I have used a few different packages over the years, and I have settled on Mendeley. Here are my reasons for liking it:
It is free (up to 2GB of storage, after which it will cost);
It can store the full pdf document as well as the metadata;
It can read the metadata automatically, though it doesn't always get it right;
It is available for Windows, Linux and Mac;
It can synchronise across multiple machines, although there are still some bugs in this option;
It has plug-ins for Word, Open Office and Libre Office;
It creates, and automatically updates, a bibtex file, which works extremely well with LaTeX;
It has full text search across papers;
It has smart filtering, tagging and automatic pdf file re-naming, which I don't really like;
It has a PDF viewer with sticky notes, text highlighting and full-screen reading options;
It has full text search across your full paper collection (indispensable);
It can import bibtex entries or complete files;
It can import and export files;
It has over 300 possible citation styles that can be used. Changing between them is as simple as pressing a few buttons. This is really cool if you need to change the referencing style for a conference or journal. Three clicks if the style is loaded, and it is easy to import more;
It has iPhone, iPad and Android apps, but not Windows phone yet;
It has various third party tools for migration of content, so it is really flexible;
It has an online copy of your library, which you can access from anywhere using a browser.
Of course there are drawbacks too:
I don't like the automatic renaming of files;
I get annoyed at the sporadic bugs on synching back and forward between machines (I have now allocated one computer as the master computer and only now synch one way, which seems to work);
Duplicate detection works well to find duplicates, but you need to check every single file to make sure you do not end up merging two different papers. I don't like that;
If you delete files, you can sometimes get a deleted ghost entry appearing in your library bibtex file, with no easy way to get rid of it;
When you sync across machines, you need to be disciplined to use a consistent citation reference, otherwise this can lead to issues.
All in all, by being careful, I can minimise the impact of the drawbacks. However, compare to other software, the benefits far outweigh these minor irritations.
@Bob thank you for detailed reply. I am also using Mendeley but these days experimenting with other freely available software. I need such a software which will export highlighted pdf text and comments into text file so that I can perform analysis on these in R.
There are many non-research specific research management software. We built Pillar Science https://pillar.science to allow us to have all our research in one place. It includes research, data and analysis management tools. I hope that helps!