iommu/amd: use full 64-bit value in build_completion_wait()
authorJohn Sperbeck <jsperbeck@google.com>
Mon, 1 Aug 2022 19:22:29 +0000 (19:22 +0000)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Thu, 15 Sep 2022 09:30:07 +0000 (11:30 +0200)
[ Upstream commit 94a568ce32038d8ff9257004bb4632e60eb43a49 ]

We started using a 64 bit completion value.  Unfortunately, we only
stored the low 32-bits, so a very large completion value would never
be matched in iommu_completion_wait().

Fixes: c69d89aff393 ("iommu/amd: Use 4K page for completion wait write-back semaphore")
Signed-off-by: John Sperbeck <jsperbeck@google.com>
Link: https://lore.kernel.org/r/20220801192229.3358786-1-jsperbeck@google.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
drivers/iommu/amd/iommu.c

index e23e70a..7154fb5 100644 (file)
@@ -852,7 +852,8 @@ static void build_completion_wait(struct iommu_cmd *cmd,
        memset(cmd, 0, sizeof(*cmd));
        cmd->data[0] = lower_32_bits(paddr) | CMD_COMPL_WAIT_STORE_MASK;
        cmd->data[1] = upper_32_bits(paddr);
-       cmd->data[2] = data;
+       cmd->data[2] = lower_32_bits(data);
+       cmd->data[3] = upper_32_bits(data);
        CMD_SET_TYPE(cmd, CMD_COMPL_WAIT);
 }