ChangeSet@1.1549, 2004-11-25 17:46:49-02:00, marcelo@logos.cnet fix get_user_pages() change typo ChangeSet@1.1548, 2004-11-25 15:21:31-02:00, marcelo@logos.cnet Changed EXTRAVERSION to -pre1 ChangeSet@1.1547, 2004-11-25 15:20:10-02:00, marcelo@logos.cnet Andrea: get_user_pages handle ZERO_PAGE PG_reserved page, BUG otherwise. ChangeSet@1.1546, 2004-11-25 14:02:21-02:00, chrisw@osdl.org [PATCH] /proc/tty/driver/serial reveals the exact number of characters used in serial links (CAN-2003-0461) /proc/tty/driver/serial in Linux 2.4.x reveals the exact number of characters used in serial links, which could allow local users to obtain potentially sensitive information such as the length of passwords. To this issue has been assigned CAN-2003-0461. ChangeSet@1.1545, 2004-11-25 13:56:49-02:00, solar@openwall.com [PATCH] Fix 32-bit syscall emulation waste of CPU resources + if (n >= (MAX_ARG_PAGES * PAGE_SIZE) / sizeof(char *)) + return -E2BIG; in 32-bit syscall emulation on some 64-bit archs are essentially the same that we've been applying to native execve(2) syscall about five years ago (at around 2.2.12). These solve a local DoS where the kernel could be caused to spend a lot of time in a busy loop scanning repeated pages of garbage supplied for program arguments. ChangeSet@1.1544, 2004-11-25 13:51:50-02:00, solar@openwall.com [PATCH] Fix SCSI tape driver return code ChangeSet@1.1543, 2004-11-25 13:16:49-02:00, vince@arm.linux.org.uk [PATCH] vga16fb: Fix frame buffer bad memory mapping The vga16fb driver uses a direct ioremap on 0xa00000 to gain access to the vga card. This is wrong on architectures other than x86, every other driver uses VGA_MAP_MEM macro from vga.h to ensure the correct memory mapping. All this patch does is add the mapping macro this has been tested and works on ARM systems The driver no longer maps parts of kernel workspace and modifies it. ChangeSet@1.1542, 2004-11-25 13:09:36-02:00, marcelo.tosatti@cyclades.com [PATCH] Ignore vma's with PageReserved pages at get_user_pages() Reserved pages which count is incremented by get_user_pages() will leak (never be freed) because page freeing path (__free_pages_ok) ignores PageReferenced pages. I dont see any practical problem with 2.4.x right now. get_user_pages() wont be called on driver created VMA's with PageReserved pages because of the VM_IO bit which is set at remap_page_range(). Its not possible to have any vma mapped by a driver without VM_IO set. But the network packet mmap was an isolated case, so I'll apply Andrea's fix just for safety, although I can't find any offender in the tree. ChangeSet@1.1541, 2004-11-25 11:58:37-02:00, marcelo@logos.cnet Marc-Christian Petersen: VM documentation update ChangeSet@1.1540, 2004-11-25 11:52:13-02:00, andrea@suse.de [PATCH] Lazily add anonymous pages to LRU Add anonymous pages lazily to LRU (on pte unmapping pass) instead of adding them immediately on page creation. This avoids severe contention on pagemap_lru_lock on high-end boxes. Andrea's comments: Having anon pages immediatly inserted in the lru allows the VM to know better when it's worthwhile to start swapping anonymous ram, it will start to swap earlier and it should swap smoother and faster, but it will decrease scalability on the >16-ways of an order of magnitude. Big SMP/NUMA definitely can't take an hit on a global spinlock at every anon page allocation. So this is off by default. ChangeSet@1.1539, 2004-11-25 11:31:14-02:00, bunk@fs.tum.de [PATCH] remove outdated Stallion contact information I got the following error when trying to contact Stallion Technologies via the email address in MAINTAINERS: <-- snip --> : Sorry, I couldn't find a mail exchanger or IP address. (#5.4.4) <-- snip --> I've checked their web page, and it only lists the cards in question as "discontinued products" and offers a Linux 2.0 and 2.2 driver version, which is hardly "Supported". The patch below (applies against both 2.4 and 2.6) removes