{"id":122993,"date":"2025-03-25T22:22:41","date_gmt":"2025-03-25T20:22:41","guid":{"rendered":"https:\/\/prohoster.info\/blog\/novosti-interneta\/reliz-yadra-linux-6-14"},"modified":"2025-03-25T22:22:41","modified_gmt":"2025-03-25T20:22:41","slug":"reliz-yadra-linux-6-14","status":"publish","type":"post","link":"https:\/\/prohoster.info\/sq\/blog\/news\/reliz-yadra-linux-6-14","title":{"rendered":"L\u00ebshimi i b\u00ebrtham\u00ebs Linux 6.14","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"
Pas dy muajsh zhvillimi, Linus Torvalds prezantoi l\u00ebshimin e b\u00ebrtham\u00ebs Linux 6.14. Nd\u00ebr ndryshimet m\u00eb t\u00eb dukshme jan\u00eb: drejtuesi ntsync me primitiv\u00eb t\u00eb sinkronizimit Windows NT, rregullimi i balancimit t\u00eb operacioneve t\u00eb leximit n\u00eb Btrfs RAID1, mb\u00ebshtetje p\u00ebr reflink n\u00eb XFS n\u00eb m\u00ebnyr\u00eb reale, mund\u00ebsia e hyrjes\/jeshjes buffer pa cache, dmem cgroup p\u00ebr limitimin e memories GPU, angazhimi i io_uring n\u00eb FUSE, delegimi i atributeve n\u00eb NFS, mb\u00ebshtetje p\u00ebr regjistrim t\u00eb atomike n\u00eb Device mapper, p\u00ebrshpejtimi i lidhjeve simbolike, menaxhimi i mund\u00ebsis\u00eb p\u00ebr ekzekutimin e skriptive, mb\u00ebshtetje p\u00ebr \u00e7ipat Qualcomm Snapdragon 8 Elite, dhe drejtuesi p\u00ebr NPU AMD. <\/p>\n
Shtesa e nj\u00eb m\u00ebnyre t\u00eb pa-cache p\u00ebr futje\/dalje, ku t\u00eb dh\u00ebnat hiqen nga kujtesa e faqes menj\u00ebher\u00eb pasi t\u00eb p\u00ebrfundojn\u00eb operacionet p\u00ebr leximin ose shkrimin e tyre. Ky ndryshim mund t\u00eb jet\u00eb i dobish\u00ebm kur p\u00ebrdoren pajisje ruajtjeje shum\u00eb t\u00eb shpejta, ku caching i operacioneve n\u00eb kujtes\u00eb \u00ebsht\u00eb i tep\u00ebrt. P\u00ebr k\u00ebto pajisje, m\u00ebnyra e re lejon p\u00ebrjashtimin e shpenzimeve t\u00eb tepruara t\u00eb kujtes\u00ebs nga cache i faqes pa p\u00ebrdorur nj\u00eb API t\u00eb komplikueshme p\u00ebr futje\/dalje t\u00eb drejtp\u00ebrdrejt\u00eb (Direct I\/O).\n
N\u00eb n\u00ebn-sistemin FUSE, q\u00eb lejon krijimin e implementimeve t\u00eb sistemeve t\u00eb skedar\u00ebve n\u00eb hap\u00ebsir\u00ebn e p\u00ebrdoruesit, \u00ebsht\u00eb shtuar mb\u00ebshtetje p\u00ebr shk\u00ebmbimin e t\u00eb dh\u00ebnave midis b\u00ebrtham\u00ebs dhe trajtuesit n\u00eb hap\u00ebsir\u00ebn e p\u00ebrdoruesit duke p\u00ebrdorur mekanizmin e futjes\/daljes io_uring. Ky ndryshim lejon p\u00ebrmir\u00ebsimin e performanc\u00ebs s\u00eb FUSE duke reduktuar kalimet e kontekstit midis b\u00ebrtham\u00ebs dhe hap\u00ebsir\u00ebs s\u00eb p\u00ebrdoruesit.\n
N\u00eb VFS \u00ebsht\u00eb implementuar nj\u00eb cache p\u00ebr madh\u00ebsin\u00eb e lidhjeve simbolike, e cila ka lejuar nj\u00eb rritje prej 1.5% n\u00eb shpejt\u00ebsin\u00eb e ekzekutimit t\u00eb operacionit readlink (n\u00eb testin me \/initrd.img n\u00eb ext4). Cache \u00ebsht\u00eb aktivizuar n\u00eb FS ext4 dhe tmpfs.\n
