[PATCH v2] xen/hvc: replace BUG_ON() with negative return value

Xen frontends shouldn't BUG() in case of illegal data received from
their backends. So replace the BUG_ON()s when reading illegal data from
the ring page with negative return values.

Signed-off-by: Juergen Gross <jgross@suse.com>
---
V2:
- drop BUG_ON() (Christophe Leroy, Greg Kroah-Hartmann)
- replace WARN_ONCE() by pr_err_once() (Greg Kroah-Hartmann)
- break out from original series
---
 drivers/tty/hvc/hvc_xen.c | 17 ++++++++++++++---
 1 file changed, 14 insertions(+), 3 deletions(-)

diff --git a/drivers/tty/hvc/hvc_xen.c b/drivers/tty/hvc/hvc_xen.c
index 92c9a476defc..8f143c09a169 100644
--- a/drivers/tty/hvc/hvc_xen.c
+++ b/drivers/tty/hvc/hvc_xen.c

@@ -86,7 +86,11 @@ static int __write_console(struct xencons_info *xencons,
 	cons = intf->out_cons;
 	prod = intf->out_prod;
 	mb();			/* update queue values before going on */
-	BUG_ON((prod - cons) > sizeof(intf->out));
+
+	if ((prod - cons) > sizeof(intf->out)) {
+		pr_err_once("xencons: Illegal ring page indices");
+		return -EINVAL;
+	}
 
 	while ((sent < len) && ((prod - cons) < sizeof(intf->out)))
 		intf->out[MASK_XENCONS_IDX(prod++, intf->out)] = data[sent++];

@@ -114,7 +118,10 @@ static int domU_write_console(uint32_t vtermno, const char *data, int len)
 	 */
 	while (len) {
 		int sent = __write_console(cons, data, len);
-		
+
+		if (sent < 0)
+			return sent;
+
 		data += sent;
 		len -= sent;
 

@@ -138,7 +145,11 @@ static int domU_read_console(uint32_t vtermno, char *buf, int len)
 	cons = intf->in_cons;
 	prod = intf->in_prod;
 	mb();			/* get pointers before reading ring */
-	BUG_ON((prod - cons) > sizeof(intf->in));
+
+	if ((prod - cons) > sizeof(intf->in)) {
+		pr_err_once("xencons: Illegal ring page indices");
+		return -EINVAL;
+	}
 
 	while (cons != prod && recv < len)
 		buf[recv++] = intf->in[MASK_XENCONS_IDX(cons++, intf->in)];

-- 
2.26.2

On 07.07.2021 11:10, Juergen Gross wrote:

Xen frontends shouldn't BUG() in case of illegal data received from
their backends. So replace the BUG_ON()s when reading illegal data from
the ring page with negative return values.

Signed-off-by: Juergen Gross <jgross@suse.com>

Reviewed-by: Jan Beulich <redacted>

Largely unrelated note: While in general the barriers here may want
switching to virt_*mb(), this particular one looks to be too heavy
anyway: a read barrier is all that's needed here afaict, just like
there's only a write barrier between ring contents and producer
writing in __write_console().

And btw, since I've got puzzled by the linuxppc-dev@ in the recipients
list, I did look up relevant entries in ./MAINTAINERS. Shouldn't the
file be part of "XEN HYPERVISOR INTERFACE"?

Jan

On 07.07.21 11:57, Jan Beulich wrote:

On 07.07.2021 11:10, Juergen Gross wrote:

quoted

Xen frontends shouldn't BUG() in case of illegal data received from
their backends. So replace the BUG_ON()s when reading illegal data from
the ring page with negative return values.

Signed-off-by: Juergen Gross <jgross@suse.com>

Reviewed-by: Jan Beulich <redacted>

quoted

--- a/drivers/tty/hvc/hvc_xen.c
+++ b/drivers/tty/hvc/hvc_xen.c

@@ -86,7 +86,11 @@ static int __write_console(struct xencons_info *xencons,
  	cons = intf->out_cons;
  	prod = intf->out_prod;
  	mb();			/* update queue values before going on */

Largely unrelated note: While in general the barriers here may want
switching to virt_*mb(), this particular one looks to be too heavy
anyway: a read barrier is all that's needed here afaict, just like
there's only a write barrier between ring contents and producer
writing in __write_console().

I agree.

And btw, since I've got puzzled by the linuxppc-dev@ in the recipients
list, I did look up relevant entries in ./MAINTAINERS. Shouldn't the
file be part of "XEN HYPERVISOR INTERFACE"?

I wouldn't mind. Greg, Jiri, what do you think?


Juergen

On 07. 07. 21, 12:40, Juergen Gross wrote:

quoted

And btw, since I've got puzzled by the linuxppc-dev@ in the recipients
list, I did look up relevant entries in ./MAINTAINERS. Shouldn't the
file be part of "XEN HYPERVISOR INTERFACE"?

I wouldn't mind. Greg, Jiri, what do you think?

/me concurs.

thanks,
-- 
js
suse labs