diff mbox

[RFC,3/4] sparc: remove __GFP_NOFAIL reuquirement

Message ID 20150302203304.GA20513@dhcp22.suse.cz
State RFC, archived
Delegated to: David Miller
Headers show

Commit Message

Michal Hocko March 2, 2015, 8:33 p.m. UTC
On Mon 02-03-15 15:04:05, David S. Miller wrote:
> From: Michal Hocko <mhocko@suse.cz>
> Date: Mon,  2 Mar 2015 14:54:42 +0100
> 
> > mdesc_kmalloc is currently requiring __GFP_NOFAIL allocation although it
> > seems that the allocation failure is handled by all callers (via
> > mdesc_alloc). __GFP_NOFAIL is a strong liability for the memory
> > allocator and so the users are discouraged to use the flag unless the
> > allocation failure is really a nogo. Drop the flag here as this doesn't
> > seem to be the case.
> > 
> > Signed-off-by: Michal Hocko <mhocko@suse.cz>
> 
> It is a serious failure.
> 
> If we miss an MDESC update due to this allocation failure, the update
> is not an event which gets retransmitted so we will lose the updated
> machine description forever.
> 
> We really need this allocation to succeed.

OK, thanks for the clarification. This wasn't clear from the commit
which has introduced this code. I will drop this patch. Would you
accept something like the following instead?
---

Comments

David Miller March 2, 2015, 8:44 p.m. UTC | #1
From: Michal Hocko <mhocko@suse.cz>
Date: Mon, 2 Mar 2015 21:33:04 +0100

> On Mon 02-03-15 15:04:05, David S. Miller wrote:
>> From: Michal Hocko <mhocko@suse.cz>
>> Date: Mon,  2 Mar 2015 14:54:42 +0100
>> 
>> > mdesc_kmalloc is currently requiring __GFP_NOFAIL allocation although it
>> > seems that the allocation failure is handled by all callers (via
>> > mdesc_alloc). __GFP_NOFAIL is a strong liability for the memory
>> > allocator and so the users are discouraged to use the flag unless the
>> > allocation failure is really a nogo. Drop the flag here as this doesn't
>> > seem to be the case.
>> > 
>> > Signed-off-by: Michal Hocko <mhocko@suse.cz>
>> 
>> It is a serious failure.
>> 
>> If we miss an MDESC update due to this allocation failure, the update
>> is not an event which gets retransmitted so we will lose the updated
>> machine description forever.
>> 
>> We really need this allocation to succeed.
> 
> OK, thanks for the clarification. This wasn't clear from the commit
> which has introduced this code. I will drop this patch. Would you
> accept something like the following instead?

Sure.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
diff mbox

Patch

diff --git a/arch/sparc/kernel/mdesc.c b/arch/sparc/kernel/mdesc.c
index 99632a87e697..26c80e18d7b1 100644
--- a/arch/sparc/kernel/mdesc.c
+++ b/arch/sparc/kernel/mdesc.c
@@ -130,26 +130,26 @@  static struct mdesc_mem_ops memblock_mdesc_ops = {
 static struct mdesc_handle *mdesc_kmalloc(unsigned int mdesc_size)
 {
 	unsigned int handle_size;
+	struct mdesc_handle *hp;
+	unsigned long addr;
 	void *base;
 
 	handle_size = (sizeof(struct mdesc_handle) -
 		       sizeof(struct mdesc_hdr) +
 		       mdesc_size);
 
+	/*
+	 * Allocation has to succeed because mdesc update would be missed
+	 * and such events are not retransmitted.
+	 */
 	base = kmalloc(handle_size + 15, GFP_KERNEL | __GFP_NOFAIL);
-	if (base) {
-		struct mdesc_handle *hp;
-		unsigned long addr;
-
-		addr = (unsigned long)base;
-		addr = (addr + 15UL) & ~15UL;
-		hp = (struct mdesc_handle *) addr;
+	addr = (unsigned long)base;
+	addr = (addr + 15UL) & ~15UL;
+	hp = (struct mdesc_handle *) addr;
 
-		mdesc_handle_init(hp, handle_size, base);
-		return hp;
-	}
+	mdesc_handle_init(hp, handle_size, base);
 
-	return NULL;
+	return hp;
 }
 
 static void mdesc_kfree(struct mdesc_handle *hp)