[lkml]   [1998]   [Oct]   [12]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: 2.1.125 Show stopper list: Draft
What about the PGE/PTE problem?  Any user can just mmap/munmap all possible
(768 under x86) page tables for their process, then use up a lot of GFP_KERNEL
memory. You end up crashing the machine because they can do this over and
over. We need to either deallocate the page tables when we deallocate references
to them, or we need to do that zero page trick, and refer to the VMA lists to
properly handled GPFs vs MMAP.


> This has a couple of errors left in it since I've not been round and
> tried to reproduce all those I can again.
> 2.1.10x APM changes
> 2.1.10x APM changes broke clock handling - clocks now drift
> apart
> Ports
> Finish merging down most of the other ports (Alpha, MIPS,
> Sparc, Sparc64..)
> Drivers
> Both the current MCA and network drivers in the kernel tree are
> horribly out of date and need updating badly, even if we cause
> some pain while doing so to debug them.
> Sound oopses
> Sound oopses on some bad load combinations and an unload
> problem somewhere. Also on a couple of other cases.
> Video4Linux
> The current .18 driver, the Diz driver and the kernel code all
> need to get fused into one very soon.
> Core dump
> Core dumps as root are still broken (rough patch exists)
> /proc
> Try rm -rf /proc (sync first)
> /proc/scsi
> cd /proc/scsi wait for a scsi module to unload. This is
> dangerous as any user can hide a daemon current dir there and
> catch an unwary sysadmin doing a scsi tape backup or using a
> scanner.
> Path MTU
> The cache doesn't expire. MTU poisoning attacks are lethally
> effective (patch exists)
> Loop module
> Hooks need improving for loadable loop modules (patch exists)
> Multicast
> Forwarder doesn't call into firewall code.
> NFS locking crashes
> NFS file locking causes crashes for some people [now SMP only]
> NFS writebehind crashes
> NFS write behind will sometimes crash nastily on retries. It
> also has bad performance problems talking to 8K page size BSD
> boxes.
> O_SYNC files are not synchronous.
> knfsd
> KNFSD fails to do some required permission and sanity checks
> that would normally be done by the generic code paths for mknod
> etc. Serious security implications.
> Ptrace
> SIGCONT sent to a process ptracing itself on an SMP machine.
> 2.1.x large fd sets
> Large fd sets allow excessive locked memory commits from both
> net and fs layers. Any user can crash the machine.
> 2.1.10x swap deadlock
> 2.1.10x can seize up looping through swap_out_vma, shrink_mmap
> and get_swap_page.
> Jiffy Handling
> Many drivers still do not handle jiffy overflows nicely
> ST driver
> The scsi tape driver is completely broken on error handling
> AHA1542 driver
> Error recovery is broken.
> CD ROM burners
> Due to locking problems with the current io_request locks you
> often can't burn CD's when you could in 2.0.x
> /proc problems
> Put security checks in the read() operators.
> ISDN4linux
> Copies data from user space with interrupts off. Occasionally
> passes non IP packets to IP error routines. See 2.0.34 fix for
> latter.
> fchmod bug
> Fchmod doesnt work on a bound unix domain socket as it does in
> Net stats
> Not all drivers record byte counters yet
> NR_OPEN default poor
> NR_OPEN is too high for the default fd limits
> Deadlocks
> Strace a process running electric fence and observe. Gdb one
> and die. (maybe fixed.. need to recheck)
> Miscellaneous fixes missing
> Assorted 2.0.31-36 fixes are missing from 2.1.x
> Not yet fully debugged under the 2.1.x VFS layer. Needs more
> testers/helpers (partially done)
> NFS over tcp
> This basically doesnt work right now. probably pull for 2.3
> 2.1.1xx CDROM Crash
> Soundblaster CDRxx CD-ROMS hang on SMP kernels. (patches out)
> SG driver
> Doesn't use vmalloc and scatter gather making scanner access
> horrible (2.3.x job)
> 2.1.x won't let you debug threads
> Some fixes for this in 2.0.34 need forward porting. (may all
> now be done)
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to
> Please read the FAQ at

Perry Harrington Linux rules all OSes. APSoft ()
email: Think Blue. /\

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
Please read the FAQ at

 \ /
  Last update: 2005-03-22 13:45    [W:0.128 / U:0.548 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site