[PATCH] simscsi breakage in __scsi_alloc_queue()

[PATCH] simscsi breakage in __scsi_alloc_queue()

[Al Viro]

With the current mainline we end up oopsing in __scsi_alloc_queue(), since
host->dma_dev is left NULL.  I'm not sure if this is the right fix, but
it gives the behaviour equivalent to what we used to have before
commit 6f381fa344911d5a234b13574433cf23036f9467 ([SCSI] scsi_lib: use
correct DMA device in __scsi_alloc_queue).  And with that applied the
damn thing appears to work correctly...

Current mainline barfs on dma_set_seg_boundary(dev, shost->dma_boundary),
since dev ends up being NULL.  AFAICS, the same problem affects any
driver that does scsi_add_host(host, NULL) and there are other such
beasts as well, so they probably also need fixing...

Again, I'm not sure that this is the right way to deal with that and I'd
prefer to have it reviewed by SCSI folks before it goes into the tree.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
---
diff --git a/arch/ia64/hp/sim/simscsi.c b/arch/ia64/hp/sim/simscsi.c
index 331de72..fe2b4d0 100644
--- a/arch/ia64/hp/sim/simscsi.c
+++ b/arch/ia64/hp/sim/simscsi.c
@@ -13,6 +13,7 @@
 #include <linux/interrupt.h>
 #include <linux/kernel.h>
 #include <linux/timer.h>
+#include <linux/platform_device.h>
 #include <asm/irq.h>
 #include "hpsim_ssc.h"
 
@@ -358,7 +359,7 @@ simscsi_init(void)
 	if (!host)
 		return -ENOMEM;
 
-	error = scsi_add_host(host, NULL);
+	error = scsi_add_host_with_dma(host, NULL, &platform_bus);
 	if (error)
 		goto free_host;
 	scsi_scan_host(host);

Re: [PATCH] simscsi breakage in __scsi_alloc_queue()

[James Bottomley]

On Fri, 2012-05-04 at 04:53 +0100, Al Viro wrote:
> With the current mainline we end up oopsing in __scsi_alloc_queue(), since
> host->dma_dev is left NULL.  I'm not sure if this is the right fix, but
> it gives the behaviour equivalent to what we used to have before
> commit 6f381fa344911d5a234b13574433cf23036f9467 ([SCSI] scsi_lib: use
> correct DMA device in __scsi_alloc_queue).  And with that applied the
> damn thing appears to work correctly...
> 
> Current mainline barfs on dma_set_seg_boundary(dev, shost->dma_boundary),
> since dev ends up being NULL.  AFAICS, the same problem affects any
> driver that does scsi_add_host(host, NULL) and there are other such
> beasts as well, so they probably also need fixing...
> 
> Again, I'm not sure that this is the right way to deal with that and I'd
> prefer to have it reviewed by SCSI folks before it goes into the tree.
> 
> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
> ---
> diff --git a/arch/ia64/hp/sim/simscsi.c b/arch/ia64/hp/sim/simscsi.c
> index 331de72..fe2b4d0 100644
> --- a/arch/ia64/hp/sim/simscsi.c
> +++ b/arch/ia64/hp/sim/simscsi.c
> @@ -13,6 +13,7 @@
>  #include <linux/interrupt.h>
>  #include <linux/kernel.h>
>  #include <linux/timer.h>
> +#include <linux/platform_device.h>
>  #include <asm/irq.h>
>  #include "hpsim_ssc.h"
>  
> @@ -358,7 +359,7 @@ simscsi_init(void)
>  	if (!host)
>  		return -ENOMEM;
>  
> -	error = scsi_add_host(host, NULL);
> +	error = scsi_add_host_with_dma(host, NULL, &platform_bus);

It's the right fix, but it's not general enough.  This should make sure
none of the legacy bus devices oops:

James

---

diff --git a/drivers/scsi/hosts.c b/drivers/scsi/hosts.c
index 351dc0b..a3a056a 100644
--- a/drivers/scsi/hosts.c
+++ b/drivers/scsi/hosts.c
@@ -218,6 +218,9 @@ int scsi_add_host_with_dma(struct Scsi_Host *shost, struct device *dev,
 
 	if (!shost->shost_gendev.parent)
 		shost->shost_gendev.parent = dev ? dev : &platform_bus;
+	if (!dma_dev)
+		dma_dev = shost->shost_gendev.parent;
+
 	shost->dma_dev = dma_dev;
 
 	error = device_add(&shost->shost_gendev);