ChangeSet@1.1503, 2004-08-19 13:11:29-03:00, thor@math.TU-Berlin.DE [PATCH] NetMOS 9805 ParPort interface here's a tiny patch against parport/parport_pc.c. It adds support for the NetMOS 9805 chip, used in several popular parallel port extension cards available here in germany. The patch below has been found working in a beige G3 Mac and a Canon BJC just fine. ChangeSet@1.1502, 2004-08-19 12:52:25-03:00, rainer.weikusat@sncag.com [PATCH] bkgoodman@bradgoodman.com: MTD cfi_cmdset_0002.c - Duplicate cleanup in error path "bradgoodman.com" writes: > Patch to 2.4.x: Corrects an obvious error where all of the cleanups are done > twice in the event of a chip programming error. This can result in > kernel BUG() getting called on subsequent programming attempts. > > +++ linux-2.4.22/drivers/mtd/chips/cfi_cmdset_0002.new > @@ -549,11 +549,6 @@ > } > } else { > printk(KERN_WARNING "Waiting for write to complete timed out in do_write_oneword."); > - > - chip->state = FL_READY; > - wake_up(&chip->wq); > - cfi_spin_unlock(chip->mutex); > - DISABLE_VPP(map); > ret = -EIO; > } > } I suggest to return after the first cleanup. That way, it is consistent with the other low-level chip access functions. But the algorithm is per se buggy, anyway, because except if DQ5 was raised before, the chip is not 'ready' (for reading array data), but still in programming mode and will remain there until the 'embedded programming algorithm' stops, because (according to the docs) a reset command will not be accepted until DQ5 has been raised and the opportunityto check for that is gone after the syscall returned to the caller. ChangeSet@1.1501, 2004-08-19 12:24:20-03:00, bunk@fs.tum.de [PATCH] disallow modular BINFMT_ELF with no 2.4 kernel BINFMT_ELF=m actually worked, you always get a <-- snip --> depmod: *** Unresolved symbols in /lib/modules/2.4.25-pre4/kernel/fs/binfmt_elf.o depmod: smp_num_siblings depmod: put_files_struct depmod: steal_locks <-- snip --> Since BINFMT_ELF=m is a very unusual case, the patch below simply disallows modular BINFMT_ELF. A similar patch I sent was already accepted into 2.6. diffstat output: Documentation/Configure.help | 6 ------ arch/alpha/config.in | 2 +- arch/arm/config.in | 2 +- arch/cris/config.in | 2 +- arch/i386/config.in | 2 +- arch/ia64/config.in | 2 +- arch/m68k/config.in | 2 +- arch/mips/config-shared.in | 2 +- arch/parisc/config.in | 2 +- arch/s390/config.in | 2 +- arch/s390x/config.in | 2 +- arch/sh/config.in | 2 +- arch/sh64/config.in | 2 +- arch/sparc/config.in | 2 +- arch/sparc64/config.in | 2 +- arch/x86_64/config.in | 2 +- 16 files changed, 15 insertions(+), 21 deletions(-) ChangeSet@1.1493.1.11, 2004-08-19 12:15:49-03:00, alan@lxorguk.ukuu.org.uk [PATCH] ad1816 sound driver web page and email address update ChangeSet@1.1493.1.10, 2004-08-19 12:05:14-03:00, david.martinez@rediris.es [PATCH] Update ftape webpage ChangeSet@1.1493.1.9, 2004-08-19 11:26:00-03:00, mikpe@csd.uu.se [PATCH] gcc34 inline failure fixes Marcelo, This patch fixes a number inlining failures with gcc-3.4.1 and the 2.4.28-pre1 kernel. There are five kinds of changes: - Trivial removals of unusable inlines. + get_buffer_flushtime(): defined in fs/buffer.c, only called from fs/jdb/journal.c + nmi_watchdog_tick(): defined in arch/{i386,x86_64}/kernel/nmi.c, only called from arch/{i386,x86_64}/kernel/traps.c + SELECT_INTERRUPT(): defined in drivers/ide/ide-iops.c, only called from drivers/ide/ide-io.c + QUIRK_LIST(): defined in drivers/ide/ide-iops.c, only called from drivers/ide/ide-probe.c - Removals of pointless inlines: + rtl8139_start_thread(): semi-heavy operation called infrequently + SELECT_DRIVE(): defined in drivers/ide/ide-iops.c, called from ide-iops.c and several other files; the calls in ide-iops.c are all in non-performance critical sections + SELECT_MASK(): similar to SELECT_DRIVE() + __mmdrop(): defined in kernel/fork.c, used in fork.c and several other files; the calls in fork.c are not performance critial enough to require inlining - Reordering to allow inlinining: + drivers/scsi/sg.c: move sg_jif_to_ms(), sg_alloc_kiovec(), and sg_free_kiovec() to before their first call sites + net/sunrpc/xprt.c: move do_xprt_reserve() to before its first and only call site - Wrappers to allow partial inlining of critical functions: + blk_get_queue(): split to allow internal calls in ll_rw_block.c to call the inlinable version + blk_seg_merge_ok(): similar to blk_get_queue() + ip_finish_output(): similar to blk_get_queue() - Replacing extern inline/normal functions duplication with only a single set of static inlines: + parport_pc defines a number of low-level functions both as extern inlines with bodies, and as normal functions in one file where those functions' addresses are exported. This causes errors from gcc-3.4. Most of these changes are from the 2.6 kernels. In some cases (the "wrappers to allow partial inlining" cases) the changes enable 2.4 to continue to (partially) inline a function even when 2.6 has elected to drop the inline attribute from that function. Compiled and booted on i386 SMP and UP, x86_64 SMP and UP, and ppc32 UP. Signed-off-by: Mikael Pettersson ChangeSet@1.1472.1.9, 2004-08-18 01:30:57-04:00, dougg@torque.net [libata] fix INQUIRY handling Changes: - send vendor, product and rev strings back for 36 byte INQUIRYs - set the additional length field to indicate 96 byte response is available ChangeSet@1.1472.1.8, 2004-08-18 01:28:59-04:00, jgarzik@pobox.com [libata] fix error recovery reference count and in-recovery flag This bug could potentially lead to soft hangs (processes stuck in D state) if an error occurred. Yet another undocumented nuance of the ->eh_strategy_handler. Thanks to excellent detective work by Brad Campbell tracking this down. Thanks also to Doug Ledford. ChangeSet@1.1498, 2004-08-16 17:10:30-07:00, kaber@trash.net [PKT_SCHED]: Remove unnecessary memsets in packet schedulers This patch removes some more unnecessary memsets in packet schedulers. The qdisc's private data is already set to 0 in qdisc_create/qdisc_create_dflt. Signed-off-by: Patrick McHardy Signed-off-by: David S. Miller ChangeSet@1.1493.1.7, 2004-08-16 19:37:23-03:00, neilb@cse.unsw.edu.au [PATCH] mark NFS/TCP not EXPERIMENTAL NFS/TCP has been used widely for a long time now, and is now enabled by many distros and typically recommended for lossy or congested networks. This patch removes the EXPERIMENTAL tag, updates the help text, and changes arch defconfigs to set it by default (note that several already did so). From: Greg Banks Signed-off-by: Neil Brown ### Diffstat output ./Documentation/Configure.help | 7 ++++--- ./arch/i386/defconfig | 2 +- ./arch/ia64/defconfig | 2 +- ./arch/mips/defconfig | 2 +- ./arch/mips64/defconfig | 2 +- ./arch/ppc/defconfig | 2 +- ./arch/ppc64/defconfig | 2 +- ./arch/s390/defconfig | 2 +- ./arch/s390x/defconfig | 2 +- ./arch/sh64/defconfig | 2 +- ./fs/Config.in | 2 +- 11 files changed, 14 insertions(+), 13 deletions(-) ChangeSet@1.1493.2.1, 2004-08-16 13:34:59-07:00, davem@nuts.davemloft.net [SPARC64]: Add missing nop to itlb_base.S Signed-off-by: David S. Miller ChangeSet@1.1493.1.6, 2004-08-16 15:05:31-03:00, sezeroz@ttnet.net.tr [PATCH] ns83820.c warning fixes ns83820.c warning fix from 2.6 ChangeSet@1.1493.1.5, 2004-08-16 