lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  PHC 
Open Source and information security mailing list archives
 
Hash Suite for Android: free password hash cracker in your pocket
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date:   Sat,  5 Dec 2020 20:17:43 +0100
From:   Rasmus Villemoes <rasmus.villemoes@...vas.dk>
To:     Li Yang <leoyang.li@....com>,
        "David S. Miller" <davem@...emloft.net>,
        Jakub Kicinski <kuba@...nel.org>
Cc:     Zhao Qiang <qiang.zhao@....com>,
        Vladimir Oltean <vladimir.oltean@....com>,
        Rasmus Villemoes <rasmus.villemoes@...vas.dk>,
        netdev@...r.kernel.org, linuxppc-dev@...ts.ozlabs.org,
        linux-kernel@...r.kernel.org
Subject: [PATCH 20/20] ethernet: ucc_geth: simplify rx/tx allocations

Since kmalloc() is nowadays [1] guaranteed to return naturally
aligned (i.e., aligned to the size itself) memory for power-of-2
sizes, we don't need to over-allocate the align amount, compute an
aligned address within the allocation, and (for later freeing) also
storing the original pointer [2].

Instead, just round up the length we want to allocate to the alignment
requirements, then round that up to the next power of 2. In theory,
this could allocate up to about twice as much memory as we needed.  In
practice, (a) kmalloc() would in most cases anyway return a
power-of-2-sized allocation and (b) with the default values of the
bdRingLen[RT]x fields, the length is already itself a power of 2
greater than the alignment.

So we actually end up saving memory compared to the current
situtation (e.g. for tx, we currently allocate 128+32 bytes, which
kmalloc() likely rounds up to 192 or 256; with this patch, we just
allocate 128 bytes.) Also struct ucc_geth_private becomes a little
smaller.

[1] 59bb47985c1d ("mm, sl[aou]b: guarantee natural alignment for
kmalloc(power-of-two)")

[2] That storing was anyway done in a u32, which works on 32 bit
machines, but is not very elegant and certainly makes a reader of the
code pause for a while.

Signed-off-by: Rasmus Villemoes <rasmus.villemoes@...vas.dk>
---
 drivers/net/ethernet/freescale/ucc_geth.c | 50 ++++++++---------------
 drivers/net/ethernet/freescale/ucc_geth.h |  2 -
 2 files changed, 17 insertions(+), 35 deletions(-)

diff --git a/drivers/net/ethernet/freescale/ucc_geth.c b/drivers/net/ethernet/freescale/ucc_geth.c
index 3cad2255bd97..a4ed9209513e 100644
--- a/drivers/net/ethernet/freescale/ucc_geth.c
+++ b/drivers/net/ethernet/freescale/ucc_geth.c
@@ -1835,7 +1835,7 @@ static void ucc_geth_free_rx(struct ucc_geth_private *ugeth)
 
 			kfree(ugeth->rx_skbuff[i]);
 
-			kfree((void *)ugeth->rx_bd_ring_offset[i]);
+			kfree(ugeth->p_rx_bd_ring[i]);
 			ugeth->p_rx_bd_ring[i] = NULL;
 		}
 	}
@@ -1872,10 +1872,8 @@ static void ucc_geth_free_tx(struct ucc_geth_private *ugeth)
 
 		kfree(ugeth->tx_skbuff[i]);
 
-		if (ugeth->p_tx_bd_ring[i]) {
-			kfree((void *)ugeth->tx_bd_ring_offset[i]);
-			ugeth->p_tx_bd_ring[i] = NULL;
-		}
+		kfree(ugeth->p_tx_bd_ring[i]);
+		ugeth->p_tx_bd_ring[i] = NULL;
 	}
 
 }
@@ -2150,25 +2148,15 @@ static int ucc_geth_alloc_tx(struct ucc_geth_private *ugeth)
 
