netdev.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Michael Chan <michael.chan@broadcom.com>
To: davem@davemloft.net
Cc: netdev@vger.kernel.org, kuba@kernel.org
Subject: [PATCH net-next 5/5] bnxt_en: Enable batch mode when using HWRM_NVM_MODIFY to flash packages.
Date: Sun, 13 Dec 2020 06:51:46 -0500	[thread overview]
Message-ID: <1607860306-17244-6-git-send-email-michael.chan@broadcom.com> (raw)
In-Reply-To: <1607860306-17244-1-git-send-email-michael.chan@broadcom.com>

The current scheme allocates a DMA buffer as big as the requested
firmware package file and DMAs the contents to firmware in one
operation.  The buffer size can be several hundred kilo bytes and
the driver may not be able to allocate the memory.  This will cause
firmware upgrade to fail.

Improve the scheme by using smaller DMA blocks and calling firmware to
DMA each block in a batch mode.  Older firmware can cause excessive
NVRAM erases if the block size is too small so we try to allocate a
256K buffer to begin with and size it down successively if we cannot
allocate the memory.

Reviewed-by: Edwin Peer <edwin.peer@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
---
 .../net/ethernet/broadcom/bnxt/bnxt_ethtool.c | 49 +++++++++++++++----
 1 file changed, 40 insertions(+), 9 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c b/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c
index 38ab882715c4..9ff79d5d14c4 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c
@@ -2432,6 +2432,10 @@ static int bnxt_flash_firmware_from_file(struct net_device *dev,
 	return rc;
 }
 
+#define BNXT_PKG_DMA_SIZE	0x40000
+#define BNXT_NVM_MORE_FLAG	(cpu_to_le16(NVM_MODIFY_REQ_FLAGS_BATCH_MODE))
+#define BNXT_NVM_LAST_FLAG	(cpu_to_le16(NVM_MODIFY_REQ_FLAGS_BATCH_LAST))
+
 int bnxt_flash_package_from_fw_obj(struct net_device *dev, const struct firmware *fw,
 				   u32 install_type)
 {
@@ -2442,6 +2446,7 @@ int bnxt_flash_package_from_fw_obj(struct net_device *dev, const struct firmware
 	bool defrag_attempted = false;
 	dma_addr_t dma_handle;
 	u8 *kmem = NULL;
+	u32 modify_len;
 	u32 item_len;
 	int rc = 0;
 	u16 index;
@@ -2450,8 +2455,19 @@ int bnxt_flash_package_from_fw_obj(struct net_device *dev, const struct firmware
 
 	bnxt_hwrm_cmd_hdr_init(bp, &modify, HWRM_NVM_MODIFY, -1, -1);
 
-	kmem = dma_alloc_coherent(&bp->pdev->dev, fw->size, &dma_handle,
-				  GFP_KERNEL);
+	/* Try allocating a large DMA buffer first.  Older fw will
+	 * cause excessive NVRAM erases when using small blocks.
+	 */
+	modify_len = roundup_pow_of_two(fw->size);
+	modify_len = min_t(u32, modify_len, BNXT_PKG_DMA_SIZE);
+	while (1) {
+		kmem = dma_alloc_coherent(&bp->pdev->dev, modify_len,
+					  &dma_handle, GFP_KERNEL);
+		if (!kmem && modify_len > PAGE_SIZE)
+			modify_len /= 2;
+		else
+			break;
+	}
 	if (!kmem)
 		return -ENOMEM;
 
@@ -2463,6 +2479,8 @@ int bnxt_flash_package_from_fw_obj(struct net_device *dev, const struct firmware
 	install.install_type = cpu_to_le32(install_type);
 
 	do {
+		u32 copied = 0, len = modify_len;
+
 		rc = bnxt_find_nvram_item(dev, BNX_DIR_TYPE_UPDATE,
 					  BNX_DIR_ORDINAL_FIRST,
 					  BNX_DIR_EXT_NONE,
@@ -2479,14 +2497,26 @@ int bnxt_flash_package_from_fw_obj(struct net_device *dev, const struct firmware
 		}
 
 		modify.dir_idx = cpu_to_le16(index);
-		modify.len = cpu_to_le32(fw->size);
 
-		memcpy(kmem, fw->data, fw->size);
-		rc = hwrm_send_message(bp, &modify, sizeof(modify),
-				       FLASH_PACKAGE_TIMEOUT);
-		if (rc)
-			break;
+		if (fw->size > modify_len)
+			modify.flags = BNXT_NVM_MORE_FLAG;
+		while (copied < fw->size) {
+			u32 balance = fw->size - copied;
 
+			if (balance <= modify_len) {
+				len = balance;
+				if (copied)
+					modify.flags |= BNXT_NVM_LAST_FLAG;
+			}
+			memcpy(kmem, fw->data + copied, len);
+			modify.len = cpu_to_le32(len);
+			modify.offset = cpu_to_le32(copied);
+			rc = hwrm_send_message(bp, &modify, sizeof(modify),
+					       FLASH_PACKAGE_TIMEOUT);
+			if (rc)
+				goto pkg_abort;
+			copied += len;
+		}
 		mutex_lock(&bp->hwrm_cmd_lock);
 		rc = _hwrm_send_message_silent(bp, &install, sizeof(install),
 					       INSTALL_PACKAGE_TIMEOUT);
@@ -2530,7 +2560,8 @@ int bnxt_flash_package_from_fw_obj(struct net_device *dev, const struct firmware
 		mutex_unlock(&bp->hwrm_cmd_lock);
 	} while (defrag_attempted && !rc);
 
-	dma_free_coherent(&bp->pdev->dev, fw->size, kmem, dma_handle);
+pkg_abort:
+	dma_free_coherent(&bp->pdev->dev, modify_len, kmem, dma_handle);
 	if (resp.result) {
 		netdev_err(dev, "PKG install error = %d, problem_item = %d\n",
 			   (s8)resp.result, (int)resp.problem_item);
-- 
2.18.1


  parent reply	other threads:[~2020-12-13 12:00 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-12-13 11:51 [PATCH net-next 0/5] bnxt_en: Improve firmware flashing Michael Chan
2020-12-13 11:51 ` [PATCH net-next 1/5] bnxt_en: Refactor bnxt_flash_nvram Michael Chan
2020-12-13 11:51 ` [PATCH net-next 2/5] bnxt_en: Rearrange the logic in bnxt_flash_package_from_fw_obj() Michael Chan
2020-12-13 11:51 ` [PATCH net-next 3/5] bnxt_en: Restructure bnxt_flash_package_from_fw_obj() to execute in a loop Michael Chan
2020-12-13 11:51 ` [PATCH net-next 4/5] bnxt_en: Retry installing FW package under NO_SPACE error condition Michael Chan
2020-12-13 11:51 ` Michael Chan [this message]
2020-12-15  3:10 ` [PATCH net-next 0/5] bnxt_en: Improve firmware flashing patchwork-bot+netdevbpf

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1607860306-17244-6-git-send-email-michael.chan@broadcom.com \
    --to=michael.chan@broadcom.com \
    --cc=davem@davemloft.net \
    --cc=kuba@kernel.org \
    --cc=netdev@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).