"With the help of Google, Yahoo, Bing and others, we found 770 unique URIs that had been cached and which contained leaked memory... The leaked memory has been purged with the help of the search engines."
I wonder if the search engines found any interesting patterns of searches performed whose results contained these 770 unique URIs.
Calling this a "memory leak" is rather misleading. The problem is literally leaking the contents of memory, yes, but the term memory leak usually means something else, something that is usually not as problematic as dumping random memory to the internet.
No. If you have any of those features enabled Email obfuscation, Server-side Excludes and Automatic HTTPS Rewrites as well as you have an invalid HTML markup. Per their blog post it is 0.00003% of the total requests.
I thought those were the conditions that caused PII to be included in the HTML response, but the PII could come from any customer as it's a buffer overrun situation. Damn this is confusing
Basically yes. If some of your traffic was flowing through a server on which any customer had the vulnerable features enabled, parts of your traffic may have been dumped.
See also the Project Zero ticket https://bugs.chromium.org/p/project-zero/issues/detail?id=11... and associated Hacker News discussion https://news.ycombinator.com/item?id=13718752
I wonder if the search engines found any interesting patterns of searches performed whose results contained these 770 unique URIs.
I've always wondered if Google did that. Seems obvious enough to just watch people using premade searches and see what they are up to.
@jgrahamc:
What are your thoughts about Tavis' remark that Cloudflare's communications (the "draft", post-mortem, etc.) "severely downplays" this issue?
Calling this a "memory leak" is rather misleading. The problem is literally leaking the contents of memory, yes, but the term memory leak usually means something else, something that is usually not as problematic as dumping random memory to the internet.
It's interesting that there was another recent Cloudflare bug caused by using an "==" vs "<=" or ">=".
It's good to know that programs can get into states that you don't expect.
https://blog.cloudflare.com/how-and-why-the-leap-second-affe...
I'm a Cloudflare (free) customer, and I've had no email or contact of any kind from Cloudflare.
My understanding from reading this and Travis' post is that all Cloudflare customers are affected, is that correct?
No. If you have any of those features enabled Email obfuscation, Server-side Excludes and Automatic HTTPS Rewrites as well as you have an invalid HTML markup. Per their blog post it is 0.00003% of the total requests.
I thought those were the conditions that caused PII to be included in the HTML response, but the PII could come from any customer as it's a buffer overrun situation. Damn this is confusing
Basically yes. If some of your traffic was flowing through a server on which any customer had the vulnerable features enabled, parts of your traffic may have been dumped.
I wonder if they rewrite their unsafe C code with Rust, would this bug still likely to happen?