KVM: SEV: Do cache maintenance on the source VM during intra-host migrationgrafted

Manually perform cache maintenance on the source VM during intra-host
migration to ensure no stale data is left in CPU caches after the VM is
destroyed.  Because the source VM is "converted" to a non-SEV VM, KVM's
memory reclaim flows won't trigger cache maintenance, e.g. when all guest
memory is reclaimed in response to detaching from the mmu_notifier.

Note, relying on the destination VM to do cache maintenance isn't an option
as KVM doesn't require identical guest memory configurations, i.e. the
source VM may have access to memory that the destination VM does not.
Enforcing equivalent memory configurations is infeasible, as it would
require a *deep* comparison of memslots, e.g. to verify that not only are
the memslot identical, but what the memslots point at is also identical.

Fixes: b56639318bb2 ("KVM: SEV: Add support for SEV intra host migration")
Cc: stable@vger.kernel.org
Reported-by: Stefan Teodorescu <fane@google.com>
Signed-off-by: Sean Christopherson <seanjc@google.com>
Message-ID: <20260923163721.1584779-3-seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

1 files changed, 290 insertions, 0 deletions

diff --git a/arch/alpha/lib/stxcpy.S b/arch/alpha/lib/stxcpy.S
new file mode 100644
index 000000000..58723b0a3
--- /dev/null
+++ b/arch/alpha/lib/stxcpy.S
@@ -0,0 +1,290 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * arch/alpha/lib/stxcpy.S
+ * Contributed by Richard Henderson (rth@tamu.edu)
+ *
+ * Copy a null-terminated string from SRC to DST.
+ *
+ * This is an internal routine used by strcpy, stpcpy, and strcat.
+ * As such, it uses special linkage conventions to make implementation
+ * of these public functions more efficient.
+ *
+ * On input:
+ *	t9 = return address
+ *	a0 = DST
+ *	a1 = SRC
+ *
+ * On output:
+ *	t12 = bitmask (with one bit set) indicating the last byte written
+ *	a0  = unaligned address of the last *word* written
+ *
+ * Furthermore, v0, a3-a5, t11, and t12 are untouched.
+ */
+
+#include <asm/regdef.h>
+
+	.set noat
+	.set noreorder
+
+	.text
+
+/* There is a problem with either gdb (as of 4.16) or gas (as of 2.7) that
+   doesn't like putting the entry point for a procedure somewhere in the
+   middle of the procedure descriptor.  Work around this by putting the
+   aligned copy in its own procedure descriptor */
+
+	.ent stxcpy_aligned
+	.align 3
+stxcpy_aligned:
+	.frame sp, 0, t9
+	.prologue 0
+
+	/* On entry to this basic block:
+	   t0 == the first destination word for masking back in
+	   t1 == the first source word.  */
+
+	/* Create the 1st output word and detect 0's in the 1st input word.  */
+	lda	t2, -1		# e1    : build a mask against false zero
+	mskqh	t2, a1, t2	# e0    :   detection in the src word
+	mskqh	t1, a1, t3	# e0    :
+	ornot	t1, t2, t2	# .. e1 :
+	mskql	t0, a1, t0	# e0    : assemble the first output word
+	cmpbge	zero, t2, t8	# .. e1 : bits set iff null found
+	or	t0, t3, t1	# e0    :
+	bne	t8, $a_eos	# .. e1 :
+
+	/* On entry to this basic block:
+	   t0 == the first destination word for masking back in
+	   t1 == a source word not containing a null.  */
+
+$a_loop:
+	stq_u	t1, 0(a0)	# e0    :
+	addq	a0, 8, a0	# .. e1 :
+	ldq_u	t1, 0(a1)	# e0    :
+	addq	a1, 8, a1	# .. e1 :
+	cmpbge	zero, t1, t8	# e0 (stall)
+	beq	t8, $a_loop	# .. e1 (zdb)
+
+	/* Take care of the final (partial) word store.
+	   On entry to this basic block we have:
+	   t1 == the source word containing the null
+	   t8 == the cmpbge mask that found it.  */
+$a_eos:
