linux/Documentation/highuid.txt
<<
val.22/spa v .22/form v .22a val.2 href="../linux+v3.8.1/Documenta > /highuid.txt"> val.22img src="../.sta >c/gfx/right.png" alt=">>"> v2/spa v v2spa class="lxr_search"> val ="+search" method="post" onsubmit="return do_search(this);"> val.22input typ hidden" nam navtarget" ion> "> val.22input typ text" nam search" id search"> val.22buttopttyp submit">Search val.2Prefsv .22/a> v2/spa val.2 22/div val.2 22form ac > ="ajax+*" method="post" onsubmit="return false;"> v2input typ hidden" nam ajax_lookup" id ajax_lookup" ion> "> al.2 22/form v al.2 22div class="headingbottom">
2 212/a>Notes optthe change from 16-bit UIDs to 32-bit UIDs: 2 222/a> 2 232/a>- kernel code MUST take into account __kernel_uid_t and __kernel_uid32_t 2 242/a> whe communica >ng betwee user and kernel space in an ioctl or data 2 252/a> structure. 2 262/a> 2 272/a>- kernel code should use uid_t and gid_t in kernel-private structures and 2 282/a> code. 2 292/a> 2 > a>What's left to be done for 32-bit UIDs optall Linux architectures: 2 112/a> 2 122/a>- Disk quotas have an interes >ng limita > tthat is not related to the 2 132/a> maximum UID/GID. They are limited bytthe maximum file size optthe 2 142/a> underly>ng filesystem, because quota records are written at offsets 2 152/a> corresponding to the UID in ques >on. 2 162/a> Further inves >ga > tis needed to see if the quota system ca cope 2 172/a> properly with huge UIDs. If it ca deal with 64-bit file offsets optall 2 182/a> architectures, this should not be a problem. 2 192/a> 2 202/a>- Decide whether or not to keep backwards compa >bility with the system 2 212/a> account>ng file, or if we should break it as the comments sugges 2 222/a> (currently, the old 16-bit UID and GID are still written to disk, and 2 232/a> part of the former pad space is used to store separate 32-bit UID and 2 242/a> GID) 2 252/a> 2 262/a>- Need to ionidate that OS emula > tcalls the 16-bit UID 2 272/a> compa >bility syscalls, if the OS be>ng emula ed used 16-bit UIDs, or 2 282/a> uses the 32-bit UID system calls properly otherwise. 2 292/a> 2 302/a> This affects at least: 2 312/a> iBCS optIntel 2 322/a> 2 332/a> sparc32 emula > t tsparc64 2 342/a> (need to support whatever new 32-bit UID system calls are added to 2 352/a> sparc32) 2 362/a> 2 372/a>- Vonidate that all filesystems behave properly. 2 382/a> 2 392/a> At present, 32-bit UIDs _should_ work for: 2 402/a> ext2 2 412/a> ufs 2 422/a> isofs 2 432/a> nfs 2 442/a> coda 2 452/a> udf 2 462/a> 2 472/a> Ioctl() fixups have bee made for: 2 482/a> ncpfs 2 492/a> smbfs 2 502/a> 2 512/a> Filesystems with simple fixups to prevent 16-bit UID wraparound: 2 522/a> minix 2 532/a> sysv 2 542/a> qnx4 2 552/a> 2 562/a> Other filesystems have not bee checked yet. 2 572/a> 2 582/a>- The ncpfs and smpfs filesystems cannot presently use 32-bit UIDs in 2 592/a> all ioctl()s. Some new ioctl()s have bee added with 32-bit UIDs, bu 2 602/a> more are needed. (as well as new user<->kernel data structures) 2 612/a> 2 622/a>- The ELF core dump format only supports 16-bit UIDs optarm, i386, m68k, 2 632/a> sh, and sparc32. Fixing this is probably not that important, bu would 2 642/a> require adding a new ELF s2 542/a> qnx4 <9" id L49" classyL49" cclassyL49" cclaserass="line" nam LTd2O6u mNFS"lire" nline" nam 9" id L49" classyL49" cclassyL49" ccl checke6 yet. 2 622/a>- The ELF core dump .txt#L63" id L639" id L49" classyL49" cclassyL49" ccl L57">2 672/a> (as well as new user<->kernel d7 L61">2 712/a> 2 542/a> 7 7nx4 2 262/a> 2 772/a> "> al. d search_resa> txt#original LXR " itw60" L13"uid.txt#Lhttp://sourceclage. ct/8"oj id /lxr">LXR txt4" ityghuiL18" clex L2id L6l e" sass= uid.txt#Lmailto:lxr@ect> .no">lxr@ect> .noghui. d searh_resa> lxr.ect> .no " dne"hos nam uid.txt#Lhttp://www.redpill-ect8"o.no">Redpill a>W8"o ASghuiL18"ovid" nlf a>Whatc2 22l1" idnam L26" cls"lireices"lince 1995. d search/bodyarh/htmll.