2023-01-06 05:51:16

by Kees Cook

[permalink] [raw]
Subject: [PATCH] scsi: megaraid_sas: Add flexible array member for SGLs

struct MPI2_RAID_SCSI_IO_REQUEST ends with a single SGL, but expects to
copy multiple. Add a flexible array member so the compiler can reason
about the size of the memcpy(). This will avoid the run-time false
positive warning:

memcpy: detected field-spanning write (size 128) of single field "&r1_cmd->io_request->SGL" at drivers/scsi/megaraid/megaraid_sas_fusion.c:3326 (size 16)

This change results in no binary output differences.

Reported-by: Holger Kiehl <[email protected]>
Link: https://lore.kernel.org/all/[email protected]/
Cc: Kashyap Desai <[email protected]>
Cc: Sumit Saxena <[email protected]>
Cc: Shivasharan S <[email protected]>
Cc: "James E.J. Bottomley" <[email protected]>
Cc: "Martin K. Petersen" <[email protected]>
Cc: [email protected]
Cc: [email protected]
Signed-off-by: Kees Cook <[email protected]>
---
Holger, are you able to test this change? I expect it should do the
trick, but I don't have the hardware.
---
drivers/scsi/megaraid/megaraid_sas_fusion.c | 2 +-
drivers/scsi/megaraid/megaraid_sas_fusion.h | 5 ++++-
2 files changed, 5 insertions(+), 2 deletions(-)

diff --git a/drivers/scsi/megaraid/megaraid_sas_fusion.c b/drivers/scsi/megaraid/megaraid_sas_fusion.c
index fe70f8f11435..6597e118c805 100644
--- a/drivers/scsi/megaraid/megaraid_sas_fusion.c
+++ b/drivers/scsi/megaraid/megaraid_sas_fusion.c
@@ -3323,7 +3323,7 @@ static void megasas_prepare_secondRaid1_IO(struct megasas_instance *instance,
/* copy the io request frame as well as 8 SGEs data for r1 command*/
memcpy(r1_cmd->io_request, cmd->io_request,
(sizeof(struct MPI2_RAID_SCSI_IO_REQUEST)));
- memcpy(&r1_cmd->io_request->SGL, &cmd->io_request->SGL,
+ memcpy(r1_cmd->io_request->SGLs, cmd->io_request->SGLs,
(fusion->max_sge_in_main_msg * sizeof(union MPI2_SGE_IO_UNION)));
/*sense buffer is different for r1 command*/
r1_cmd->io_request->SenseBufferLowAddress =
diff --git a/drivers/scsi/megaraid/megaraid_sas_fusion.h b/drivers/scsi/megaraid/megaraid_sas_fusion.h
index 49e9a9048ee7..b677d80e5874 100644
--- a/drivers/scsi/megaraid/megaraid_sas_fusion.h
+++ b/drivers/scsi/megaraid/megaraid_sas_fusion.h
@@ -526,7 +526,10 @@ struct MPI2_RAID_SCSI_IO_REQUEST {
__le32 Control; /* 0x3C */
union MPI2_SCSI_IO_CDB_UNION CDB; /* 0x40 */
union RAID_CONTEXT_UNION RaidContext; /* 0x60 */
- union MPI2_SGE_IO_UNION SGL; /* 0x80 */
+ union {
+ union MPI2_SGE_IO_UNION SGL; /* 0x80 */
+ DECLARE_FLEX_ARRAY(union MPI2_SGE_IO_UNION, SGLs);
+ };
};

/*
--
2.34.1


2023-01-06 11:56:32

by Holger Kiehl

[permalink] [raw]
Subject: Re: [PATCH] scsi: megaraid_sas: Add flexible array member for SGLs

On Thu, 5 Jan 2023, Kees Cook wrote:

> struct MPI2_RAID_SCSI_IO_REQUEST ends with a single SGL, but expects to
> copy multiple. Add a flexible array member so the compiler can reason
> about the size of the memcpy(). This will avoid the run-time false
> positive warning:
>
> memcpy: detected field-spanning write (size 128) of single field "&r1_cmd->io_request->SGL" at drivers/scsi/megaraid/megaraid_sas_fusion.c:3326 (size 16)
>
> This change results in no binary output differences.
>
> Reported-by: Holger Kiehl <[email protected]>
> Link: https://lore.kernel.org/all/[email protected]/
> Cc: Kashyap Desai <[email protected]>
> Cc: Sumit Saxena <[email protected]>
> Cc: Shivasharan S <[email protected]>
> Cc: "James E.J. Bottomley" <[email protected]>
> Cc: "Martin K. Petersen" <[email protected]>
> Cc: [email protected]
> Cc: [email protected]
> Signed-off-by: Kees Cook <[email protected]>
> ---
> Holger, are you able to test this change? I expect it should do the
> trick, but I don't have the hardware.
>
Yes, that does work. I no longer see 'memcpy: detected field-spanning
write (size 128)'. Tested this on 6.1.4-rc1. Did not see any other
regression.

Regards,
Holger

2023-01-19 00:15:07

by Martin K. Petersen

[permalink] [raw]
Subject: Re: [PATCH] scsi: megaraid_sas: Add flexible array member for SGLs


Kees,

> struct MPI2_RAID_SCSI_IO_REQUEST ends with a single SGL, but expects
> to copy multiple. Add a flexible array member so the compiler can
> reason about the size of the memcpy(). This will avoid the run-time
> false positive warning:

Applied to 6.3/scsi-staging, thanks!

--
Martin K. Petersen Oracle Linux Engineering

2023-01-27 03:23:41

by Martin K. Petersen

[permalink] [raw]
Subject: Re: [PATCH] scsi: megaraid_sas: Add flexible array member for SGLs

On Thu, 05 Jan 2023 21:32:00 -0800, Kees Cook wrote:

> struct MPI2_RAID_SCSI_IO_REQUEST ends with a single SGL, but expects to
> copy multiple. Add a flexible array member so the compiler can reason
> about the size of the memcpy(). This will avoid the run-time false
> positive warning:
>
> memcpy: detected field-spanning write (size 128) of single field "&r1_cmd->io_request->SGL" at drivers/scsi/megaraid/megaraid_sas_fusion.c:3326 (size 16)
>
> [...]

Applied to 6.3/scsi-queue, thanks!

[1/1] scsi: megaraid_sas: Add flexible array member for SGLs
https://git.kernel.org/mkp/scsi/c/a9a3629592ab

--
Martin K. Petersen Oracle Linux Engineering