P\u00ebrmir\u00ebsuar performanca e operacioneve t\u00eb leximit n\u00eb sistemet e skedar\u00ebve NETFS, CIFS dhe AFS (Andrew File System).\n
N\u00eb Squashfs \u00ebsht\u00eb p\u00ebrfshir\u00eb nj\u00eb mod i ngarkimit direkt t\u00eb bllokimeve n\u00eb cached let\u00ebr (SQUASHFS_FILE_DIRECT), i cili lejon t\u00eb anashkalohet nj\u00eb cached ndar\u00ebs read_page. Ky ndryshim ka lejuar t\u00eb reduktohet madh\u00ebsia e memories s\u00eb p\u00ebrdorur gjat\u00eb pun\u00ebs me Squashfs.\n
N\u00eb thirrjen e sistemit statx() \u00ebsht\u00eb implementuar flaga STATX_DIO_READ_ALIGN, p\u00ebr t\u00eb p\u00ebrcaktuar k\u00ebrkes\u00ebn p\u00ebr rregullimin e operacioneve t\u00eb leximit nga skedari.\n
N\u00eb sistemin e skedar\u00ebve Bcachefs \u00ebsht\u00eb p\u00ebrmir\u00ebsuar dhe stabilizuar formati i strukturave t\u00eb diskut. \u00c7do ndryshim i m\u00ebtejsh\u00ebm i formatit do t\u00eb konsiderohet si opsional dhe do t\u00eb implementohet n\u00eb form\u00ebn e shtesave opsionale. Shpejt\u00ebsia e verifikimit t\u00eb integritetit t\u00eb FS \u00ebsht\u00eb rritur ndjesh\u00ebm. P\u00ebrve\u00e7 k\u00ebsaj, n\u00eb Bcachefs \u00ebsht\u00eb p\u00ebrmir\u00ebsuar funksionimi n\u00eb modin vet\u00ebm p\u00ebr lexim; jan\u00eb eliminuar problemet q\u00eb \u00e7ojn\u00eb n\u00eb qasje n\u00eb memorie pas lirimit t\u00eb saj (use after free); jan\u00eb zgjidhur problemet me referencat reflink n\u00eb fsck; \u00ebsht\u00eb rregulluar trajtimi i rinisjes s\u00eb transaksionit.\n
Moduli md-linear \u00ebsht\u00eb rikthyer, i destinuar p\u00ebr t\u00eb bashkuar pajisjet e bllokut. Ky modul ishte shpallur m\u00eb par\u00eb si i vjetruar dhe ishte fshir\u00eb nga b\u00ebrthama 6.8, por si\u00e7 duket, ishte k\u00ebrkuar dhe prandaj tani \u00ebsht\u00eb rikthyer.\n
Sistemet e skedar\u00ebve F2FS dhe SQUASHFS jan\u00eb kaluar n\u00eb p\u00ebrdorimin e flet\u00ebve t\u00eb faqes s\u00eb memories (page folios).\n
Sistemet e skedar\u00ebve OCFS2 dhe DLMFS jan\u00eb kaluar n\u00eb p\u00ebrdorimin e API-t t\u00eb ri p\u00ebr montimin e pjes\u00ebve.\n
N\u00eb sistemin e mapimit t\u00eb pajisjeve dhe n\u00eb modulet dm-mirror, dm-io, dm-table, dm-linear, dm-stripe dhe dm-raid1 \u00ebsht\u00eb implementuar mb\u00ebshtetje p\u00ebr shkruarjen atomike. <\/ul>\n
Memorie dhe sh\u00ebrbime sistemike<\/b>\n
\n
Integrimi n\u00eb thelbin e driver-it ntsync \u00ebsht\u00eb p\u00ebrfunduar, i cili implementon pajisjen simbolike \/dev\/ntsync dhe nj\u00eb grup primitivash p\u00ebr sinkronizimin e p\u00ebrdorur n\u00eb thelbin e Windows NT. Realizimi i k\u00ebtyre primitiveve n\u00eb nivelin e thelbit lejon nj\u00eb rritje t\u00eb dukshme t\u00eb performanc\u00ebs s\u00eb loj\u00ebrave Windows q\u00eb ekzekutohen me ndihm\u00ebn e Wine. Rritja e performanc\u00ebs arrihet duke eliminuar shpenzimet e lidhura me p\u00ebrdorimin e RPC n\u00eb hap\u00ebsir\u00ebn e p\u00ebrdoruesit. Krijimi i nj\u00eb driver-i t\u00eb ve\u00e7ant\u00eb p\u00ebr thelbin e Linux shpjegohet nga v\u00ebshtir\u00ebsia e implementimit t\u00eb sakt\u00eb t\u00eb API-s\u00eb s\u00eb sinkronizimit NT mbi primitive ekzistuese n\u00eb thelb.\n
