bug#31785: Multiple client 'build-paths' RPCs can lead to daemon deadlock

[Ludovic Courtès]

Here’s another example from berlin today:

--8<---------------cut here---------------start------------->8---
ludo@berlin$ sudo guix processes
SessionPID: 38649
ClientPID: 38611
ClientCommand: /gnu/store/7x9lrzs2l00mfcr3ya7hg1i06i7lq9pd-profile/bin/guile \ 
/home/mbakke/guix/master/scripts/guix build --no-grafts --keep-going 
--max-silent-time=36000 --timeout=216000 -s armhf-linux nmoldyn r-msnid r-yamss 
netcdf-fortran python-netcdf4
LockHeld: /gnu/store/v6hcm1gvv3gdfbnq6cd4kn1i6ip3y74s-netcdf-fortran-4.4.4.lock
LockHeld: /gnu/store/ydxl0ws9kk6ix58m99ib7qhi157yq49h-r-msnid-1.18.1.lock
ChildProcess: 33181: 
/gnu/store/9alic3caqhay3h8mx4iihpmyj6ymqpcx-guile-2.2.4/bin/guile \ 
/gnu/store/vyhfp0gj4rx74yw0ybalzj8bvmqgp9pr-guix-1.0.1-4.c902458/bin/.guix-real 
offload x86_64-linux 36000 1 216000

[...]

SessionPID: 44043
ClientPID: 44023
ClientCommand: /gnu/store/7x9lrzs2l00mfcr3ya7hg1i06i7lq9pd-profile/bin/guile \ 
/home/mbakke/guix/master/scripts/guix build --no-grafts --keep-going 
--max-silent-time=36000 --timeout=216000 -s armhf-linux netcdf-fortran r-msnid 
nmoldyn r-yamss
LockHeld: /gnu/store/69jjabypcpx83482rh3yp4a9ncjimw6b-r-yamss-1.10.0.lock
LockHeld: /gnu/store/ibdbwxb1i8yal8ilqxxgnzh6bxpinic1-nmoldyn-3.0.11.lock
ChildProcess: 48997: 
/gnu/store/9alic3caqhay3h8mx4iihpmyj6ymqpcx-guile-2.2.4/bin/guile \ 
/gnu/store/vyhfp0gj4rx74yw0ybalzj8bvmqgp9pr-guix-1.0.1-4.c902458/bin/.guix-real 
offload x86_64-linux 36000 1 216000

ludo@berlin ~$ sudo strace -p 38649 -e openat,fcntl
strace: Process 38649 attached
openat(AT_FDCWD, 
"/gnu/store/ibdbwxb1i8yal8ilqxxgnzh6bxpinic1-nmoldyn-3.0.11.lock", 
O_RDWR|O_CREAT, 0600) = 10
fcntl(10, F_GETFD)                      = 0
fcntl(10, F_SETFD, FD_CLOEXEC)          = 0
fcntl(10, F_SETLK, {l_type=F_WRLCK, l_whence=SEEK_SET, l_start=0, l_len=0}) = 
-1 EAGAIN (Resource temporarily unavailable)
openat(AT_FDCWD, 
"/gnu/store/69jjabypcpx83482rh3yp4a9ncjimw6b-r-yamss-1.10.0.lock", 
O_RDWR|O_CREAT, 0600) = 10
fcntl(10, F_GETFD)                      = 0
fcntl(10, F_SETFD, FD_CLOEXEC)          = 0
fcntl(10, F_SETLK, {l_type=F_WRLCK, l_whence=SEEK_SET, l_start=0, l_len=0}) = 
-1 EAGAIN (Resource temporarily unavailable)
^Cstrace: Process 38649 detached

ludo@berlin ~$ sudo strace -p 44043 -e openat,fcntl
strace: Process 44043 attached
openat(AT_FDCWD, 
"/gnu/store/ydxl0ws9kk6ix58m99ib7qhi157yq49h-r-msnid-1.18.1.lock", 
O_RDWR|O_CREAT, 0600) = 10
fcntl(10, F_GETFD)                      = 0
fcntl(10, F_SETFD, FD_CLOEXEC)          = 0
fcntl(10, F_SETLK, {l_type=F_WRLCK, l_whence=SEEK_SET, l_start=0, l_len=0}) = 
-1 EAGAIN (Resource temporarily unavailable)
^Cstrace: Process 44043 detached
--8<---------------cut here---------------end--------------->8---

A very palpable deadlock.

Ludo’.