Christoph Hellwig, maintainer of the DMA subsystem, KVM, Slab Allocator, and PowerPC architecture in the Linux kernel, who at one time was a member of the technical committee of the Linux Foundation and was a plaintiff in a legal dispute with VMware regarding the GPL, refused to accept patches related to supporting driver development in the Rust language. The proposed patches added wrappers over several functions of the DMA subsystem, allowing DMA to be used in drivers written in Rust. <\/p>\n
The reason for the refusal mentioned was the complexity of maintaining code when wrappers are present in other languages and the desire to keep the programming interfaces to DMA in a readable form in C, without being obscured by unclear wrappers. Christoph suggested directly calling the original C API of DMA in each Rust driver, to avoid creating additional abstractions that kernel maintainers would have to depend on. <\/p>\n
V\u00fdvoj\u00e1\u0159i patch\u016f uvedli, \u017ee si vezmou ve\u0161kerou pr\u00e1ci na \u00fadr\u017eb\u011b k\u00f3du v Rustu na sv\u00e1 bedra, jsou p\u0159ipraveni tyto patchy spravovat sami a p\u0159esunuli obalovac\u00ed k\u00f3d do samostatn\u00e9ho podadres\u00e1\u0159e (rust\/kernel\/dma.rs). V reakci na to Christoph uvalil veto («Nacked-by») na p\u0159ijet\u00ed patch\u016f souvisej\u00edc\u00edch s Rustem a uvedl, \u017ee nepot\u0159ebuje dal\u0161\u00edho spr\u00e1vce. Christoph prohl\u00e1sil, \u017ee pokud si v\u00fdvoj\u00e1\u0159i obal\u016f p\u0159ej\u00ed dos\u00e1hnout nemo\u017enosti \u00fadr\u017eby Linuxu kv\u016fli m\u00edch\u00e1n\u00ed n\u011bkolika jazyk\u016f v jednom k\u00f3dov\u00e9m z\u00e1kladu, m\u011bli by to d\u011blat ve sv\u00e9m ovlada\u010di a ne\u0161\u00ed\u0159it tuto rakovinu do hlavn\u00edch subsyst\u00e9m\u016f j\u00e1dra. <\/p>\n
At the same time, Christoph clarified that he has nothing against the Rust language and considers it one of the best new languages, but he is against mixing code in different languages. According to Christoph, he supports the creation of new projects in Rust, but is against mixing Rust with large codebases in C, as such mixing significantly reduces the maintainability of the kernel as an integrated project. <\/p>\n
Podstata probl\u00e9m\u016f se spr\u00e1vou spo\u010d\u00edv\u00e1 v tom, \u017ee Rust wrappery \u010din\u00ed spr\u00e1vce z\u00e1visl\u00e9 na k\u00f3du napsan\u00e9m v jazyce Rust. Na prvn\u00ed pohled se zd\u00e1, \u017ee wrappery jsou pouze nadstavbami nad C strukturami a funkcemi, kter\u00e9 nijak neovliv\u0148uj\u00ed v\u00fdvoj a \u00fadr\u017ebu k\u00f3du v C. Ale to nen\u00ed pravda. V p\u0159\u00edpad\u011b p\u0159\u00edtomnosti takov\u00fdch wrapper\u016f mus\u00ed v\u00fdvoj\u00e1\u0159i subsyst\u00e9m\u016f napsan\u00fdch v C br\u00e1t v \u00favahu vliv sv\u00fdch zm\u011bn na pokra\u010dov\u00e1n\u00ed funk\u010dnosti wrapper\u016f. Jak\u00e1koli zm\u011bna datov\u00fdch struktur nebo vnit\u0159n\u00edch funkc\u00ed v C m\u016f\u017ee v\u00e9st k nutnosti zm\u011bny k\u00f3du wrapper\u016f, a proto je t\u0159eba sledovat a synchronizovat zm\u011bny v C k\u00f3du, kter\u00e9 ovliv\u0148uj\u00ed wrappery, s k\u00f3dem v Rustu. Mnoho spr\u00e1vc\u016f nen\u00ed p\u0159ipraveno vz\u00edt si na sebe dodate\u010dnou odpov\u011bdnost za opravu probl\u00e9m\u016f, kter\u00e9 vznikaj\u00ed v k\u00f3du v Rustu, a nehodl\u00e1 tr\u00e1vit sv\u016fj \u010das sledov\u00e1n\u00edm stavu Rust wrapper\u016f. <\/p>\n
