I've only dabbled, so I'm happy to have people with more linux-side knowledge to call me out on any inaccuracies here, but...
io_uring is effectively as "secure" as any other syscall unto itself. The issue is that the mechanism by which io_uring makes its syscalls as part of its submission/completion queues means that those underlying syscalls can't be filtered by seccomp. The real question is your security posture.
If you're writing a hypervisor that's intended to partition resources between underlying users in a secure fashion, the ability for io_uring to bypass seccomp is largely a non-starter. But if you own the machine and you just want to run an application on it (i.e. an HTTP server that uses io_uring for file/network io) you should largely be in the clear.
I don't consider myself fully qualified to speak to this, so please take it with a grain of salt.
From what I gather it seems like you could potentially create scenarios where TOCTOU is indeed a problem, but in considering the situations where it could come up I do feel like all my ideas are somewhat contrived in nature. And even when noodling on it I very much get the feeling that I return to my previous statement: consider what you're building. I think that the potential for TOCTOU could potentially compromise a hypervisor's security (i.e. letting an arbitrary number of user on a system make arbitrary io_uring calls) and even if I couldn't demonstrate how that could be weaponized I would avoid it. However, if you're writing an application that's going to do a read(2) or something, I don't see TOCTOU being a uniquely io_uring problem.
Similar sentiments here. I can't find much common ground with Charlie Kirk but that doesn't merit an assassination. Unfortunate all around, and a situation not too dissimilar from the Mangione case (in the context of what happened, not necessarily why).
That said, while I don't condone it I can't say I'm surprised by it. It seems stoking divisions is a large part of the modern media landscape and all it takes is one person with the motive and the means.
The GP did not say anything about war continuing due to logical outcomes or nationally-motivated ones. Most wars simply are personal pursuits wrapped in convenient excuses.
Curious how we're defining "democracy" and "free market" with this one. I wonder how countries with a pure democracy and an actually free market compare to the republic and regulated market we have in the US.
People frequently misunderstand "constitutional democracy" as being substantially different from "republic" but that's usually an ESL error that can be fixed quickly.
The vast majority of markets in the US are hardly free. Every single large company in the US is heavily government subsidized, market protectionism is rife, and regulatory capture and artificial moat-building is the norm. I think it's quite a stretch to day we have a free market. Maybe a 'free-er' market.
> The US is a constitutional democracy with a free market and I consider it successful.
Out of all the definitions you gave, I feel you left out the most important. How exactly are you defining “successful”? Considering the current state of the US, that one seems really important.
I'm only really describing the due diligence I do to keep people safe who might rely on my OSS work. I didn't realize I was so far ahead of the defense industry...
They actually did NOT repeal the portion relevant to software development being considered research - it was actually made permanent and is pretty trivial to find by searching for "software development" against the text of the bill. What did change was an additional section to how research is capitalized, with domestic research being able to deduct immediately while other (i.e. foreign) research is subject to the amortization rules.
In all fairness, plenty of people leaving the trappings of Big Tech are pretty capable of uprooting and doing business elsewhere; California offers relatively little other than what it has from existing inertia.
Agglomeration effects are real. Research has constantly shown that people working in high productivity area generate more wealth than the same people working outside high productivity areas.
The real problem, IMO, is that workers in the tech industry have voluntarily destroyed these agglomeration benefits by using the short period of power workers had during/after the pandemic to insist on work from home instead of better worker protections.
As a result, it’s very possible that high productivity areas in the U.S. are no longer that, and remote work has reduced all Americana’s productivity, making it impossible to justify the higher cost of living and salaries in big cities.
I dunno. I wish I could find places outside California like Iowa City or St Louis appealing, a simple million will get you a mansion with like fountains, a guest house, pool, tennis courts... In SF that's a two bedroom condo
Except climate. I was on a road trip this summer and stayed in Iowa City while driving East and then St Louis driving West. Downtown Iowa City was appealing. I had a good hotel at a bargain price near the St Louis airport although the area was not the best.
But it was hot. My time back in coastal California has been mostly in the 70s. Camping in Marin County this week, the high was in the 60s. Refreshing after heat warnings in Nebraska, Illinois and Ontario.
The majority of “SF” tech works in the South Bay that has plenty of 100f days. Also almost all the condos there have thin walls and no AC. So that means 80f all night in your $3000/mo 3rd floor crap box.
I do agree there is no winter in the SF bay though. That’s pretty nice.
Agreed about jobs in the South Bay. I commuted from Midpeninsula southward for 30 years.
100 degrees is unusual although it happens. I never lived in the South Bay so can't speak about AC. My house, built in 1947, didn't have it. It was a pain, maybe 2 weeks of the year, at most. It was in the 50s, most days of the year. Low at night in the summer, high in the winter.
I had early guidance that the climate was mild and I would live in a smaller house than I was used to but I would be outside all year. Check.
I didn't understand that there are only two seasons: spring when it rained sometimes and summer when it didn't rain but might be foggy in the morning.
My first year, I craved the rain after six months without. I've adjusted although the first seasonal rain is enjoyable like the first snow in the Midwest. But I don't miss the next 6 months of Midwest winter.
Partially! The text of the OBBBA actually made permanent the section of the tax code that treats software development as research and amortizes accordingly. However, because nothing can just be straightforward, it also allows domestic research expenses to be deducted immediately instead of amortized.
Definitely a much better tax situation but also not one we would've been in if not for the TCJA, and we still have an exposed oblique (the removal of the domestic research exclusion) that could put us back in the same spot unless the software development as research section is removed.
Not going to lie, this is one of the few reasons I use LLMs at all. Even if I feel like I have a decent idea if I don't have anyone around to listen I'll just lob thoughts at an AI just to ask for alternatives, dissenting opinions, critiques, etc. Typically much of the output are things I already considered, but even that can be validating itself as a sort of reminder that I did think things through. And on some occasions it does raise things I wouldn't have considered which can be great to stop and chew on before proceeding.
How do you reconcile the fact that LLMs can be pretty fair-weather? Meaning, while they can serve as a sounding board and often raise perspectives you might not have thought of, they don't have much conviction and will change their tune if you push them in the other direction enough.
My approach is to compliment the LLM on something I've not thought of and ask it to sell me on the approach, expound on its position, and ask probing questions. If I get a feeling something's off I just go do independent research like normal.
io_uring is effectively as "secure" as any other syscall unto itself. The issue is that the mechanism by which io_uring makes its syscalls as part of its submission/completion queues means that those underlying syscalls can't be filtered by seccomp. The real question is your security posture.
If you're writing a hypervisor that's intended to partition resources between underlying users in a secure fashion, the ability for io_uring to bypass seccomp is largely a non-starter. But if you own the machine and you just want to run an application on it (i.e. an HTTP server that uses io_uring for file/network io) you should largely be in the clear.