+	negq	t8, t6		# e0    : find low bit set
+	and	t8, t6, t12	# e1 (stall)
+
+	/* For the sake of the cache, don't read a destination word
+	   if we're not going to need it.  */
+	and	t12, 0x80, t6	# e0    :
+	bne	t6, 1f		# .. e1 (zdb)
+
+	/* We're doing a partial word store and so need to combine
+	   our source and original destination words.  */
+	ldq_u	t0, 0(a0)	# e0    :
+	subq	t12, 1, t6	# .. e1 :
+	zapnot	t1, t6, t1	# e0    : clear src bytes >= null
+	or	t12, t6, t8	# .. e1 :
+	zap	t0, t8, t0	# e0    : clear dst bytes <= null
+	or	t0, t1, t1	# e1    :
+
+1:	stq_u	t1, 0(a0)	# e0    :
+	ret	(t9)		# .. e1 :
+
+	.end stxcpy_aligned
+
+	.align 3
+	.ent __stxcpy
+	.globl __stxcpy
+__stxcpy:
+	.frame sp, 0, t9
+	.prologue 0
+
+	/* Are source and destination co-aligned?  */
+	xor	a0, a1, t0	# e0    :
+	unop			#       :
+	and	t0, 7, t0	# e0    :
+	bne	t0, $unaligned	# .. e1 :
+
+	/* We are co-aligned; take care of a partial first word.  */
+	ldq_u	t1, 0(a1)	# e0    : load first src word
+	and	a0, 7, t0	# .. e1 : take care not to load a word ...
+	addq	a1, 8, a1		# e0    :
+	beq	t0, stxcpy_aligned	# .. e1 : ... if we wont need it
+	ldq_u	t0, 0(a0)	# e0    :
+	br	stxcpy_aligned	# .. e1 :
+
+
+/* The source and destination are not co-aligned.  Align the destination
+   and cope.  We have to be very careful about not reading too much and
+   causing a SEGV.  */
+
+	.align 3
+$u_head:
+	/* We know just enough now to be able to assemble the first
+	   full source word.  We can still find a zero at the end of it
+	   that prevents us from outputting the whole thing.
+
+	   On entry to this basic block:
+	   t0 == the first dest word, for masking back in, if needed else 0
+	   t1 == the low bits of the first source word
+	   t6 == bytemask that is -1 in dest word bytes */
+
+	ldq_u	t2, 8(a1)	# e0    :
+	addq	a1, 8, a1	# .. e1 :
+
+	extql	t1, a1, t1	# e0    :
+	extqh	t2, a1, t4	# e0    :
+	mskql	t0, a0, t0	# e0    :
+	or	t1, t4, t1	# .. e1 :
+	mskqh	t1, a0, t1	# e0    :
+	or	t0, t1, t1	# e1    :
+
+	or	t1, t6, t6	# e0    :
+	cmpbge	zero, t6, t8	# .. e1 :
+	lda	t6, -1		# e0    : for masking just below
+	bne	t8, $u_final	# .. e1 :
+
+	mskql	t6, a1, t6		# e0    : mask out the bits we have
+	or	t6, t2, t2		# e1    :   already extracted before
+	cmpbge	zero, t2, t8		# e0    :   testing eos
+	bne	t8, $u_late_head_exit	# .. e1 (zdb)
+
+	/* Finally, we've got all the stupid leading edge cases taken care
+	   of and we can set up to enter the main loop.  */
+
+	stq_u	t1, 0(a0)	# e0    : store first output word
+	addq	a0, 8, a0	# .. e1 :
+	extql	t2, a1, t0	# e0    : position ho-bits of lo word
+	ldq_u	t2, 8(a1)	# .. e1 : read next high-order source word
+	addq	a1, 8, a1	# e0    :
+	cmpbge	zero, t2, t8	# .. e1 :
+	nop			# e0    :
+	bne	t8, $u_eos	# .. e1 :
+
+	/* Unaligned copy main loop.  In order to avoid reading too much,
+	   the loop is structured to detect zeros in aligned source words.
+	   This has, unfortunately, effectively pulled half of a loop
+	   iteration out into the head and half into the tail, but it does
+	   prevent nastiness from accumulating in the very thing we want
+	   to run as fast as possible.
+
+	   On entry to this basic block:
+	   t0 == the shifted high-order bits from the previous source word
+	   t2 == the unshifted current source word
+
+	   We further know that t2 does not contain a null terminator.  */
+
+	.align 3
+$u_loop:
+	extqh	t2, a1, t1	# e0    : extract high bits for current word
+	addq	a1, 8, a1	# .. e1 :
+	extql	t2, a1, t3	# e0    : extract low bits for next time
