that’s bad
objdump indeed different project (binutils), but it have to work, even on modern binaries.
ok, give me u-boot file
FreeBSD always helps :)))))
it’s uploaded in gdrive-link above
look like it is net_set_ip_header, seems due not aligned access
r0 : 7efd874e & 0x3 = 0x2 - must be 0, or it have to be accesses with mem ops with w or b suffix
do tftp crashes too?
Using mtk_eth device
TFTP from server 192.168.0.10; our IP address is 192.168.0.18
Filename 'u-boot-mtk_r64_sd_rtl8367_gcc8.3.bin'.
Load address: 0x41dffe00
Loading: data abort
pc : [<7ef5f6b4>] lr : [<7ef5f760>]
sp : 7cf29af8 ip : 00000033 fp : 00000045
r10: 00000dd7 r9 : 7cf29f40 r8 : 7efd8740
r7 : 00000dd7 r6 : 00000045 r5 : 00000044 r4 : 7efd874e
r3 : 14000045 r2 : 1200a8c0 r1 : 0a00a8c0 r0 : 7efd874e
Flags: nZCv IRQs off FIQs off Mode SVC_32
Resetting CPU ...
yes
how did you get this output? if it is in net_set_ip_header it should affect also the other switch
can i somehow force aligned access? function itself looks not wrong
./net/net.c
1380 void net_set_ip_header(uchar *pkt, IPaddr_t dest, IPaddr_t source)
1381 {
1382 struct ip_udp_hdr *ip = (struct ip_udp_hdr *)pkt;
1383
1384 /*
1385 * Construct an IP header.
1386 */
1387 /* IP_HDR_SIZE / 4 (not including UDP) */
1388 ip->ip_hl_v = 0x45;
1389 ip->ip_tos = 0;
1390 ip->ip_len = htons(IP_HDR_SIZE);
1391 ip->ip_id = htons(NetIPID++);
1392 ip->ip_off = htons(IP_FLAGS_DFRAG); /* Don't fragment */
1393 ip->ip_ttl = 255;
1394 ip->ip_sum = 0;
1395 /* already in network byte order */
1396 NetCopyIP((void *)&ip->ip_src, &source);
1397 /* already in network byte order */
1398 NetCopyIP((void *)&ip->ip_dst, &dest);
1399 }
i guess crash happens in the last 4 lines (=2 without comments)
try to add that line at beginning of net_set_ip_header
printf(“pkt=%p\n”, pkt);
I think it will be
7efd874e
or a bit moved by added code with printf
rtk_switch_init ret = 0!!!!!!!!!!!!
Set RTL8367S SGMII 2.5Gbps
0x1b000014 = 0x00110214
Using mtk_eth device
pkt=7efd878e
data abort
disable optimization? -O0 in Makefile KBUILD_FLAGS?
rtk_switch_init ret = 0!!!!!!!!!!!!
Set RTL8367S SGMII 2.5Gbps
0x1b000014 = 0x00110214
Using mtk_eth device
pkt=7efd89ce
ping failed; host 192.168.0.10 is not alive
BPI-R64>
address also not aligned and ping not possible, but also no crash/reset, and not that fast as with -Os…sometimes it hangs a bit
second ping goes through
BPI-R64> ping 192.168.0.10
Using mtk_eth device
pkt=7efd89ce
host 192.168.0.10 is alive
BPI-R64>
seems switch takes some time for setup…too long for first ping
uboot-binary is ~200kb larger, at least it is proofed, that optimization brings this problem, not alignment
https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html
-Os (previous value)
Optimize for size. -Os enables all -O2 optimizations except those that often increase code size:
-falign-functions -falign-jumps
-falign-labels -falign-loops
-fprefetch-loop-arrays -freorder-blocks-algorithm=stc
It also enables -finline-functions, causes the compiler to tune for code size rather than execution speed, and performs further optimizations designed to reduce code size.
-O2:
-falign-functions -falign-jumps
-falign-labels -falign-loops
-fcaller-saves
-fcode-hoisting
-fcrossjumping
-fcse-follow-jumps -fcse-skip-blocks
-fdelete-null-pointer-checks
-fdevirtualize -fdevirtualize-speculatively
-fexpensive-optimizations
-ffinite-loops
-fgcse -fgcse-lm
-fhoist-adjacent-loads
-finline-small-functions
-findirect-inlining
-fipa-bit-cp -fipa-cp -fipa-icf
-fipa-ra -fipa-sra -fipa-vrp
-fisolate-erroneous-paths-dereference
-flra-remat
-foptimize-sibling-calls
-foptimize-strlen
-fpartial-inlining
-fpeephole2
-freorder-blocks-algorithm=stc
-freorder-blocks-and-partition -freorder-functions
-frerun-cse-after-loop
-fschedule-insns -fschedule-insns2
-fsched-interblock -fsched-spec
-fstore-merging
-fstrict-aliasing
-fthread-jumps
-ftree-builtin-call-dce
-ftree-pre
-ftree-switch-conversion -ftree-tail-merge
-ftree-vrp
is there any option i should disable? when using -Os
except
-falign-functions -falign-jumps
-falign-labels -falign-loops
sounds wrong in this context
mhm, O2 also crashes ;( O1 does not crash
1 Like
https://lists.denx.de/pipermail/u-boot/2017-July/297998.html
Conclusion:
This bug can be fixed by either:
1) Always ensure proper alignment of the udp packet header
pointers, which are passed to the net_set_ip_header() function
and similar functions. Some further investigation is necessary.
or
2) Just add a "packed" attribute to the ip_udp_hdr struct. In this
case the compiler will always assume the smallest 1 byte alignment.
It may have some performance implications though.Yeah, usually we are transferring gigabytes of traffic with U-Boot and in most cases U-Boot’s network performance close to line speed :)))))))))))
Of course just select 2nd
I try it next week,i’m currently away from home and my repo has -O1 as workaround 
tried to change this like this way, but same result (crash)
--- a/include/net.h
+++ b/include/net.h
@@ -273,7 +273,7 @@ struct ip_hdr {
/*
* Internet Protocol (IP) + UDP header.
*/
-struct ip_udp_hdr {
+struct __attribute__((__packed__)) ip_udp_hdr {
uchar ip_hl_v; /* header length and version */
uchar ip_tos; /* type of service */
ushort ip_len; /* total length */
i have also set -mno-unaligned-access in KBUILD_FLAGS (where i had changed the optimization), same result
-O1 works, but -O2/-Os doesn’t, can’t we just disable one or more options from O2 to have best optimization without crash? which may cause the crash?
edit: nailed it down to this: -fstore-merging
i got no crash if using
KBUILD_CFLAGS += -Os -fno-store-mergingBut it seems still unfixed in upstream…
Even later releases? Or you about BPI one?
i’ve found no commit in uboot for this issue…thats why i asked tom (CC to mailinglist)