Not the first, see eg. Burroughs mainframes, early 60's.
JOVIAL was another language, used for systems programming on the militay side, predating C and proven.
Pascal was contemporary with C and UCSD p-System was released just 2 years later than Unix. It was also more portable and retargetable than Unix.
If anyone here is interested in experimenting with a Bliss compiler, there's one installed on the gein.vistech.net VMS server, and there are free accounts available there. Visit http://deathrow.vistech.net for account registration details.
FWIW, the UCSD P-System and p-code and the Pascal compiler was easily portable, but the tools were also very buggy (at least) on the Terak boxes, and — even for the era — very slow. Folks were always working around some bug or another, or some slowness.
True story: I took over a numerical simulation on a 2D grid that ran slowly, because it implemented the 2D grid as a linked list. If you wanted to refer to the node directly below the current one, and the grid was 60x60, you had to follow 60 links to do it! I looked down on the guy who wrote it as being somehow hung up on linked lists, like it was his favorite data structure from school or something. Only later did I find out that he had no choice - he couldn't allocate a variable-sized array. So we re-wrote it to allocate the biggest array we had memory to hold, and to only use part of it.
But if we were trying to write a portable operating system, or something of that sort, our kludge would not have worked. Allocating the biggest possible size needed is not good behavior for an OS...