在 2022/4/14 12:39, Ramon Fried 写道:
On Sat, Apr 2, 2022 at 5:58 AM <qianfangui...@163.com> wrote:
From: qianfan Zhao <qianfangui...@163.com>
tftp_handler do eth_halt when TFTP_ERROR, but eth_halt will remove eth
device if it is an usb network. usbeth's private data will be unaccessable
when usb_eth_free_pkt, touch it will trigger data abort.
Next is the console messages:
=> tftp xxx
...
Loading: *
TFTP error: 'open failed: No such file or directory' (1)
Not retrying...
data abort
pc : [<9feb6ba2>] lr : [<9feb6b9f>]
Fix it.
Signed-off-by: qianfan Zhao <qianfangui...@163.com>
---
net/eth-uclass.c | 7 +++++++
1 file changed, 7 insertions(+)
diff --git a/net/eth-uclass.c b/net/eth-uclass.c
index 2b88b6c145..c9bba4f8de 100644
--- a/net/eth-uclass.c
+++ b/net/eth-uclass.c
@@ -407,6 +407,13 @@ int eth_rx(void)
flags = 0;
if (ret > 0)
net_process_received_packet(packet, ret);
+
+ /* ethernet maybe halted when packet_handler, check again */
+ if (!eth_is_active(current)) {
+ ret = 0;
+ break;
+ }
+
if (ret >= 0 && eth_get_ops(current)->free_pkt)
eth_get_ops(current)->free_pkt(current, packet, ret);
if (ret <= 0)
--
2.17.1
I wonder how I never experienced this behavior. I have used USB based
ethernet dongles.
And I used tftp and got errors but never an abort .Is it possible that
the problem is with the specific ethernet driver you are using ?
Can you specify which one you used ?
I had check the pc location of data abort, it's on dlmalloc. The bug had
changed those memory already freed. Cause dlmalloc trigger data abort.
This bug is happens on am335x device. nothing is specific, use
am335x_evm_defconfig can reproduce this.