Shtuar nj\u00eb kontrolues t\u00eb ri cgroup DMEM p\u00ebr llogaritjen e ndar\u00eb t\u00eb zonave t\u00eb memories s\u00eb pajisjeve, si\u00e7 \u00ebsht\u00eb GPU. DMEM lejon krijimin e cgroup-eve t\u00eb ve\u00e7anta p\u00ebr detyra t\u00eb ndryshme q\u00eb punojn\u00eb me GPU, n\u00eb m\u00ebnyr\u00eb q\u00eb ato t\u00eb mund t\u00eb ekzekutohen pa ndikim mbi nj\u00ebra-tjetr\u00ebn. Kjo mund\u00ebsi e re zgjidh problemin e p\u00ebrfundimit t\u00eb detyrave me GPU kur memorja e disponueshme p\u00ebrfundon, duke llogaritur ndihm\u00ebn e reflektuar t\u00eb memories s\u00eb GPU dhe memories s\u00eb p\u00ebrdorur nga driver-at e CPU n\u00eb cgroup-e t\u00eb ve\u00e7anta.\n
Optimizimet jan\u00eb b\u00ebr\u00eb n\u00eb skalimin e operacionit t\u00eb pastrimit (flush) t\u00eb caches TLB (Translation Lookaside Buffer), q\u00eb p\u00ebrdoret p\u00ebr t\u00eb p\u00ebrshpejtuar konvertimin e adresave virtuale n\u00eb fizike. Optimizimet e reja p\u00ebrfshijn\u00eb p\u00ebrdit\u00ebsimin e vonuar t\u00eb disa strukturave t\u00eb dh\u00ebnash gjat\u00eb kalimit t\u00eb kontekstit, duke lejuar p\u00ebrmir\u00ebsimin e performanc\u00ebs gjat\u00eb kalimit t\u00eb disa provave.\n
Performanca e mekanizmit MGLRU (Multi-Generational LRU) \u00ebsht\u00eb rritur, i cili p\u00ebrdoret p\u00ebr t\u00eb identifikuar cilat faqe memorije p\u00ebrdoren dhe cilat mund t\u00eb shkarkohen n\u00eb hap\u00ebsir\u00ebn e shkarkimit.\n
N\u00eb skenar\u00ebt e nd\u00ebrtimit \u00ebsht\u00eb propozuar nj\u00eb kod i ri p\u00ebr gjenerimin e versioneve t\u00eb simboleve t\u00eb debugimit p\u00ebr modul\u00ebt e ngarkuesh\u00ebm, i cili tani p\u00ebrdor informacion nga regjistrat e debugimit n\u00eb formatin DWARF, n\u00eb vend q\u00eb t\u00eb analizoj\u00eb kodin burimor drejtp\u00ebrdrejt. Ky ndryshim lejon aplikimin e versionimit t\u00eb simboleve t\u00eb debugimit p\u00ebr modul\u00ebt e shkruar n\u00eb gjuh\u00ebn Rust. Implementimi i vjet\u00ebr gjithashtu \u00ebsht\u00eb l\u00ebn\u00eb n\u00eb b\u00ebrtham\u00eb, dhe zgjedhja e gjeneratorit b\u00ebhet n\u00eb nivelin e mund\u00ebsive t\u00eb nd\u00ebrtimit.\n
\u00cbsht\u00eb ofruar mund\u00ebsia e p\u00ebrdorimit t\u00eb parametrave sysctl pid_max me hap\u00ebsira identitetesh t\u00eb procesit (ID namespace). Parametri pid_max \u00ebsht\u00eb i destinuar p\u00ebr t\u00eb kufizuar vler\u00ebn maksimale t\u00eb identifikator\u00ebve t\u00eb proceseve (PID) dhe tani mund t\u00eb p\u00ebrdoret p\u00ebr t\u00eb kufizuar numrin maksimal t\u00eb proceseve t\u00eb ekzekutuara n\u00eb hap\u00ebsir\u00ebn e caktuar identitetesh. Trajtimi i parametrave kryhet n\u00eb m\u00ebnyr\u00eb hierarkike, dmth. kufizimet n\u00eb hap\u00ebsirat e jashtme identitetesh zgjaten n\u00eb hap\u00ebsirat e brendshme identitetesh.\n
