[PATCH 1/1] Properly check return values of kmalloc and vmbus_recvpacket

[Jesper Juhl]

On Thu, 9 Dec 2010, Ky Srinivasan wrote:

> 
> 
> >>> On 12/9/2010 at  3:23 PM, in message
> <1291926209-17120-1-git-send-email-hjanssen at microsoft.com>, Hank Janssen
> <hjanssen at microsoft.com> wrote: 
> > Correct ugly oversight, we need to check the return values of kmalloc
> > and vmbus_recvpacket and return if they fail. I also tightened up the
> > call to kmalloc. 
> > 
> > Thanks to Evgeniy Polyakov <zbr at ioremap.net>  for pointing this out.
> > 
> > Signed-off-by: Haiyang Zhang <haiyangz at microsoft.com>
> > Signed-off-by: Hank Janssen <hjanssen at microsoft.com>
> > 
> > ---
> >  drivers/staging/hv/hv_utils.c |   48 ++++++++++++++++++++++++++++------------
> >  1 files changed, 33 insertions(+), 15 deletions(-)
> > 
> > diff --git a/drivers/staging/hv/hv_utils.c b/drivers/staging/hv/hv_utils.c
> > index 53e1e29..ac68575 100644
> > --- a/drivers/staging/hv/hv_utils.c
> > +++ b/drivers/staging/hv/hv_utils.c
> > @@ -43,21 +43,27 @@ static void shutdown_onchannelcallback(void *context)
> >  {
> >  	struct vmbus_channel *channel = context;
> >  	u8 *buf;
> > -	u32 buflen, recvlen;
> > +	u32 recvlen;
> >  	u64 requestid;
> >  	u8  execute_shutdown = false;
> > +	int ret = 0;
> >  
> >  	struct shutdown_msg_data *shutdown_msg;
> >  
> >  	struct icmsg_hdr *icmsghdrp;
> >  	struct icmsg_negotiate *negop = NULL;
> >  
> > -	buflen = PAGE_SIZE;
> > -	buf = kmalloc(buflen, GFP_ATOMIC);
> > +	buf = kmalloc(PAGE_SIZE, GFP_ATOMIC);
> >  
> > -	vmbus_recvpacket(channel, buf, buflen, &recvlen, &requestid);
> > +	if (!buf) {
> > +		printk(KERN_INFO
> > +		       "Unable to allocate memory for shutdown_onchannelcallback");
> > +		return;
> > +	}
> > +
> > +	ret = vmbus_recvpacket(channel, buf, PAGE_SIZE, &recvlen, &requestid);
> >  
> > -	if (recvlen > 0) {
> > +	if (ret == 0 && recvlen > 0) {
> >  		DPRINT_DBG(VMBUS, "shutdown packet: len=%d, requestid=%lld",
> >  			   recvlen, requestid);
> >  
> > @@ -151,17 +157,23 @@ static void timesync_onchannelcallback(void *context)
> >  {
> >  	struct vmbus_channel *channel = context;
> >  	u8 *buf;
> > -	u32 buflen, recvlen;
> > +	u32 recvlen;
> >  	u64 requestid;
> >  	struct icmsg_hdr *icmsghdrp;
> >  	struct ictimesync_data *timedatap;
> > +	int ret = 0;
> >  
> > -	buflen = PAGE_SIZE;
> > -	buf = kmalloc(buflen, GFP_ATOMIC);
> > +	buf = kmalloc(PAGE_SIZE, GFP_ATOMIC);
> >  
> > -	vmbus_recvpacket(channel, buf, buflen, &recvlen, &requestid);
> > +	if (!buf) {
> > +		printk(KERN_INFO
> > +		       "Unable to allocate memory for timesync_onchannelcallback");
> > +		return;
> > +	}
> >  
> > -	if (recvlen > 0) {
> > +	ret = vmbus_recvpacket(channel, buf, PAGE_SIZE, &recvlen, &requestid);
> > +
> > +	if (ret == 0 && recvlen > 0) {
> >  		DPRINT_DBG(VMBUS, "timesync packet: recvlen=%d, requestid=%lld",
> >  			recvlen, requestid);
> >  
> > @@ -197,17 +209,23 @@ static void heartbeat_onchannelcallback(void *context)
> >  {
> >  	struct vmbus_channel *channel = context;
> >  	u8 *buf;
> > -	u32 buflen, recvlen;
> > +	u32 recvlen;
> >  	u64 requestid;
> >  	struct icmsg_hdr *icmsghdrp;
> >  	struct heartbeat_msg_data *heartbeat_msg;
> > +	int ret = 0;
> > +
> > +	buf = kmalloc(PAGE_SIZE, GFP_ATOMIC);
> >  
> > -	buflen = PAGE_SIZE;
> > -	buf = kmalloc(buflen, GFP_ATOMIC);
> > +	if (!buf) {
> > +		printk(KERN_INFO
> > +		    "Unable to allocate memory for heartbeat_onchannelcallback");
> > +		return;
> > +	}
> >  
> > -	vmbus_recvpacket(channel, buf, buflen, &recvlen, &requestid);
> > +	ret = vmbus_recvpacket(channel, buf, PAGE_SIZE, &recvlen, &requestid);
> >  
> > -	if (recvlen > 0) {
> > +	if (ret == 0 && recvlen > 0) {
> >  		DPRINT_DBG(VMBUS, "heartbeat packet: len=%d, requestid=%lld",
> >  			   recvlen, requestid);
> >  
> I am wondering if it might be better to allocate a page during module 
> startup for dealing with some of these services. Since the protocol 
> guarantees that there cannot be more than one outstanding request from 
> the host side, having a pre-allocated receive buffer would be a more 
> robust solution - we don't have to allocate memory when we cannot afford 
> to sleep and thereby don't have to deal with a class of failure 
> conditions that are not easy to deal with. For instance not being able 
> to shut the guest down because of low memory situation would be bad.
> 

IMVHO; nicer to have a module fail at load time with ENOMEM than having 
failures later. If memory requirement is known at load time (and is fairly 
low), just allocate what's needed then and then there's no unexpected 
failure later.

-- 
Jesper Juhl <jj at chaosbits.net>            http://www.chaosbits.net/
Don't top-post http://www.catb.org/~esr/jargon/html/T/top-post.html
Plain text mails only, please.