[PATCH] megaraid.c

[PATCH] megaraid.c

[Praveen Srinivasan]

Hi,
This patch fixes an instance where an allocation is checked, but only after 
the pointer is memset() - moving the memset further down in the function 
fixes this.

Praveen Srinivasan

--- ../linux-fresh/./drivers/scsi/megaraid.c    Fri Apr 27 13:59:18 2001
+++ ./drivers/scsi/megaraid.c   Wed May 23 12:31:01 2001
@@ -4115,7 +4115,7 @@
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,0)        /* 0x020400 */
                scsicmd = (Scsi_Cmnd *) kmalloc (sizeof (Scsi_Cmnd),
                                                 GFP_KERNEL | GFP_DMA);
-               memset (scsicmd, 0, sizeof (Scsi_Cmnd));
+
 #else
                scsicmd = (Scsi_Cmnd *) scsi_init_malloc (sizeof (Scsi_Cmnd),
                                                          GFP_ATOMIC | 
GFP_DMA);
@@ -4127,6 +4127,9 @@
                        return -ENOMEM;
                }
 
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,0) /* 0x020400 */
+               memset (scsicmd, 0, sizeof (Scsi_Cmnd));
+#endif
                scsicmd->host = NULL;
 
                /*

Re: [PATCH] megaraid.c

[Alan Cox]

> This patch fixes an instance where an allocation is checked, but only after 
> the pointer is memset() - moving the memset further down in the function 
> fixes this.

There are a ton of these in the scsi code in Linus tree, Im still merging them
please grab the -ac patch before you waste time on this - there are about
30 or 40 of these already fixed.

The SCSI one is _much_ more problematic because there are pieces of quite high
level code that blindly assumed a scsi command alloc always works etc..

Re: [PATCH] megaraid.c

[Alan Cox]

> This patch fixes an instance where an allocation is checked, but only after 
> the pointer is memset() - moving the memset further down in the function 
>

Linus - please skip this one. There is a newer (1.15d) megaraid driver ready
to go on to you this weekend and it fixes this, pci api usage and security 
holes