the bouncing email address from all files and removes the outdated MAINTAINERS entry. diffstat output: MAINTAINERS | 5 ----- drivers/char/istallion.c | 2 +- drivers/char/stallion.c | 2 +- include/linux/cd1400.h | 2 +- include/linux/cdk.h | 2 +- include/linux/comstats.h | 2 +- include/linux/istallion.h | 2 +- include/linux/sc26198.h | 2 +- include/linux/stallion.h | 2 +- 9 files changed, 8 insertions(+), 13 deletions(-) Signed-off-by: Adrian Bunk ChangeSet@1.1538, 2004-11-25 11:23:28-02:00, a.pugachev@pcs-net.net [PATCH] via82cxxx audio procfs code selection fix Fix via82cxxx_audio procfs code selection ChangeSet@1.1537, 2004-11-24 18:02:56-02:00, manfred99@gmx.ch [PATCH] Tigran Aivazian: backport sigmatch() issue in microcode.c [PATCH] fix to microcode driver for the old CPUs. Here is a patch against Linux 2.6.7 which fixes the sigmatch() macro to work for the relatively old processors as well, which have 'pf == 0' (processor flags as read from MSR 0x17), For example, the processors failing without this patch are Pentium II 300 MHz (Klamath) with family/model/stepping 6/3/4 and 6/3/3. ChangeSet@1.1536, 2004-11-24 17:23:10-02:00, mikpe@csd.uu.se [PATCH] gcc34 fastcall mismatch fixes for rwsem-spinlock Kernel 2.4.28 fails to compile with gcc-3.4 when CONFIG_M386=y because of fastcall mismatches between include/linux/rwsem-spinlock.h and lib/rwsem-spinlock.c. This patch fixes that. Signed-off-by: Mikael Pettersson lib/rwsem-spinlock.c | 14 +++++++------- 1 files changed, 7 insertions(+), 7 deletions(-) ChangeSet@1.1535, 2004-11-24 04:18:12-02:00, sezeroz@ttnet.net.tr [PATCH] ide-scsi update from ac/rh: Added transform for reading ATAPI tape drive block limits ide-scsi v0.94 from -ac/redhat, affecting ide-tape drives, * Ver 0.94 Sep 09 03 Added transform for reading ATAPI tape drive block * limits (ATAPI tapes report block limits in mode * page 0x2A, not by "read block limits" command) ChangeSet@1.1534, 2004-11-23 12:46:38-08:00, herbert@gondor.apana.org.au [NETLINK]: Invoke netlink_proto_init() correctly in non-modular case. Signed-off-by: Herbert Xu Signed-off-by: David S. Miller ChangeSet@1.1533, 2004-11-23 12:45:23-08:00, davem@nuts.davemloft.net [NETLINK]: sed 's/->sk_/->//' in af_netlink.c Signed-off-by: David S. Miller ChangeSet@1.1532, 2004-11-23 10:49:32-02:00, sezeroz@ttnet.net.tr [PATCH] Mark InterMezzo as orphan ChangeSet@1.1531, 2004-11-23 08:01:10-02:00, barryn@pobox.com [PATCH] binfmt_elf.c fix for 32-bit apps with large bss This is a 2.4.27-2.4.28 port of this patch: > [PATCH] binfmt_elf.c fix for 32-bit apps with large bss > > From: Julie DeWandel > > A problem exists where a 32-bit application can have a huge bss, one that > is so large that an overflow of the TASK_SIZE happens. But in this case, > the overflow is not detected in load_elf_binary(). Instead, because > arithmetic is being done using 32-bit containers, a truncation occurs and > the program gets loaded when it shouldn't have been. Subsequent execution > yields unpredictable results. > > The attached patch fixes this problem by checking for the overflow > condition and sending a SIGKILL to the application if the overflow is > detected. This problem can in theory exist when loading the elf > interpreter as well, so a similar check was added there. Signed-off-by: Barry K. Nathan ChangeSet@1.1530, 2004-11-23 08:00:49-02:00, barryn@pobox.com [PATCH] Fix ELF exec with huge bss This is a 2.4.27-2.4.28 port of the following patch: http://linux.bkbits.net:8080/linux-2.5/cset@3ff112802L-9-rs0BbkozDnTnpch9w > [PATCH] fix ELF exec with huge bss > > From: Roland McGrath > > The following test program will crash every time if dynamically linked. > I think this bites all 32-bit platforms, including 32-bit executables on > 64-bit platforms that support them (and could in theory bite 64-bit > platforms with bss sizes beyond the bounds of comprehension). > > volatile char hugebss[1080000000]; > main() { printf("%p..