Situace se zhor\u0161en\u00edm \u00fadr\u017eby nen\u00ed teoretick\u00e1. Diskuze se z\u00fa\u010dastnil Jason Gunthorpe, udr\u017eovatel TPM, VFIO a Infiniband z firmy NVIDIA, kter\u00fd p\u0159ivedl p\u0159\u00edklad, kdy Linus Torvalds odm\u00edtl pull request se zm\u011bnami v subsyst\u00e9mu spr\u00e1vy pam\u011bti, proto\u017ee tato zm\u011bna vedla k chyb\u011b p\u0159i pokusu o sestaven\u00ed j\u00e1dra se zapnutou podporou Rust. Chyba nastala, proto\u017ee spr\u00e1vci k\u00f3du v Rustu nep\u0159idali pot\u0159ebn\u00e9 zm\u011bny do gener\u00e1toru wrapper\u016f (bindgen). Tak se spr\u00e1vcov\u00e9 subsyst\u00e9mu spr\u00e1vy pam\u011bti stali z\u00e1visl\u00fdmi na voliteln\u00e9m ciz\u00edm k\u00f3du v j\u00e1d\u0159e, za kter\u00fd odpov\u00eddaj\u00ed jin\u00ed lid\u00e9, co\u017e m\u011blo za n\u00e1sledek dokonce \u00faplnou z\u00e1vislost na k\u00f3du, kter\u00fd byl z pohledu C k\u00f3du a j\u00e1dra jako celku zcela spr\u00e1vn\u00fd. <\/p>\n
Odm\u00edtnut\u00ed p\u0159ijmout k\u00f3d obal\u016f nad vol\u00e1n\u00edmi DMA dostalo v\u00fdvoj\u00e1\u0159e projektu Rust for Linux do slep\u00e9 uli\u010dky, proto\u017ee bez takov\u00fdch obal\u016f bude obt\u00ed\u017en\u00e9 vyv\u00edjet plnohodnotn\u00e9 ovlada\u010de v jazyce Rust. Hector Martin, maintainer k\u00f3du pro podporu ARM \u010dip\u016f Apple a vedouc\u00ed projektu Ashai Linux, navrhl jako alternativu vy\u0159e\u0161en\u00ed konfliktu dos\u00e1hnout p\u0159ijet\u00ed obal\u016f p\u0159\u00edmo p\u0159es Linuse Torvaldse, obch\u00e1zej\u00edc\u00edho spr\u00e1vce subsyst\u00e9mu DMA. Pokud by Linus souhlasil s t\u00edmto poru\u0161en\u00edm hierarchie a zaveden\u00e9 praxe, mohlo by to v\u00e9st k krizi \u0159\u00edzen\u00ed v\u00fdvoje j\u00e1dra, a pokud by odm\u00edtl \u2014 zastav\u00ed prosp\u011bch Rustu v j\u00e1d\u0159e. <\/p>\n
Jako mo\u017enost zm\u00ednil Hector p\u0159ita\u017een\u00ed Christophy k odpov\u011bdnosti za poru\u0161en\u00ed kodexu chov\u00e1n\u00ed kv\u016fli koment\u00e1\u0159i, ve kter\u00e9m Christoph p\u0159irovnal Rust k rakovinov\u00e9mu n\u00e1doru. D\u00e1le Hector napsal, \u017ee je unaven\u00fd ze v\u0161ech byrokratick\u00fdch zdr\u017een\u00ed, nen\u00ed p\u0159ipraven jednodu\u0161e d\u016fv\u011b\u0159ovat zaveden\u00fdm proces\u016fm a nazna\u010dil zapojen\u00ed soci\u00e1ln\u00edch m\u00e9di\u00ed. Dave Airlie, udr\u017eovatel subsyst\u00e9mu DRM, poradil, aby se konflikt nep\u0159ece\u0148oval a pochopil, \u017ee toxick\u00e9 chov\u00e1n\u00ed je nep\u0159ijateln\u00e9 na obou stran\u00e1ch, bez ohledu na to, zda je \u00fa\u010dastn\u00edk diskuse v pr\u00e1vu nebo ne. <\/p>\n
Diskuze se z\u00fa\u010dastnil Linus Torvalds, kter\u00fd uvedl, \u017ee probl\u00e9m mo\u017en\u00e1 spoil\u00ed s\u00e1m Hector svou nadm\u011brnou sebed\u016fv\u011brou, \u017ee v\u00ed n\u011bco l\u00e9pe ne\u017e ostatn\u00ed, a nikoli v aktu\u00e1ln\u00edm procesu v\u00fdvoje j\u00e1dra, kter\u00fd funguje. Tento proces m\u00e1 sv\u00e9 probl\u00e9my, ale to je jednodu\u0161e \u017eivotn\u00ed realita \u2014 v \u017eivot\u011b nen\u00ed nic dokonal\u00e9. Pokusy o \u0161ikanu prost\u0159ednictv\u00edm soci\u00e1ln\u00edch s\u00edt\u00ed jsou jedn\u00edm z d\u016fvod\u016f, pro\u010d Linus ztratil z\u00e1jem m\u00edt jak\u00fdkoli vztah k p\u0159\u00edstupu Hectora. Pro Linuse maj\u00ed v\u00fdznam pouze technick\u00e9 diskuse a patche, nikoli vyv\u00edjen\u00ed tlaku prost\u0159ednictv\u00edm soci\u00e1ln\u00edch s\u00edt\u00ed. <\/p>\n