There is no "dark l" sound in Proto-Germanic language and Proto-Indo-European language.
There are some sound changes between Proto-Germanic and Old English that are taken as evidence for the existence of a "dark" (i.e. velarized, uvularized or pharyngealized) allophone of /l/ (and /r/) in the pre-Old English time period.
This period is estimated to be c. AD 475–900. This includes changes from the split between Old English and Old Frisian (c. AD 475) up through historic early West Saxon of AD 900:
Breaking of front vowels.
- Most generally, before /x, w/, and /r, l/ + consonant (assumed to be velarized [rˠ, ɫ] in these circumstances), but exact conditioning factors vary from vowel to vowel.
Initial result was a falling diphthong ending in /u/, but this was followed by diphthong height harmonization, producing short /æ̆ɑ̆/, /ĕŏ/, /ĭŭ/ from short /æ/, /e/, /i/, long /æɑ/, /eo/, /iu/ from long /æː/, /eː/, /iː/.
- Written ea, eo, io, where length is not distinguished graphically.
Result in some dialects, for example Anglian, was back vowels rather than diphthongs. West Saxon ceald; but Anglian cald > ModE cold.
("Phonological history of Old English": Old English period)
The Wikipedia article "Phonological history of Old English" has a section "Breaking and retraction" that goes into a bit more detail. I can't personally vouch for its accuracy, but it mentions some details like "breaking does not occur before ll produced by West Germanic gemination (the /i/ or /j/ in the following syllable prevents breaking)" and "Short /e/ [...] is broken before l only in the combination lh and sometimes lc" that support the idea that vowel breaking was related to velarization.
So it would seem that it's safe to say that "dark l" existed at some point in one of the ancestors of English spoken before the turn of the first millennium, but I don't know how detailed a picture we can give of how its use varied between dialects and changed over time within dialects.
As far as I know, there is not much evidence for or against the existence of "dark l" as a phone in Proto-Germanic or Proto-Indo-European. We can find the phone in a number of other Germanic or Indo-European languages, but the rules for its distribution tend to vary between languages (just as they vary between different varieties of English). "Dark l" has a tendency to vocalize to /u/, /ɯ/, or /o/ (or non-syllabic equivalents) or to cause changes to the pronunciation of preceding vowels, but as far as I know, this is not an inevitable outcome, and I have the impression that it is not impossible for "dark l" to change to "light l" (even if the reverse change may be more common). For example, if I remember correctly, Latin "l pinguis", which is typically interpreted as a "dark l" of some kind based on its effects on adjacent vowels, has "light l" as a reflex in some Romance languages in certain contexts. So I don't think absence of evidence for a "dark l" phone in a proto-language can be interpreted as definitive evidence against the existence of "dark l" as a possible allophonic realization of *l.