ripatel

48 posts

ripatel banner
ripatel

ripatel

@fd_ripatel

Katılım Haziran 2024
90 Takip Edilen222 Takipçiler
Visegrád 24
Visegrád 24@visegrad24·
🇸🇾🇩🇪 Syria has rejected Germany’s plan to return over 700,000 Syrian refugees to their home country. "We categorically reject any attempts at forced deportation,” said Syrian Foreign Minister Asaad al-Shaibani. He added that Syrians in Germany are a “strategic resource” rather than a burden.
Visegrád 24 tweet media
English
3K
2.5K
6.2K
752.8K
miles
miles@wedtm·
Bro, this is how it works. You don’t get the massive payout that black hats do, but you’re also not looking over your shoulder every night or wearing an orange jumpsuit. Not to mention that the $800m isn’t *theirs*, what money are they going to pay the $250k bounty with? At what point does it flip from a disclosure to extortion?
English
22
2
268
65.6K
Basil the Great
Basil the Great@BasilTheGreat·
🚨NEWS: A Muslim man who tried to rape a 14 year old girl in Manchester named Mohammed Ayaz, originally from Pakistan, has been found dead just before he was due to be sentenced for his crime He was caught by a child protection trap group
Basil the Great tweet media
English
11.3K
11.8K
86.8K
4.4M
Martin Sellner
Martin Sellner@MartinSellner_·
Ordnungsruf für Remigration Die irren Altparteien wollen verbieten, dass im Nationalrat der Begriff Remigration verwendet wird. Die Hälfte der Österreicher will umfassende Remigration. Die Mehrheit will eine Remigrationspartei wählen. Und im Parlament soll man den Begriff nicht erwähnen dürfen...
Martin Sellner tweet media
Deutsch
14
71
700
10.2K
ripatel
ripatel@fd_ripatel·
@readylayerone @alessandrod Frankendancer has been running XDP for 2 years Agave improved replay quite a lot by copying a bunch of other FD features, like speculative execution (a bit incorrectly hence 4.0 patch releases on mainnet) They have not yet caught up to full FD's performance but still good work
English
0
0
3
120
Alessandro Decina
Alessandro Decina@alessandrod·
I've been critical of them in the past, but wow, this is the biggest frankendancer perf improvement to date. 4 times faster! credit where credit is due, amazing work team! (this is them rebasing on top of agave 4.0 and picking up our perf work for free lol)
Alessandro Decina tweet media
English
8
5
106
4K
ripatel
ripatel@fd_ripatel·
@deanmlittle IBT has a 0x3e notrack prefix that disables enforcement for individual jumps. i.e. notrack jmp rax Weakens the sandbox but at least the rest of the app gets function pointer integrity still.
English
1
0
0
90
Dean 利迪恩 (⚛️,🐱) | sbpf/acc
@fd_ripatel static offsets don’t have this problem in the first place. how would you do this with callx when the call target is indirected and could be supplied completely dynamically? that makes *every* instruction a valid call target.
English
1
0
0
102
ripatel
ripatel@fd_ripatel·
Linux 7.0 still does not support Intel CET indirect branch tracking (IBT) in usermode. Intel CPUs supporting IBT first appeared in 2021. Here's my patch series: @wii.dev/T/" target="_blank" rel="nofollow noopener">lore.kernel.org/lkml/202605171…
ripatel tweet media
English
2
12
64
21K
ripatel
ripatel@fd_ripatel·
@deanmlittle The Agave JIT already tracks valid call targets and enforces them. They just need to emit endbr64 instructions.
English
1
0
0
118
Dean 利迪恩 (⚛️,🐱) | sbpf/acc
@fd_ripatel i was researching this recently to see if we could use it to kill jit spraying attacks without all the expensive sanitization we currently do but it seems you'd need to either put it before each instruction boundary or somehow know all possible valid callx targets aot. 😢
English
1
0
3
320
ripatel
ripatel@fd_ripatel·
@canadianbryan very cool, hopefully some motivation for Linux to accept my patches 🙂
English
1
0
2
60
Bryan Steele 🦋
Bryan Steele 🦋@canadianbryan·
OpenBSD has had IBT support since the latter part of 2023, for Tiger Lake+ (Intel 11th gen and newer). openbsd.org/74.html It's enabled by default in both the kernel and usermode (including ports, w/ less than ~100 marked USE_NOBTCFI, some aarch64 BTI only).
ripatel@fd_ripatel

Linux 7.0 still does not support Intel CET indirect branch tracking (IBT) in usermode. Intel CPUs supporting IBT first appeared in 2021. Here's my patch series: @wii.dev/T/" target="_blank" rel="nofollow noopener">lore.kernel.org/lkml/202605171…