15:04:52-03:00, sezeroz@ttnet.net.tr [PATCH] pm3fb and kaweth missing casts ChangeSet@1.1493.1.4, 2004-08-16 14:51:05-03:00, neilb@cse.unsw.edu.au [PATCH] Use llseek instead of f_pos= for directory seeking. nfsd currently just sets f_pos when seeking in a directory. This bypasses any checking and other handling that a filesystem might want to do. So instead, we use 'llseek' copied out of fs/read_write.c use that, both to seek at the start, and the find the new position at the end. If it were not too intrusive, would could export llseek as vfs_llseek as was done in 2.6.... Thanks to "Derrick Schommer" "Trond Myklebust" Signed-off-by: Neil Brown ### Diffstat output ./fs/nfsd/vfs.c | 35 ++++++++++++++++++++++++++++------- 1 files changed, 28 insertions(+), 7 deletions(-) ChangeSet@1.1493.1.3, 2004-08-16 14:50:04-03:00, neilb@cse.unsw.edu.au [PATCH] Fixed possibly xdr parsing error if write size exceed 2^31 xdr_argsize_check needs to cope with the possibility that the pointer has wrapped and could be below buf->base. Signed-off-by: Neil Brown ### Diffstat output ./fs/nfsd/nfs3xdr.c | 2 +- ./include/linux/nfsd/xdr3.h | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) ChangeSet@1.1493.1.2, 2004-08-16 14:47:13-03:00, neilb@cse.unsw.edu.au [PATCH] Allow larger NFSd MAXBLKSIZE on architectures with larger PAGE_SIZE This patch makes NFSSVC_MAXBLKSIZE depend on PAGE_SIZE so that machines with large page sizes can take advantage of that feature to serve NFS with larger blocksizes, increasing performance and avoiding a fallback to synchronous traffic between machines with page sizes greater than 8K. Also, documents the actual constraints on NFSSVC_MAXBLKSIZE. From: Greg Banks Signed-off-by: Neil Brown ### Diffstat output ./include/linux/nfsd/const.h | 12 ++++++++++-- 1 files changed, 10 insertions(+), 2 deletions(-) ChangeSet@1.1493.1.1, 2004-08-16 14:45:44-03:00, pbadari@us.ibm.com [PATCH] scsi PHYS_4G merging doesn't work We recently found that, BH_PHYS_4G() handling in scsi-merge code is broken. Instead of creating new segment when the IO crosses 4G boundary, its forcing to create a new request. So we end up not merging IOs and start doing small IOs. Only requirement is, driver can't handle crossing 4G boundary in a single segment - but we can have multiple segments doing IOs all over the place. Here is the patch to fix it (suggested by Jens Axboe). ChangeSet@1.1472.1.7, 2004-08-16 12:32:14-04:00, akpm@osdl.org [PATCH] libata build fix drivers/scsi/libata-core.c: In function `swap_buf_le16': drivers/scsi/libata-core.c:2073: `words' undeclared (first use in this function) drivers/scsi/libata-core.c:2073: (Each undeclared identifier is reported only once Signed-off-by: Andrew Morton ChangeSet@1.1497, 2004-08-15 19:29:29-07:00, kaber@trash.net [NET_SCHED]: O(1) children vtoff adjustment in HFSC scheduler Signed-off-by: Patrick McHardy Signed-off-by: David S. Miller ChangeSet@1.1496, 2004-08-15 19:29:12-07:00, kaber@trash.net [NET_SCHED]: Replace actlist by rbtrees in HFSC scheduler. Signed-off-by: Patrick McHardy Signed-off-by: David S. Miller ChangeSet@1.1495, 2004-08-15 19:28:45-07:00, kaber@trash.net [NET_SCHED]: Replace eligible list by rbtree in HFSC scheduler. Signed-off-by: Patrick McHardy Signed-off-by: David S. Miller ChangeSet@1.1494, 2004-08-15 19:28:06-07:00, kaber@trash.net [RBTREE]: Add rb_{first,last,prev,next} Signed-off-by: Patrick McHardy Signed-off-by: David S. Miller ChangeSet@1.1493, 2004-08-15 21:30:53-03:00, linville@tuxdriver.com [PATCH] Add IOI Media Bay to SCSI quirk list ChangeSet@1.1492, 2004-08-15 16:43:15-03:00, marcelo@logos.cnet Changed Makefile to 2.4.28-pre1 TAG: v2.4.28-pre1