
On Wed, Nov 08, 2023 at 09:00:23PM -0300, Fabio Estevam wrote:
[Adding Tom and Marek]
On Wed, Nov 8, 2023 at 7:22 PM Tim Harvey tharvey@gateworks.com wrote:
Greetings,
I've recently given wget/tcp a try and find the results inconsistent.
Transferring the same file 15 times in a row will eventually give me an incomplete transfer with packet/byte count wrong but showing 'Transfer Successful'.
Here is current master on an imx8mp-venice-gw74xx board using IMX8MP EQOS enet dev:
u-boot=> time wget $loadaddr /tftpboot/venice/core-image-minimal.ext4.gz && md5sum $loadaddr $filesize HTTP/1.0 200 OK| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Packets received 64368, Transfer Successful Bytes transferred = 93198937 (58e1a59 hex)
time: 3.103 seconds md5 for 48200000 ... 4dae1a58 ==> 69bedfe899148c43c09a98ac950dd0ed
....
u-boot=> time wget $loadaddr /tftpboot/venice/core-image-minimal.ext4.gz && md5sum $loadaddr $filesize HTTP/1.0 200 OK| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Packets received 64368, Transfer Successful Bytes transferred = 26984682 (19bc0ea hex)
time: 3.011 seconds md5 for 48200000 ... 49bbc0e9 ==> c9978ff116605a7729a1cb14ef51fc20
Anyone else seeing this?
Yes, I am seeing the same behavior on an imx8mn board using the FEC driver.
I am transferring a 412MB file. Doing the wget transfers several times in a row, I also see the incorrect number of bytes transferred:
Interesting. What if you take wget out of the equation and use "tftp" to grab a large file with a known checksum, repeatedly (note that "rhash -C" will give a crc32 in Linux so you can confirm the correct value and loop a get, crc32, compare, repeat loop with a little work) ? I ask because I was reminded today that for me at least, the RPi ethernet isn't reliable in U-Boot for me if you stress it too hard. So I'd like to try see if this is "wget bug" or "network driver/related bug".