+	addq	a0, 8, a0	# .. e1 :
+	or	t0, t1, t1	# e0    : current dst word now complete
+	ldq_u	t2, 0(a1)	# .. e1 : load high word for next time
+	stq_u	t1, -8(a0)	# e0    : save the current word
+	mov	t3, t0		# .. e1 :
+	cmpbge	zero, t2, t8	# e0    : test new word for eos
+	beq	t8, $u_loop	# .. e1 :
+
+	/* We've found a zero somewhere in the source word we just read.
+	   If it resides in the lower half, we have one (probably partial)
+	   word to write out, and if it resides in the upper half, we
+	   have one full and one partial word left to write out.
+
+	   On entry to this basic block:
+	   t0 == the shifted high-order bits from the previous source word
+	   t2 == the unshifted current source word.  */
+$u_eos:
+	extqh	t2, a1, t1	# e0    :
+	or	t0, t1, t1	# e1    : first (partial) source word complete
+
+	cmpbge	zero, t1, t8	# e0    : is the null in this first bit?
+	bne	t8, $u_final	# .. e1 (zdb)
+
+$u_late_head_exit:
+	stq_u	t1, 0(a0)	# e0    : the null was in the high-order bits
+	addq	a0, 8, a0	# .. e1 :
+	extql	t2, a1, t1	# e0    :
+	cmpbge	zero, t1, t8	# .. e1 :
+
+	/* Take care of a final (probably partial) result word.
+	   On entry to this basic block:
+	   t1 == assembled source word
+	   t8 == cmpbge mask that found the null.  */
+$u_final:
+	negq	t8, t6		# e0    : isolate low bit set
+	and	t6, t8, t12	# e1    :
+
+	and	t12, 0x80, t6	# e0    : avoid dest word load if we can
+	bne	t6, 1f		# .. e1 (zdb)
+
+	ldq_u	t0, 0(a0)	# e0    :
+	subq	t12, 1, t6	# .. e1 :
+	or	t6, t12, t8	# e0    :
+	zapnot	t1, t6, t1	# .. e1 : kill source bytes >= null
+	zap	t0, t8, t0	# e0    : kill dest bytes <= null
+	or	t0, t1, t1	# e1    :
+
+1:	stq_u	t1, 0(a0)	# e0    :
+	ret	(t9)		# .. e1 :
+
+	/* Unaligned copy entry point.  */
+	.align 3
+$unaligned:
+
+	ldq_u	t1, 0(a1)	# e0    : load first source word
+
+	and	a0, 7, t4	# .. e1 : find dest misalignment
+	and	a1, 7, t5	# e0    : find src misalignment
+
+	/* Conditionally load the first destination word and a bytemask
+	   with 0xff indicating that the destination byte is sacrosanct.  */
+
+	mov	zero, t0	# .. e1 :
+	mov	zero, t6	# e0    :
+	beq	t4, 1f		# .. e1 :
+	ldq_u	t0, 0(a0)	# e0    :
+	lda	t6, -1		# .. e1 :
+	mskql	t6, a0, t6	# e0    :
+1:
+	subq	a1, t4, a1	# .. e1 : sub dest misalignment from src addr
+
+	/* If source misalignment is larger than dest misalignment, we need
+	   extra startup checks to avoid SEGV.  */
+
+	cmplt	t4, t5, t12	# e0    :
+	beq	t12, $u_head	# .. e1 (zdb)
+
+	lda	t2, -1		# e1    : mask out leading garbage in source
+	mskqh	t2, t5, t2	# e0    :
+	nop			# e0    :
+	ornot	t1, t2, t3	# .. e1 :
+	cmpbge	zero, t3, t8	# e0    : is there a zero?
+	beq	t8, $u_head	# .. e1 (zdb)
+
+	/* At this point we've found a zero in the first partial word of
+	   the source.  We need to isolate the valid source data and mask
+	   it into the original destination data.  (Incidentally, we know
+	   that we'll need at least one byte of that original dest word.) */
+
+	ldq_u	t0, 0(a0)	# e0    :
+
+	negq	t8, t6		# .. e1 : build bitmask of bytes <= zero
+	and	t6, t8, t12	# e0    :
+	and	a1, 7, t5	# .. e1 :
+	subq	t12, 1, t6	# e0    :
+	or	t6, t12, t8	# e1    :
+	srl	t12, t5, t12	# e0    : adjust final null return value
+
+	zapnot	t2, t8, t2	# .. e1 : prepare source word; mirror changes
+	and	t1, t2, t1	# e1    : to source validity mask
+	extql	t2, a1, t2	# .. e0 :
+	extql	t1, a1, t1	# e0    :
+
+	andnot	t0, t2, t0	# .. e1 : zero place for source to reside
+	or	t0, t1, t1	# e1    : and put it there
+	stq_u	t1, 0(a0)	# .. e0 :
+	ret	(t9)		# e1    :
+
+	.end __stxcpy