%p\n", &hugebss[0], &hugebss[sizeof hugebss]); > system("cat /proc/$PPID/maps"); > hugebss[sizeof hugebss - 1] = 1; > return 23; > } > > The problem is that the kernel maps ld.so at 0x40000000 or some such place, > before it maps the bss. Here the bss is so large that it overlaps and > clobbers that mapping. I've changed it to map the bss before it loads the > interpreter, so that part of the address space is reserved before ld.so's > mapping (which doesn't really care where it goes) is done. > > This patch also adds error checking to the bss setup (and interpreter's bss > setup). With the aforementioned change but no error checking, "ulimit -v > 65536; ./hugebss" will crash in the store after the `system' call, because > the kernel will have failed to allocate the bss and ignored the error, so > the program runs without those pages being mapped at all. With this change > it dies with a SIGKILL as for a failure to set up stack pages. It might be > even better to try to detect the case earlier so that execve can return an > error before it has wiped out the address space. But that seems like it > would always be fragile and miss some corner cases, so I did not try to add > such complexity. Signed-off-by: Barry K. Nathan ChangeSet@1.1529, 2004-11-23 07:58:24-02:00, sezeroz@ttnet.net.tr [PATCH] intermezzo, fixes from cvs While messing with intermezzo, I ran into this patch at sourceforge cvs which still isn't in 2.4. Please review, and apply if appropriate. from intermezzo cvs repo at Sourceforge: * Applied and tested Domen Puncer's patch to fs25/fs24 part code for list related processing * Applied and tested Renaud Duhaut(rd@duhaut.com)'s patch to solve fs24's InterMezzo's SLAB allocator leakage problem, on 2.4 kernel, now it can insmod and run, then rmmod. Insmod again won't cause Oops any more. :-) =================================================================== RCS file: /cvsroot/intermezzo/intermezzo/fs24/fileset.c,v retrieving revision 1.17 retrieving revision 1.18 ChangeSet@1.1528, 2004-11-23 07:56:18-02:00, sezeroz@ttnet.net.tr [PATCH] intermezzo, backport some more fixes from 2.6 intermezzo more fixes from 2.6: - Don't leak a pathname ref on error - Don't do putname() on a nameidata. ChangeSet@1.1527, 2004-11-23 07:55:29-02:00, sezeroz@ttnet.net.tr [PATCH] intermezzo, backport some fixes from 2.6 intermezzo, 2.6 backports: o unreachable code in methods.c o double semicolon in journal_xfs.c o potential datarace possibility and leak in psdev.c ChangeSet@1.1526, 2004-11-23 05:04:07-02:00, zaitcev@redhat.com [PATCH] USB: ohci fix by Jes&Pete for Jessie In usb-ohci of 2.4.28 it is possibe for del_urb called from a timer to access a thread stack under a stack pointer, when trying to wake the thread. On ia64 this causes an exception, which causes Jessie's Altix to collapse on reboot if a USB keyboard is not connected. It is obviously a bug everywhere, but for some reason it always worked dandy on i386. In my opinion, the root cause is a misguided attempt to save memory by allocating pointers to waitqueue heads intstead of waitqueue heads themselves, so I solve the problem by allocating waitqueue heads. This problem was analyzed and fixed by Jes Sorensen in a different way which wasn't radical enough for me, so this is a counter patch. ChangeSet@1.1525, 2004-11-23 05:03:34-02:00, zaitcev@redhat.com [PATCH] USB: fix ohci_complete_add This is a fix for a ludicrously stupid bug in my code in usb-ohci, which only affects 2.4, fortunately. The problem should be obvious from the code: when adding an element to the queue, an URB is lost if the queue contains two or more elements already. The fix is to implement the queue correctly: add to the tail and do not corrupt the list. Fortunately for us, this situation is rare and its impact is limited. More than two URBs have to complete in the same interrupt for this to happen, and this typically takes several devices operating simultaneously. When it