English
3
3
13
1K
ripatel
ripatel@fd_ripatel·
This came up while adding control flow integrity to Firedancer's sandbox. One of the best Linux x86 jails out there, enforcing process isolation, namespaces, strict seccomp, landlock, MPK (user protection keys), CFI, and soon more. github.com/firedancer-io/…
English
0
2
8
768
ripatel
ripatel@fd_ripatel·
IBT enforces that indirect calls and jumps can't land on any arbitrary instruction, but one of these: [f3 0f 1e fa] endbr64 "jmp rax" to elsewhere causes a segfault SEGV_CPERR, defending against control flow hijacking (e.g. corrupt function pointer).
English
1
1
10
2.1K
frank
frank@soundsonacid·
the day is finally here
frank tweet media
English
3
2
21
1.3K
ripatel
ripatel@fd_ripatel·
@toly @SolanaSensei @solana @SolanaFndn @bw_solana 😴😴😴 Why do let the majority client pick protocol limits instead of giving validators a config option? Frankendancer has supported XDP for 2 years, so we're waiting on nothing. Just switch back to Agave once they sort out their performance issues
English
2
0
2
240
Solana Sensei
Solana Sensei@SolanaSensei·
Solana accounted for 46% of all crypto transactions in March 2026. 3.73B out of 8.08B total. OPOS
Solana Sensei tweet media
English
80
21
253
16.4K
dr cavey phd ∿
dr cavey phd ∿@cavemanloverboy·
Big news from the SOLANA world — Sinon over at JUMP FIREDANCER just built the FASTEST Falcon implementation EVER, totally from scratch, using AVX512 (very advanced, very powerful stuff, most people have no idea). And get this — they’re giving it AWAY, upstreaming straight to the author's repo. That’s called CLASS, folks. Expect MANY more incredible open source contributions. Tremendous work. The best!​​​​​​​​​​​​​​​​ Thank you for your attention to this matter!
ripatel@fd_ripatel

@JohnAlanWoods Solana (Sinon from @jump_firedancer) wrote *the* fastest Falcon implementation from scratch, using AVX512. The author is upstreaming their contributions: github.com/algorand/falco… Expect more good OSS contributions from them 🙂

English
4
11
86
16.9K
cts🌸
cts🌸@gf_256·
Fun fact: there's an easter egg that unlocks 1 free month of twitter premium Quickly press L R X Cmd+⏎ only works on desktop website
English
11
2
118
20.9K
ripatel
ripatel@fd_ripatel·
@RaphaelBrinkert @ergroovt Ja natürlich ist das Absahnen von Steuergeldern sehr profitabel für die diebischen Freunde der regierenden Parteien.
Deutsch
0
0
1
63
Rapha
Rapha@RaphaelBrinkert·
@ergroovt Das meiste sind Deals, wovon zugleich die deutschen Unternehmen profitieren. #Servicetweet
Deutsch
7
0
21
3.3K
DerB
DerB@ergroovt·
Deutschland zahlte 2025 26 Milliarden € Entwicklungshilfe in alle Welt. Hat jemand eine Idee, was man für die deutschen Pflegekassen machen könnte?
DerB tweet media
DER SPIEGEL@derspiegel

In der Pflege fehlen laut Gesundheitsministerin Warken sechs Milliarden Euro im kommenden Jahr. Erste Sparideen klingen entsprechend hart und sollen unter anderem die Kosten bei der Heimunterbringung betreffen. #ref=rss" target="_blank" rel="nofollow noopener">spiegel.de/wirtschaft/ref…

Deutsch
138
1.7K
5.5K
176.5K
ripatel
ripatel@fd_ripatel·
@WiMJapan Es ist nie zu spät zu kämpfen solange dein Herz noch schlägt
Deutsch
0
0
10
222
Isaac
Isaac@isaacrrr7·
Los Europeos no hacen hijos porque no pueden mantenerlos. Los Musulmanes hacen hijos porque se los mantienen los Europeos.
Español
589
19.1K
98.5K
1.3M
ripatel
ripatel@fd_ripatel·
@cavemanloverboy Bursting PoH at the start of the block was a bug but it improves performance, so I guess it can stay ... until PoH is finally gone
English
0
0
1
48
dr cavey phd ∿
dr cavey phd ∿@cavemanloverboy·
the funny part about tick distribution stuff is that the most IBRL thing to do is to ignore poh pacing entirely, get all of your tick shreds out as soon as possible to the network, and then just record your txns whenever they come through in the last few ticks (but throughout your slot in wall time) this is the default behavior in full fd. it hashes ASAP until the last two ticks, and then reserves the rest of the space for mixing in records until the pack tile says it's done. to do otherwise is to withhold tick shreds. recall that other nodes have to run poh verify during replay, so withholding them means they must do this 10-50ms of work later into the slot. ibrl.wtf/validator/DNVZ…
English
4
1
14
1.4K