- Linux-stable-mirror - lists.linaro.org

Re: [Linux-stable-mirror] [PATCH] drm/vc4: Fix NULL pointer dereference in vc4_save_hang_state()

by Boris Brezillon

On Wed, 17 Jan 2018 12:03:15 -0800 Eric Anholt <eric(a)anholt.net> wrote: > Boris Brezillon <boris.brezillon(a)free-electrons.com> writes: > > > When saving BOs in the hang state we skip one entry of the > > kernel_state->bo[] array, thus leaving it to NULL. This leads to a NULL > > pointer dereference when, later in this function, we iterate over all > > BOs to check their ->madv state. > > > > Fixes: ca26d28bbaa3 ("drm/vc4: improve throughput by pipelining binning and rendering jobs") > > Cc: <stable(a)vger.kernel.org> > > Signed-off-by: Boris Brezillon <boris.brezillon(a)free-electrons.com> > > --- > > drivers/gpu/drm/vc4/vc4_gem.c | 2 +- > > 1 file changed, 1 insertion(+), 1 deletion(-) > > > > diff --git a/drivers/gpu/drm/vc4/vc4_gem.c b/drivers/gpu/drm/vc4/vc4_gem.c > > index 6c32c89a83a9..19ac7fe0e5db 100644 > > --- a/drivers/gpu/drm/vc4/vc4_gem.c > > +++ b/drivers/gpu/drm/vc4/vc4_gem.c > > @@ -208,7 +208,7 @@ vc4_save_hang_state(struct drm_device *dev) > > kernel_state->bo[j + prev_idx] = &bo->base.base; > > j++; > > } > > - prev_idx = j + 1; > > + prev_idx = j; > > Could we replace the whole "[j + prev_idx]" with a "[k++]" and maybe a > WARN_ON_ONCE(k != state->bo_count) at the end? Sure. > > I really need to figure out if I can come up with a way to make IGT > cases for GPU hangs on vc4, despite the validation. I managed to trigger the NULL pointer dereference while debugging the perfmon stuff, but it's fixed now, so I don't have a way to easily force a reset. > I found a bug in > GPU reset due to BCL hangs when doing vc5, but I don't have a testcase. > Maybe some submit flags that overwrite the BCL or RCL to do an infinite > loop?

7 years, 5 months

1
0
0 0

[Linux-stable-mirror] [PATCH] usbip: list: don't list devices attached to vhci_hcd

by Shuah Khan

usbip host lists devices attached to vhci_hcd on the same server when user does attach over localhost or specifies the server as the remote. usbip attach -r localhost -b busid or usbip attach -r servername (or server IP) Fix it to check and not list devices that are attached to vhci_hcd. Cc: stable(a)vger.kernel.org Signed-off-by: Shuah Khan <shuahkh(a)osg.samsung.com> --- tools/usb/usbip/src/usbip_list.c | 9 +++++++++ 1 file changed, 9 insertions(+) diff --git a/tools/usb/usbip/src/usbip_list.c b/tools/usb/usbip/src/usbip_list.c index f1b38e866dd7..d65a9f444174 100644 --- a/tools/usb/usbip/src/usbip_list.c +++ b/tools/usb/usbip/src/usbip_list.c @@ -187,6 +187,7 @@ static int list_devices(bool parsable) const char *busid; char product_name[128]; int ret = -1; + const char *devpath; /* Create libudev context. */ udev = udev_new(); @@ -209,6 +210,14 @@ static int list_devices(bool parsable) path = udev_list_entry_get_name(dev_list_entry); dev = udev_device_new_from_syspath(udev, path); + /* Ignore devices attached to vhci_hcd */ + devpath = udev_device_get_devpath(dev); + if (strstr(devpath, USBIP_VHCI_DRV_NAME)) { + dbg("Skip the device %s already attached to %s\n", + devpath, USBIP_VHCI_DRV_NAME); + continue; + } + /* Get device information. */ idVendor = udev_device_get_sysattr_value(dev, "idVendor"); idProduct = udev_device_get_sysattr_value(dev, "idProduct"); -- 2.14.1