happens, there is no memory leak and no oops, just a "lost callback" sort of situation. I got to know about this when a customer reported that a when three USB-serial adapters are connected to a system, they start to lose interrupts when traffic gets heavy. ChangeSet@1.1524, 2004-11-23 04:56:29-02:00, marcelo@logos.cnet Cset exclude: m.c.p@kernel.linux-systeme.com|ChangeSet|20041122173550|59288 ChangeSet@1.1523, 2004-11-23 04:55:38-02:00, mjagdis@eris-associates.co.uk [PATCH] Mike Jagdis CREDITS email address change Plus one *ancient* device driver that has my footprints in it :-). The patch is against 2.4.28 but it seems to apply cleanly to the last 2.6 release modulo some 10s of lines of fuzz. Mike ChangeSet@1.1458.1.8, 2004-11-22 21:06:37-05:00, len.brown@intel.com [ACPI] BIOS workaround allowing devices to use reserved IO ports Author: David Shaohua Li http://bugzilla.kernel.org/show_bug.cgi?id=3049 ChangeSet@1.1519.1.2, 2004-11-22 15:54:26-08:00, manfred99@gmx.ch [ATM]: Force -n option in gzip invocation. Signed-off-by: David S. Miller ChangeSet@1.1519.1.1, 2004-11-22 15:53:22-08:00, manfred99@gmx.ch [DECNET]: dn_neigh.c needs linux/module.h Signed-off-by: David S. Miller ChangeSet@1.1516.1.20, 2004-11-22 15:51:28-02:00, sezeroz@ttnet.net.tr [PATCH] hamradio scc warning type fix include/linux/scc.h: See the TIMER_OFF definition in scc.h and if (scc->kiss.mintime != TIMER_OFF) and similar checks against TIMER_OFF in drivers/net/hamradio/scc.c. ChangeSet@1.1516.1.19, 2004-11-22 15:44:41-02:00, marcelo@logos.cnet O.Sezer: cpqphp_pci.c size warning fix ChangeSet@1.1516.1.18, 2004-11-22 15:35:50-02:00, m.c.p@kernel.linux-systeme.com [PATCH] ASM: Warning: indirect lcall without `*' --Boundary-00=_u53nBq6Cct7UpSO Content-Transfer-Encoding: 7bit Content-Type: text/plain; charset="iso-8859-15" Content-Disposition: inline Hi Marcelo, these warnings are in 2.4 since centuries. [i386: pci-pc.c]: Fix up asm warnings 'indirect lcall without `*' {standard input}: Assembler messages: {standard input}:1014: Warning: indirect lcall without `*' {standard input}:1091: Warning: indirect lcall without `*' {standard input}:1176: Warning: indirect lcall without `*' {standard input}:1255: Warning: indirect lcall without `*' {standard input}:1271: Warning: indirect lcall without `*' {standard input}:1281: Warning: indirect lcall without `*' {standard input}:1349: Warning: indirect lcall without `*' {standard input}:1364: Warning: indirect lcall without `*' {standard input}:1375: Warning: indirect lcall without `*' {standard input}:1874: Warning: indirect lcall without `*' {standard input}:1960: Warning: indirect lcall without `*' Since I hate warnings, fix it up please :) Signed-off-by: Marc-Christian Petersen ciao, Marc --Boundary-00=_u53nBq6Cct7UpSO Content-Transfer-Encoding: 7bit Content-Type: text/x-diff; charset="iso-8859-15"; name="999_asm-indirect-lcall-without-star.patch" Content-Disposition: attachment; filename="999_asm-indirect-lcall-without-star.patch" [i386: pci-pc.c]: Fix up asm warnings 'indirect lcall without `*' {standard input}: Assembler messages: {standard input}:1014: Warning: indirect lcall without `*' {standard input}:1091: Warning: indirect lcall without `*' {standard input}:1176: Warning: indirect lcall without `*' {standard input}:1255: Warning: indirect lcall without `*' {standard input}:1271: Warning: indirect lcall without `*' {standard input}:1281: Warning: indirect lcall without `*' {standard input}:1349: Warning: indirect lcall without `*' {standard input}:1364: Warning: indirect lcall without `*' {standard input}:1375: Warning: indirect lcall without `*' {standard input}:1874: Warning: indirect lcall without `*' {standard input}:1960: Warning: indirect lcall without `*' Since I hate warnings, fix it up please :) Signed-off-by: Marc-Christian Petersen ChangeSet@1.1516.1.17, 2004-11-22 15:32:00-02:00, sezeroz@ttnet.net.tr [PATCH] OPTI Viper-M/N+ chipset support (by Michael