Resource temporarily unavailable / Protocol not available when running redis with app-elfloader

[razvand]

Describe the bug

When using app-elfloader (via the run_app.sh script in run-app-elfloader) to run the redis dynamic build and redis static build, a Resource temporarily unavailable / Protocol not available issue appears.

Steps to reproduce

You can easily just use the ./do.sh script in the unikraft-upb/scripts repository:

$ ./do.sh setup
$ ./do.sh build
$ ./do.sh run_built redis

Then run:

$ redis-cli -h 172.44.0.2

And you'll get the error message.

Expected behavior

It should work to connect via redis-cli.

Which architectures were you using or does this bug affect?

x86_64

Which platforms were you using or does this bug affect?

kvm

Relevant log output

read(fd:3, <out>buf:0x4000efca7, 1) = Resource temporarily unavailable (-11)
gettimeofday(0x4000efcb0, 0x0, ...) = 0x0
epoll_wait(0x5, 0x10018f8000, ...) = 0x0
time(0x0, 0x10018f8000, ...) = 0x64416089
gettimeofday(0x4000efcb0, 0x0, ...) = 0x0
time(0x0, 0x0, ...) = 0x64416089
gettimeofday(0x4000efbe0, 0x0, ...) = 0x0
gettimeofday(0x4000efbd0, 0x0, ...) = 0x0
gettimeofday(0x4000efbd0, 0x0, ...) = 0x0
gettimeofday(0x4000efbd0, 0x0, ...) = 0x0
gettimeofday(0x4000efbd0, 0x0, ...) = 0x0
gettimeofday(0x4000efbd0, 0x0, ...) = 0x0
gettimeofday(0x4000efbd0, 0x0, ...) = 0x0
gettimeofday(0x4000efbd0, 0x0, ...) = 0x0
getpid() = pid:0
openat(AT_FDCWD, "/proc/0/stat", O_RDONLY) = fd:7
read(fd:7, <out>"16235 (bash) S 13550 162"..., 4096) = 353
close(fd:7) = OK
gettimeofday(0x4000efbd0, 0x0, ...) = 0x0
gettimeofday(0x4000efb50, 0x0, ...) = 0x0
gettimeofday(0x4000efb50, 0x0, ...) = 0x0
gettimeofday(0x4000efc60, 0x0, ...) = 0x0
gettimeofday(0x4000efc40, 0x0, ...) = 0x0
read(fd:3, <out>buf:0x4000efca7, 1) = Resource temporarily unavailable (-11)
gettimeofday(0x4000efcb0, 0x0, ...) = 0x0
epoll_wait(0x5, 0x10018f8000, ...) = 0x1
accept(0x6, 0x4000efb50, ...) = 0x7
fcntl(0x7, 0x3, ...) = 0x8002
fcntl(0x7, 0x4, ...) = 0x0
setsockopt(0x7, 0x6, ...) = 0x0
setsockopt(0x7, 0x1, ...) = 0x0
setsockopt(0x7, 0x6, ...) = Protocol not available (-92)
epoll_ctl(0x5, 0x1, ...) = 0x0
qemu-system-x86_64: terminating on signal 2

[marcrittinghaus]

The static one works. I don't see the read-error and I can connect with redis-cli. Make sure you merge the lib-lwip PR

• Add SOCK_NONBLOCK to accept4 and socket() lib-lwip#33

The dynamic redis-server from the dynamic-apps repo crashes for me with a stack corruption during startup

close(fd:3) = OK
mmap(NULL, 8192, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, fd:-1, 0) = va:0x1001111000
arch_prctl(0x1002, 0x1001111bc0, ...) = 0x0
mprotect(va:0x1001107000, 16384, PROT_READ) = OK
mprotect(va:0x1000d1a000, 4096, PROT_READ) = OK
mprotect(va:0x1000afd000, 8192, PROT_READ) = OK
mprotect(va:0x10008c9000, 4096, PROT_READ) = OK
mprotect(va:0x10006c3000, 4096, PROT_READ) = OK
mprotect(va:0x100030a000, 8192, PROT_READ) = OK
mprotect(va:0x40016e000, 4096, PROT_READ) = OK
set_tid_address(0x1001111e90, 0x4001bfe50, ...) = 0x0
set_robust_list(0x1001111ea0, 0x18, ...) = Function not implemented (-38)
rt_sigaction(0x20, 0x4001bfb90, ...) = 0x0
rt_sigaction(0x21, 0x4001bfb90, ...) = 0x0
rt_sigprocmask(0x1, 0x4001bfd08, ...) = 0x0
prlimit64(0x0, 0x3, ...) = 0x0

*** stack smashing detected ***: <unknown> terminated

writev(0x2, 0x4001bfb90, ...) = 0x36
mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, fd:-1, 0) = va:0x1001113000
rt_sigprocmask(0x1, 0x4001bfa60, ...) = 0x0
rt_sigprocmask(0x0, 0x4001bf9c0, ...) = 0x0
getpid() = pid:0
gettid() = pid:0
tgkill(0x0, 0x0, ...) = Function not implemented (-38)
rt_sigprocmask(0x2, 0x4001bf940, ...) = 0x0
rt_sigaction(0x6, 0x4001bf910, ...) = 0x0
rt_sigprocmask(0x0, 0x4001bf9c0, ...) = 0x0
getpid() = pid:0
gettid() = pid:0
tgkill(0x0, 0x0, ...) = Function not implemented (-38)
rt_sigprocmask(0x2, 0x4001bf940, ...) = 0x0

[nderjung]

unikraft/lib-lwip#33 is now merged. @razvand can you please re-check?

[razvand]

unikraft/lib-lwip#33