Kur p\u00ebrdoret p\u00ebr t\u00eb nisur procesin e thirrjes sistemike execveat n\u00eb sistemin e skedar\u00ebve \/proc, tani do t\u00eb shfaqet emri i skedarit t\u00eb nisur, jo numri i descriptor-it t\u00eb skedarit.\n
N\u00eb kodin burimor t\u00eb b\u00ebrtham\u00ebs (n\u00eb katalogun samples\/vfs) \u00ebsht\u00eb shtuar utilitarja mountinfo, e cila demostrim p\u00ebrdorimin e thirrjeve sistemike statmount() dhe listmount().\n
N\u00eb n\u00ebn-sistemin BPF jan\u00eb propozuar funksionet e reja bpf_local_irq_save() dhe bpf_local_irq_restore() p\u00ebr t\u00eb \u00e7aktivizuar p\u00ebrkoh\u00ebsisht nd\u00ebrprerjet n\u00eb CPU-n\u00eb lokale. Funksionet mund t\u00eb p\u00ebrdoren p\u00ebr t\u00eb implementuar struktura, trajtimi i t\u00eb cilave nuk pezullohet nga nd\u00ebrprerjet.\n
N\u00eb thirrjen sistemike madvise() me p\u00ebrdorimin e flagjeve MADV_DONTNEED dhe MADV_FREE, \u00ebsht\u00eb siguruar \u00e7lirimi i tabelave t\u00eb faqeve t\u00eb memories q\u00eb lidhen me intervalin e adresave q\u00eb po \u00e7lirohen, pasi n\u00eb disa situata faqet e zbraz\u00ebta t\u00eb memories mund t\u00eb z\u00ebn\u00eb nj\u00eb sasi t\u00eb konsiderueshme memorjeje.\n
P\u00ebr arkitektur\u00ebn OpenRISC \u00ebsht\u00eb implementuar mb\u00ebshtetje p\u00ebr mekanizmin e sekuencave t\u00eb ri-nisura (rseq, restartable sequences), i cili \u00ebsht\u00eb i destinuar p\u00ebr realizimin e shpejt\u00eb t\u00eb operacioneve atomike, t\u00eb cilat n\u00eb rastin e nj\u00eb nd\u00ebrprerjeje nga nj\u00eb proces tjet\u00ebr pastrohen dhe p\u00ebrpiqen s\u00ebrish p\u00ebr t\u00eb u ekzekutuar.\n
\u00cbsht\u00eb b\u00ebr\u00eb nj\u00eb reorganizim i kodit me implementimin e algoritmave CRC32 dhe CRC-T10DIF, i cili nuk p\u00ebrplaset m\u00eb me n\u00ebn-sistemin crypto dhe thirret drejtp\u00ebrdrejt nga interfaci i bibliotek\u00ebs. Ndryshimi ka mund\u00ebsuar thjeshtimin e kodit dhe rritjen e efikasitetit t\u00eb tij.\n
N\u00eb sistemin e hyrjes\/shtypjes asinkrone io_uring \u00ebsht\u00eb shtuar nj\u00eb nd\u00ebrfaqe p\u00ebr t\u00eb transmetuar metadatat shtes\u00eb mbi integritetin gjat\u00eb ekzekutimit t\u00eb operacioneve t\u00eb leximit dhe shkruarjes. <\/ul>\n
N\u00eb sistemet x86 \u00ebsht\u00eb realizuar mb\u00ebshtetje p\u00ebr numra t\u00eb mbrojtur t\u00eb koh\u00ebs p\u00ebr sistemet mysafir, t\u00eb cilat nuk lejojn\u00eb t\u00eb ndryshojn\u00eb oraret sistemore t\u00eb sistemit mysafir nga ambienti prit\u00ebs. Ky mund\u00ebsi \u00ebsht\u00eb realizuar mbi baz\u00ebn e mekanizmit AMD SEV (Secure Encrypted Virtualization), i p\u00ebrdorur n\u00eb sistemet e virtualizimit p\u00ebr mbrojtje makinat virtuale<\/a> nga nd\u00ebrhyrja e hipervizorit ose administratori t\u00eb sistemit prit\u00ebs.\n