[PATCH] driver/ata: fix potential null pointer dereference on pointer last_sge

[PATCH] driver/ata: fix potential null pointer dereference on pointer last_sge

[Chengfeng Ye]

The pointer cs_desc could be null if the loop is not
executed, so there is a potential NULL-PTR dereference
issue. Fix this by adding a null check before dereference.

Signed-off-by: Chengfeng Ye <cyeaa@connect.ust.hk>
---
 drivers/ata/sata_sil24.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/ata/sata_sil24.c b/drivers/ata/sata_sil24.c
index 06a1e27c4f84..2bd595da799f 100644
--- a/drivers/ata/sata_sil24.c
+++ b/drivers/ata/sata_sil24.c
@@ -785,6 +785,8 @@ static inline void sil24_fill_sg(struct ata_queued_cmd *qc,
 		sge++;
 	}
 
+	if (!last_sge)
+		return;
 	last_sge->flags = cpu_to_le32(SGE_TRM);
 }
 
-- 
2.17.1

Re: [PATCH] driver/ata: fix potential null pointer dereference on pointer last_sge

[Damien Le Moal]

On 2021/10/25 22:13, Chengfeng Ye wrote:
> The pointer cs_desc could be null if the loop is not
> executed, so there is a potential NULL-PTR dereference
> issue. Fix this by adding a null check before dereference.
> 
> Signed-off-by: Chengfeng Ye <cyeaa@connect.ust.hk>
> ---
>  drivers/ata/sata_sil24.c | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/drivers/ata/sata_sil24.c b/drivers/ata/sata_sil24.c
> index 06a1e27c4f84..2bd595da799f 100644
> --- a/drivers/ata/sata_sil24.c
> +++ b/drivers/ata/sata_sil24.c
> @@ -785,6 +785,8 @@ static inline void sil24_fill_sg(struct ata_queued_cmd *qc,
>  		sge++;
>  	}
>  
> +	if (!last_sge)
> +		return;
>  	last_sge->flags = cpu_to_le32(SGE_TRM);
>  }

I do not think that this fix is necessary: sil24_fill_sg() is called only if the
qc has ATA_QCFLAG_DMAMAP set (see sil24_qc_prep()) and that in turn is only set
if ata_sg_setup() sees at least one sg entry. So the loop in sil24_fill_sg()
will always be executed for qc marked with ATA_QCFLAG_DMAMAP and so last_sg
cannot be NULL.

-- 
Damien Le Moal
Western Digital Research

回复: [PATCH] driver/ata: fix potential null pointer dereference on pointer last_sge

[YE Chengfeng]

Got it. Thanks for your reply in such detail.

Best Regards,
Chengfeng

-----邮件原件-----
发件人: Damien Le Moal <damien.lemoal@opensource.wdc.com> 
发送时间: 2021年10月26日 12:10
收件人: YE Chengfeng <cyeaa@connect.ust.hk>
抄送: linux-kernel@vger.kernel.org
主题: Re: [PATCH] driver/ata: fix potential null pointer dereference on pointer last_sge

On 2021/10/25 22:13, Chengfeng Ye wrote:
> The pointer cs_desc could be null if the loop is not executed, so 
> there is a potential NULL-PTR dereference issue. Fix this by adding a 
> null check before dereference.
> 
> Signed-off-by: Chengfeng Ye <cyeaa@connect.ust.hk>
> ---
>  drivers/ata/sata_sil24.c | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/drivers/ata/sata_sil24.c b/drivers/ata/sata_sil24.c index 
> 06a1e27c4f84..2bd595da799f 100644
> --- a/drivers/ata/sata_sil24.c
> +++ b/drivers/ata/sata_sil24.c
> @@ -785,6 +785,8 @@ static inline void sil24_fill_sg(struct ata_queued_cmd *qc,
>  		sge++;
>  	}
>  
> +	if (!last_sge)
> +		return;
>  	last_sge->flags = cpu_to_le32(SGE_TRM);  }

I do not think that this fix is necessary: sil24_fill_sg() is called only if the qc has ATA_QCFLAG_DMAMAP set (see sil24_qc_prep()) and that in turn is only set if ata_sg_setup() sees at least one sg entry. So the loop in sil24_fill_sg() will always be executed for qc marked with ATA_QCFLAG_DMAMAP and so last_sg cannot be NULL.

--
Damien Le Moal
Western Digital Research