Mueller) ChangeSet@1.1516.1.16, 2004-11-22 15:30:49-02:00, sezeroz@ttnet.net.tr [PATCH] Cure ISDN eicon size warning drivers/isdn/eicon/fourbri.c: The loop sez: for ( i = 0 ; i < 300 ; ++i ) Change "byte" to "int" ChangeSet@1.1516.1.15, 2004-11-22 15:24:16-02:00, sezeroz@ttnet.net.tr [PATCH] DAC960 firmware/alpha backport from 2.6 ChangeSet@1.1516.1.14, 2004-11-22 15:13:04-02:00, sezeroz@ttnet.net.tr [PATCH] Wilfried Weissmann: hptraid v0.03 from -ac/redhat - minor fixes I have some cosmetic updates for the hptraid driver. Changelog since 0.02: ===================== * register the raid volume only if all disks are available * print a warning that raid-(0+)1 failover is not supported Wilfried Weissmann http://marc.theaimsgroup.com/?l=3Dlinux-kernel&m=3D106287827628378&w=3D2 in -ac/redhat ChangeSet@1.1516.1.13, 2004-11-22 15:04:34-02:00, sezeroz@ttnet.net.tr [PATCH] ricoh.h, mem0 wrong definition v2.6 backport ChangeSet@1.1516.1.12, 2004-11-22 15:03:37-02:00, ralf@linux-mips.org [PATCH] MIPS documentation Document that CONFIG_OLD_TIME_C is gone. ChangeSet@1.1516.1.11, 2004-11-22 15:03:04-02:00, ralf@linux-mips.org [PATCH] MIPS: MAINTAINERS update o Maciej takes over maintenance of the DECFXX driver o Remove Steven Hills's entry for the Nino at his request ChangeSet@1.1516.1.10, 2004-11-22 15:02:11-02:00, ralf@linux-mips.org [PATCH] MIPS: Configure.help updates A bunch of MIPS updates for Configure.help: o CPU configuration o Turbochannel support for DEFXX family of net work cards o Turbochannel support for TGA framebuffers o Building the 64-bit kernel binary as a 64-bit ELF file ChangeSet@1.1516.1.9, 2004-11-22 14:46:31-02:00, ralf@linux-mips.org [PATCH] MIPS: sound drivers for AMD Alchemy platforms Two more sound drivers for the endless stream of AMD Alchemy platforms. ChangeSet@1.1516.1.8, 2004-11-22 14:34:54-02:00, ralf@linux-mips.org [PATCH] MIPS update o Gcc 3.4 and 3.5 bullet proofing o Support for PMC-Sierra Yosemite o Support for Momentum Jaguar o Support for Momentum Stretch o Support for AMD Alchemy DB1550 o More error checking for the MIPS addinird tool o KGDB register corruption fixes o Delete old time code o Big cleanup of code for NEC VR-series o Return the Swarm sound driver back to work o Driver for the builtin RM7000 interrupt controller o Fix build on systems without LLD / SCD o Several fixes of symbol exports to modules o Handle .rodata.* sections in linker script also o Fix swarm sound driver o Fix off by one bug in IRIX compat uname syscall ChangeSet@1.1516.1.7, 2004-11-22 05:28:41-02:00, khali@linux-fr.org [PATCH] I2C updates: get rid of unused code Two defined constants in include/linux/i2c.h aren't used anywhere, haven't ever been and won't ever be. They simply don't correspond to anything in the i2c core. Let's get rid of them. Signed-off-by: Jean Delvare ChangeSet@1.1516.1.6, 2004-11-22 05:27:59-02:00, khali@linux-fr.org [PATCH] I2C updates: lack trailing newline in logs Original report and discussion: http://archives.andrew.net.au/lm-sensors/msg28295.html Bottom line: Some debug messages in the i2c-core lack their trailing newline, which breaks the logs. Signed-off-by: Jean Delvare ChangeSet@1.1516.1.5, 2004-11-22 05:26:56-02:00, khali@linux-fr.org [PATCH] I2C updates: hardcoded buffer size should depend on define Original report and discussion: http://marc.theaimsgroup.com/?l=linux-arm-kernel&m=109816546827995&w=2 http://marc.theaimsgroup.com/?l=linux-arm-kernel&m=109926079025024&w=2 Bottom line: Two hardcoded buffer sizes in i2c_smbus_xfer_emulated (i2c-core) should depend on I2C_SMBUS_BLOCK_MAX. Else increasing I2C_SMBUS_BLOCK_MAX (in include/linux/i2c.h) will result in buffer overflows. Credits go to Tehn Yit Chin for noticing the suspicious hardcoded values. Signed-off-by: Jean Delvare ChangeSet@1.1516.1.4, 2004-11-22 05:25:45-02:00, khali@linux-fr.org [PATCH] I2C updates: i2c proc parser fix Original report and discussion: http://archives.andrew.net.au/lm-sensors/msg28053.html