.
> > > To be able to reclaim the memory, the description of the buffers has to fit
> > > in the ffa_desc_buf.
> > > Add a bounds check on the FF-A sharing path to prevent the memory reclaim
> > > from failing.
> > > 
> > > Also do_ffa_mem_xfer() does not need __always_inline
> > > 
> > > Fixes: 634d90cf0ac65 ("KVM: arm64: Handle FFA_MEM_LEND calls from the host")
> > > Cc: stable@xxxxxxxxxxxxxxx
> > > Reviewed-by: Sebastian Ene <sebastianene@xxxxxxxxxx>
> > > Signed-off-by: Snehal Koukuntla <snehalreddy@xxxxxxxxxx>
> > > ---
> > >  arch/arm64/kvm/hyp/nvhe/ffa.c | 7 ++++++-
> > >  1 file changed, 6 insertions(+), 1 deletion(-)
> > > 
> > > diff --git a/arch/arm64/kvm/hyp/nvhe/ffa.c b/arch/arm64/kvm/hyp/nvhe/ffa.c
> > > index e715c157c2c4..637425f63fd1 100644
> > > --- a/arch/arm64/kvm/hyp/nvhe/ffa.c
> > > +++ b/arch/arm64/kvm/hyp/nvhe/ffa.c
> > > @@ -426,7 +426,7 @@ static void do_ffa_mem_frag_tx(struct arm_smccc_res *res,
> > >  	return;
> > >  }
> > >  
> > > -static __always_inline void do_ffa_mem_xfer(const u64 func_id,
> > > +static void do_ffa_mem_xfer(const u64 func_id,
> > 
> > I am seeing a compilation error because of this.
> 
> Thanks for reporting this. Looks like the __always_inline was slightly
> more load bearing...
> 
> Marc, can you put something like this on top?
> 
> 
> From c2712eaa94989ae6457baad3ec459cf363ec5119 Mon Sep 17 00:00:00 2001
> From: Oliver Upton <oliver.upton@xxxxxxxxx>
> Date: Tue, 10 Sep 2024 16:45:30 +0000
> Subject: [PATCH] KVM: arm64: Drop BUILD_BUG_ON() from do_ffa_mem_xfer()
> 
> __always_inline was recently discarded from do_ffa_mem_xfer() since it
> appeared to be unnecessary. Of course, this was ~immediately proven
> wrong, as the compile-time check against @func_id depends on inlining
> for the value to be known.
> 
> Just downgrade to a WARN_ON() instead of putting the old mess back in
> place. Fix the wrapping/indentation of the function parameters while at
> it.
>
> Fixes: 39dacbeeee70 ("KVM: arm64: Add memory length checks and remove inline in do_ffa_mem_xfer")
> Reported-by: Wei-Lin Chang <r09922117@xxxxxxxxxxxxxxx>
> Signed-off-by: Oliver Upton <oliver.upton@xxxxxxxxx>
> ---
>  arch/arm64/kvm/hyp/nvhe/ffa.c | 11 ++++++-----
>  1 file changed, 6 insertions(+), 5 deletions(-)
> 
> diff --git a/arch/arm64/kvm/hyp/nvhe/ffa.c b/arch/arm64/kvm/hyp/nvhe/ffa.c
> index 637425f63fd1..316d269341f3 100644
> --- a/arch/arm64/kvm/hyp/nvhe/ffa.c
> +++ b/arch/arm64/kvm/hyp/nvhe/ffa.c
> @@ -426,9 +426,8 @@ static void do_ffa_mem_frag_tx(struct arm_smccc_res *res,
>  	return;
>  }
>  
> -static void do_ffa_mem_xfer(const u64 func_id,
> -					    struct arm_smccc_res *res,
> -					    struct kvm_cpu_context *ctxt)
> +static void do_ffa_mem_xfer(const u64 func_id, struct arm_smccc_res *res,
> +			    struct kvm_cpu_context *ctxt)
>  {
>  	DECLARE_REG(u32, len, ctxt, 1);
>  	DECLARE_REG(u32, fraglen, ctxt, 2);
> @@ -440,8 +439,10 @@ static void do_ffa_mem_xfer(const u64 func_id,
>  	u32 offset, nr_ranges;
>  	int ret = 0;
>  
> -	BUILD_BUG_ON(func_id != FFA_FN64_MEM_SHARE &&
> -		     func_id != FFA_FN64_MEM_LEND);
> +	if (WARN_ON(func_id != FFA_FN64_MEM_SHARE && func_id != FFA_FN64_MEM_LEND)) {
> +		ret = SMCCC_RET_NOT_SUPPORTED;
> +		goto out;
> +	}


I'm not overly on the WARN_ON(), as it has pretty fatal effects on
pKVM (it simply panics). What do you think of this instead, which
compiles with my prehistoric version of clang (14.0.6):

diff --git a/arch/arm64/kvm/hyp/nvhe/ffa.c b/arch/arm64/kvm/hyp/nvhe/ffa.c
index 637425f63fd1b..e433dfab882aa 100644
--- a/arch/arm64/kvm/hyp/nvhe/ffa.c
+++ b/arch/arm64/kvm/hyp/nvhe/ffa.c
@@ -426,9 +426,9 @@ static void do_ffa_mem_frag_tx(struct arm_smccc_res *res,
 	return;
 }
 
-static void do_ffa_mem_xfer(const u64 func_id,
-					    struct arm_smccc_res *res,
-					    struct kvm_cpu_context *ctxt)
+static void __do_ffa_mem_xfer(const u64 func_id,
+			      struct arm_smccc_res *res,
+			      struct kvm_cpu_context *ctxt)
 {
 	DECLARE_REG(u32, len, ctxt, 1);
 	DECLARE_REG(u32, fraglen, ctxt, 2);
@@ -440,9 +440,6 @@ static void do_ffa_mem_xfer(const u64 func_id,
 	u32 offset, nr_ranges;
 	int ret = 0;
 
-	BUILD_BUG_ON(func_id != FFA_FN64_MEM_SHARE &&
-		     func_id != FFA_FN64_MEM_LEND);
-
 	if (addr_mbz || npages_mbz || fraglen > len ||
 	    fraglen > KVM_FFA_MBOX_NR_PAGES * PAGE_SIZE) {
 		ret = FFA_RET_INVALID_PARAMETERS;
@@ -517,6 +514,13 @@ static void do_ffa_mem_xfer(const u64 func_id,
 	goto out_unlock;
 }
 
+#define do_ffa_mem_xfer(fid, res, ctxt)				\
+	do {							\
+		BUILD_BUG_ON((fid) != FFA_FN64_MEM_SHARE &&	\
+			     (fid) != FFA_FN64_MEM_LEND);	\
+		__do_ffa_mem_xfer((fid), (res), (ctxt));	\
+	} while (0);
+
 static void do_ffa_mem_reclaim(struct arm_smccc_res *res,
 			       struct kvm_cpu_context *ctxt)
 {

It preserves the build-time assertion, which was the intention of the
original author.

I can easily squash that in the original commit, avoiding the headache
of backporting both patch to stable.

	M.

-- 
Without deviation from the norm, progress is not possible.