7 years, 5 months

1
0
0 0

[Linux-stable-mirror] [PATCH] usbip: prevent bind loops on devices attached to vhci_hcd

by Shuah Khan

usbip host binds to devices attached to vhci_hcd on the same server when user does attach over localhost or specifies the server as the remote. usbip attach -r localhost -b busid or usbip attach -r servername (or server IP) Unbind followed by bind works, however device is left in a bad state with accesses via the attached busid result in errors and system hangs during shutdown. Fix it to check and bail out if the device is already attached to vhci_hcd. Cc: stable(a)vger.kernel.org Signed-off-by: Shuah Khan <shuahkh(a)osg.samsung.com> --- tools/usb/usbip/src/usbip_bind.c | 9 +++++++++ 1 file changed, 9 insertions(+) diff --git a/tools/usb/usbip/src/usbip_bind.c b/tools/usb/usbip/src/usbip_bind.c index fa46141ae68b..e121cfb1746a 100644 --- a/tools/usb/usbip/src/usbip_bind.c +++ b/tools/usb/usbip/src/usbip_bind.c @@ -144,6 +144,7 @@ static int bind_device(char *busid) int rc; struct udev *udev; struct udev_device *dev; + const char *devpath; /* Check whether the device with this bus ID exists. */ udev = udev_new(); @@ -152,8 +153,16 @@ static int bind_device(char *busid) err("device with the specified bus ID does not exist"); return -1; } + devpath = udev_device_get_devpath(dev); udev_unref(udev); + /* If the device is already attached to vhci_hcd - bail out */ + if (strstr(devpath, USBIP_VHCI_DRV_NAME)) { + err("bind loop detected: device: %s is attached to %s\n", + devpath, USBIP_VHCI_DRV_NAME); + return -1; + } + rc = unbind_other(busid); if (rc == UNBIND_ST_FAILED) { err("could not unbind driver from device on busid %s", busid); -- 2.14.1

7 years, 5 months

1
0
0 0

[Linux-stable-mirror] [PATCH] ubifs: Check ubifs_wbuf_sync() return code

by Richard Weinberger

If ubifs_wbuf_sync() fails we must not write a master node with the dirty marker cleared. Otherwise it is possible that in case of an IO error while syncing we mark the filesystem as clean and UBIFS refuses to recover upon next mount. Cc: <stable(a)vger.kernel.org> Fixes: 1e51764a3c2a ("UBIFS: add new flash file system") Signed-off-by: Richard Weinberger <richard(a)nod.at> --- fs/ubifs/super.c | 14 ++++++++++---- 1 file changed, 10 insertions(+), 4 deletions(-) diff --git a/fs/ubifs/super.c b/fs/ubifs/super.c index 0beb285b143d..468bca452d0a 100644 --- a/fs/ubifs/super.c +++ b/fs/ubifs/super.c @@ -1739,8 +1739,11 @@ static void ubifs_remount_ro(struct ubifs_info *c) dbg_save_space_info(c); - for (i = 0; i < c->jhead_cnt; i++) - ubifs_wbuf_sync(&c->jheads[i].wbuf); + for (i = 0; i < c->jhead_cnt; i++) { + err = ubifs_wbuf_sync(&c->jheads[i].wbuf); + if (err) + ubifs_ro_mode(c, err); + } c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY); c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS); @@ -1806,8 +1809,11 @@ static void ubifs_put_super(struct super_block *sb) int err; /* Synchronize write-buffers */ - for (i = 0; i < c->jhead_cnt; i++) - ubifs_wbuf_sync(&c->jheads[i].wbuf); + for (i = 0; i < c->jhead_cnt; i++) { + err = ubifs_wbuf_sync(&c->jheads[i].wbuf); + if (err) + ubifs_ro_mode(c, err); + } /* * We are being cleanly unmounted which means the -- 2.13.6