Bottom line: The real parser in i2c-proc is partly broken, but in a way it will go unnoticed in most cases (thus unspotted so far). Still worth fixing IMHO. Signed-off-by: Jean Delvare ChangeSet@1.1516.1.3, 2004-11-22 05:11:41-02:00, Andries.Brouwer@cwi.nl [PATCH] backport v2.6 largefile isofs fix People encounter daily the fact that Linux cannot handle large files on CDROM/DVD. I fixed this long ago for 2.6, but 2.4 still has the old behaviour. Below the patch as currently found in 2.6. Untested. It does two things: (i) According to the standard various fields are unsigned; make them unsigned, so that people with large files will not see negative sizes. (ii) The old code automatically throws away the high bits of the file size - it does not believe that one can have 2 GB files, and believes that the CDROM or DVD is made by a buggy mastering program. The code below does not truncate large files. If anybody still has a CDROM where this truncation is desired, it can be mounted with the old "cruft" mount option. Andries [the patch looks larger than it is because indentation changed] ChangeSet@1.1516.1.2, 2004-11-21 17:53:58-02:00, khali@linux-fr.org [PATCH] I2C updates: i2c_register_entry shouldn't rely on the procname field Original report and discussion: http://archives.andrew.net.au/lm-sensors/msg18839.html Bottom line: i2c_register_entry shouldn't rely on the procname field to detect the end of the control table, but on the ctl_name field. The latter is guaranteed to be non-zero except for the table terminator, the former can be null even in the middle of the table. The bug wasn't caught so far because all users of this function (basically the lm_sensors project's drivers) were exporting all entries through /proc so procname was never null (except for the table terminator, obviously). Credits go to Louis-Martin Cote for discovering the bug and proposing a fix. Signed-off-by: Jean Delvare ChangeSet@1.1516.1.1, 2004-11-19 17:44:39-02:00, kravetz@us.ibm.com [PATCH] Task name handling static copy v2.6 backport There is a problem with task name handling in the /proc fs. See http://www.ussg.iu.edu/hypermail/linux/kernel/0407.1/0136.html for the patch that eventually made its way into the 2.6 tree. We now have people experiencing the same problem/bug in 2.4. Here is a patch for 2.4 that implements the same fix. Please consider applying. Thanks, Signed-off-by: Mike Kravetz ChangeSet@1.1519, 2004-11-18 14:21:44-08:00, herbert@gondor.apana.org.au [NETLINK]: Backport pid hashing changes from 2.6 Signed-off-by: Herbert Xu Signed-off-by: David S. Miller ChangeSet@1.1518, 2004-11-18 14:05:40-08:00, lnville@tuxdriver.com [VLAN]: change_mtu should return 0 on success. Signed-off-by: John W. Linville Signed-off-by: David S. Miller ChangeSet@1.1517, 2004-11-18 13:54:00-08:00, davem@nuts.davemloft.net [TCP]: Receive buffer moderation fixes. 1) Make window clamp follow receive buffer growth so it does not limit window advertisements. Noticed by John Heffner and Stephen Hemminger. 2) Fix rcvmem calculation such that tcp_adv_win_scale is taken into account. Signed-off-by: David S. Miller ChangeSet@1.1516, 2004-11-18 06:34:41-02:00, jack@ucw.cz [PATCH] Configurable quota messages Here is a patch which allows root to turn of quota messages to console (some users don't like them because they disturb other console output). If you think the patch could go into 2.4 tree then please apply it. ChangeSet@1.1514, 2004-11-17 15:19:29-02:00, rddunlap@osdl.org [PATCH] oops on boot when initializing CDROM ChangeSet@1.1499.4.2, 2004-11-17 11:13:14-06:00, shaggy@austin.ibm.com JFS: fix race in jfs_commit_inode There was a race that resulted in old, deleted inodes being written to disk after the inode number had been reused. jfs_commit_inode needs to verify that the inode is still linked and dirty before committing it. Signed-off-by: Dave Kleikamp ChangeSet@1.1513, 2004-11-17 11:55:34-02:00, tonyb@cybernetics.com [PATCH] fix race condition in sg.c This patch fixes a command completion race condition in the SCSI generic driver. There are several variations of this patch floating around: #1: Original patch from me: http://marc.theaimsgroup.com/?t=105664088400001&r=1&w=2 #2: Better patch from Brian King for lk 2.6.x: http://marc.theaimsgroup.com/?t=107843139000002&r=1&w=2 #3: Patch for a related problem from Brian King for lk 2.6.x: http://marc.theaimsgroup.com/?t=108636397100004&r=1&w=2 #4: Port of combined #2 and #3 to lk 2.4.x from Tom Coughlan: http://marc.theaimsgroup.com/?t=109657360500004&r=1&w=2 #5: Correction of #4 by me: http://marc.theaimsgroup.com/?l=linux-scsi&m=109658140509138&w=2 None of these fixes have made it into lk 2.4.x yet. Below is a port of patch #2 to 2.4.x which has been extensively tested by my company. Patch #5 which includes these changes would be better, but I haven't tested it. Please apply this patch, and the additional changes for patch #5 can be merged later if and when they are tested by those interested. ChangeSet@1.1512, 2004-11-17 11:41:32-02:00, tonyb@cybernetics.com [PATCH] fix for scsi_unjam_host: Miscount of number of failed commands This is a resend of http://marc.theaimsgroup.com/?l=linux-scsi&m=108438814301446&w=2 This patch fixes a panic "scsi_unjam_host: Miscount of number of failed commands". This panic happens only under very specific circumstances: 1) A command fails with check condition 2) The low-level driver doesn't auto-retrieve the sense data 3) The sense key is "not ready" and ASC/ASCQ indicate the unit is becoming ready 4) A retry of the command fails in the same way Please apply. ChangeSet@1.1511, 2004-11-17 11:41:03-02:00, tonyb@cybernetics.com [PATCH] make SCSI error handler preserve data transfer residual This is a resend of http://marc.theaimsgroup.com/?l=linux-scsi&m=108445913200046&w=2. When a SCSI command fails, the error handler in 2.4.x may send a command such as request sense which clobbers the data transfer residual value of the original command. This patch makes the error handler save and restore the data transfer residual so that the value passed back to the upper level reflects the residual of the original command rather than the error handler's internal command. This makes a difference for userspace applications that use the /dev/sg* interface that require a correct data transfer residual count even for commands that went through error handling. Please apply. ChangeSet@1.1510, 2004-11-17 11:39:55-02:00, tonyb@cybernetics.com [PATCH] sym53c8xx_2 Ultra 160 requires LVD This patch prevents sym53c8xx_2 from attempting to negotiate Ultra 160 on a SE or HVD bus, which is invalid and confuses some devices. Please apply. ChangeSet@1.1509, 2004-11-17 11:37:00-02:00, tonyb@cybernetics.com [PATCH] sym53c8xx_2 sniff inquiry fix This is a resend of http://marc.theaimsgroup.com/?l=linux-scsi&m=108117878914404&w=2 The sym53c8xx_2 driver in lk 2.4.x is unmaintained. This is a backported fix from lk 2.6.x for a problem with sniffing inquiry data incorrectly resulting in Ultra 160 negotiation being disabled for some devices. Please apply. ChangeSet@1.1508, 2004-11-17 11:36:32-02:00, tonyb@cybernetics.com [PATCH] sym53c8xx_2 error handler fix The sym53c8xx_2 driver in lk 2.4.x is unmaintained. This is a backported fix from lk 2.6.x for a problem with the error handler never successfully completing an abort, bus reset, device reset, or host reset. This patch has been well tested at my company. Please apply. ChangeSet@1.1499.1.13, 2004-11-17 10:11:45-05:00, jgarzik@pobox.com Merge pobox.com:/garz/repo/netdev-2.4/e100e1000 into pobox.com:/garz/repo/net-drivers-2.4 ChangeSet@1.1499.1.12, 2004-11-17 10:11:22-05:00, jgarzik@pobox.com Merge pobox.com:/garz/repo/netdev-2.4/forcedeth into pobox.com:/garz/repo/net-drivers-2.4 ChangeSet@1.1499.1.11, 2004-11-17 05:49:55-02:00, marcelo@logos.cnet Changed Makefile to v2.4.28 TAG: v2.4.28