Drivers: hv: vmbus: Split memcpy of flex-array
authorKees Cook <keescook@chromium.org>
Tue, 27 Sep 2022 21:17:36 +0000 (14:17 -0700)
committerWei Liu <wei.liu@kernel.org>
Wed, 28 Sep 2022 13:35:26 +0000 (13:35 +0000)
To work around a misbehavior of the compiler's ability to see into
composite flexible array structs (as detailed in the coming memcpy()
hardening series[1]), split the memcpy() of the header and the payload
so no false positive run-time overflow warning will be generated. This
results in the already inlined memcpy getting unrolled a little more,
which very slightly increases text size:

$ size drivers/hv/vmbus_drv.o.before drivers/hv/vmbus_drv.o
   text    data     bss     dec     hex filename
  22968    5239     232   28439    6f17 drivers/hv/vmbus_drv.o.before
  23032    5239     232   28503    6f57 drivers/hv/vmbus_drv.o

Avoids the run-time false-positive warning:

  memcpy: detected field-spanning write (size 212) of single field "&ctx->msg" at drivers/hv/vmbus_drv.c:1133 (size 16)

[1] https://lore.kernel.org/linux-hardening/20220901065914.1417829-2-keescook@chromium.org/

Cc: "K. Y. Srinivasan" <kys@microsoft.com>
Cc: Haiyang Zhang <haiyangz@microsoft.com>
Cc: Stephen Hemminger <sthemmin@microsoft.com>
Cc: Wei Liu <wei.liu@kernel.org>
Cc: Dexuan Cui <decui@microsoft.com>
Cc: linux-hyperv@vger.kernel.org
Reported-by: Nathan Chancellor <nathan@kernel.org>
Reported-by: "Gustavo A. R. Silva" <gustavoars@kernel.org>
Tested-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Reviewed-by: Michael Kelley <mikelley@microsoft.com>
Link: https://lore.kernel.org/r/20220927211736.3241175-1-keescook@chromium.org
Signed-off-by: Wei Liu <wei.liu@kernel.org>
drivers/hv/vmbus_drv.c

index cfc231b..bbbc929 100644 (file)
@@ -1130,7 +1130,8 @@ void vmbus_on_msg_dpc(unsigned long data)
                        return;
 
                INIT_WORK(&ctx->work, vmbus_onmessage_work);
-               memcpy(&ctx->msg, &msg_copy, sizeof(msg->header) + payload_size);
+               ctx->msg.header = msg_copy.header;
+               memcpy(&ctx->msg.payload, msg_copy.u.payload, payload_size);
 
                /*
                 * The host can generate a rescind message while we