 	/* Allocate Tx bds */
 	for (j = 0; j < ucc_geth_tx_queues(ug_info); j++) {
-		u32 align = UCC_GETH_TX_BD_RING_ALIGNMENT;
-
-		/* Allocate in multiple of
-		   UCC_GETH_TX_BD_RING_SIZE_MEMORY_ALIGNMENT,
-		   according to spec */
-		length = ((ug_info->bdRingLenTx[j] * sizeof(struct qe_bd))
-			  / UCC_GETH_TX_BD_RING_SIZE_MEMORY_ALIGNMENT)
-		    * UCC_GETH_TX_BD_RING_SIZE_MEMORY_ALIGNMENT;
-		if ((ug_info->bdRingLenTx[j] * sizeof(struct qe_bd)) %
-		    UCC_GETH_TX_BD_RING_SIZE_MEMORY_ALIGNMENT)
-			length += UCC_GETH_TX_BD_RING_SIZE_MEMORY_ALIGNMENT;
-
-		ugeth->tx_bd_ring_offset[j] =
-			(u32) kmalloc((u32) (length + align), GFP_KERNEL);
-
-		if (ugeth->tx_bd_ring_offset[j] != 0)
-			ugeth->p_tx_bd_ring[j] =
-				(u8 __iomem *)((ugeth->tx_bd_ring_offset[j] +
-						align) & ~(align - 1));
+		u32 align = max(UCC_GETH_TX_BD_RING_ALIGNMENT,
+				UCC_GETH_TX_BD_RING_SIZE_MEMORY_ALIGNMENT);
+		u32 alloc;
+
+		length = ug_info->bdRingLenTx[j] * sizeof(struct qe_bd);
+		alloc = round_up(length, align);
+		alloc = roundup_pow_of_two(alloc);
+
+		ugeth->p_tx_bd_ring[j] = kmalloc(alloc, GFP_KERNEL);
 
 		if (!ugeth->p_tx_bd_ring[j]) {
 			if (netif_msg_ifup(ugeth))
@@ -2176,9 +2164,7 @@ static int ucc_geth_alloc_tx(struct ucc_geth_private *ugeth)
 			return -ENOMEM;
 		}
 		/* Zero unused end of bd ring, according to spec */
-		memset_io((void __iomem *)(ugeth->p_tx_bd_ring[j] +
-		       ug_info->bdRingLenTx[j] * sizeof(struct qe_bd)), 0,
-		       length - ug_info->bdRingLenTx[j] * sizeof(struct qe_bd));
+		memset(ugeth->p_tx_bd_ring[j] + length, 0, alloc - length);
 	}
 
 	/* Init Tx bds */
@@ -2225,15 +2211,13 @@ static int ucc_geth_alloc_rx(struct ucc_geth_private *ugeth)
 	/* Allocate Rx bds */
 	for (j = 0; j < ucc_geth_rx_queues(ug_info); j++) {
 		u32 align = UCC_GETH_RX_BD_RING_ALIGNMENT;
+		u32 alloc;
 
 		length = ug_info->bdRingLenRx[j] * sizeof(struct qe_bd);
-		ugeth->rx_bd_ring_offset[j] =
-			(u32) kmalloc((u32) (length + align), GFP_KERNEL);
-		if (ugeth->rx_bd_ring_offset[j] != 0)
-			ugeth->p_rx_bd_ring[j] =
-				(u8 __iomem *)((ugeth->rx_bd_ring_offset[j] +
-						align) & ~(align - 1));
+		alloc = round_up(length, align);
+		alloc = roundup_pow_of_two(alloc);
 
+		ugeth->p_rx_bd_ring[j] = kmalloc(alloc, GFP_KERNEL);
 		if (!ugeth->p_rx_bd_ring[j]) {
 			if (netif_msg_ifup(ugeth))
 				pr_err("Can not allocate memory for Rx bd rings\n");
diff --git a/drivers/net/ethernet/freescale/ucc_geth.h b/drivers/net/ethernet/freescale/ucc_geth.h
index 6539fed9cc22..ccc4ca1ae9b6 100644
--- a/drivers/net/ethernet/freescale/ucc_geth.h
+++ b/drivers/net/ethernet/freescale/ucc_geth.h
@@ -1182,9 +1182,7 @@ struct ucc_geth_private {
 	struct ucc_geth_rx_bd_queues_entry __iomem *p_rx_bd_qs_tbl;
 	u32 rx_bd_qs_tbl_offset;
 	u8 __iomem *p_tx_bd_ring[NUM_TX_QUEUES];
-	u32 tx_bd_ring_offset[NUM_TX_QUEUES];
 	u8 __iomem *p_rx_bd_ring[NUM_RX_QUEUES];
-	u32 rx_bd_ring_offset[NUM_RX_QUEUES];
 	u8 __iomem *confBd[NUM_TX_QUEUES];
 	u8 __iomem *txBd[NUM_TX_QUEUES];
 	u8 __iomem *rxBd[NUM_RX_QUEUES];
-- 
2.23.0

Powered by blists - more mailing lists