7 years, 5 months

1
0
0 0

[Linux-stable-mirror] v4.4.112 build: 0 failures 0 warnings (v4.4.112)

by Build bot for Mark Brown

Tree/Branch: v4.4.112 Git describe: v4.4.112 Commit: 42375c1120 Linux 4.4.112 Build Time: 79 min 23 sec Passed: 9 / 9 (100.00 %) Failed: 0 / 9 ( 0.00 %) Errors: 0 Warnings: 0 Section Mismatches: 0 ------------------------------------------------------------------------------- defconfigs with issues (other than build errors): ------------------------------------------------------------------------------- =============================================================================== Detailed per-defconfig build reports below: ------------------------------------------------------------------------------- Passed with no errors, warnings or mismatches: arm64-allnoconfig arm64-allmodconfig arm-multi_v5_defconfig arm-multi_v7_defconfig x86_64-defconfig arm-allmodconfig arm-allnoconfig x86_64-allnoconfig arm64-defconfig close failed in file object destructor: sys.excepthook is missing lost sys.stderr

7 years, 5 months

1
0
0 0

[Linux-stable-mirror] Please apply commit 1b5c7ef3d0d0 ("drm/nouveau/disp/gf119: add missing drive vfunc ptr") on 4.14

by Sven Joachim

Hi Greg, Could you please pick up commit 1b5c7ef3d0d0 ("drm/nouveau/disp/gf119: add missing drive vfunc ptr") for the 4.14 series? It fixes https://bugs.freedesktop.org/show_bug.cgi?id=103421 which seems to break nouveau for everyone with a GF119 card. This problem has also been reported twice in Debian[1,2], and the Debian kernel team has already applied the patch. Cheers, Sven 1. https://bugs.debian.org/880660 2. https://bugs.debian.org/886727

7 years, 5 months

2
1
0 0

[Linux-stable-mirror] Patch "drm/nouveau/disp/gf119: add missing drive vfunc ptr" has been added to the 4.14-stable tree

by gregkh＠linuxfoundation.org

This is a note to let you know that I've just added the patch titled drm/nouveau/disp/gf119: add missing drive vfunc ptr to the 4.14-stable tree which can be found at: http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=sum… The filename of the patch is: drm-nouveau-disp-gf119-add-missing-drive-vfunc-ptr.patch and it can be found in the queue-4.14 subdirectory. If you, or anyone else, feels it should not be added to the stable tree, please let <stable(a)vger.kernel.org> know about it. >From 1b5c7ef3d0d0610bda9b63263f7c5b7178d11015 Mon Sep 17 00:00:00 2001 From: Rob Clark <robdclark(a)gmail.com> Date: Sat, 6 Jan 2018 10:59:41 -0500 Subject: drm/nouveau/disp/gf119: add missing drive vfunc ptr From: Rob Clark <robdclark(a)gmail.com> commit 1b5c7ef3d0d0610bda9b63263f7c5b7178d11015 upstream. Fixes broken dp on GF119: Call Trace: ? nvkm_dp_train_drive+0x183/0x2c0 [nouveau] nvkm_dp_acquire+0x4f3/0xcd0 [nouveau] nv50_disp_super_2_2+0x5d/0x470 [nouveau] ? nvkm_devinit_pll_set+0xf/0x20 [nouveau] gf119_disp_super+0x19c/0x2f0 [nouveau] process_one_work+0x193/0x3c0 worker_thread+0x35/0x3b0 kthread+0x125/0x140 ? process_one_work+0x3c0/0x3c0 ? kthread_park+0x60/0x60 ret_from_fork+0x25/0x30 Code: Bad RIP value. RIP: (null) RSP: ffffb1e243e4bc38 CR2: 0000000000000000 Fixes: af85389c614a drm/nouveau/disp: shuffle functions around Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103421 Signed-off-by: Rob Clark <robdclark(a)gmail.com> Signed-off-by: Ben Skeggs <bskeggs(a)redhat.com> Cc: Sven Joachim <svenjoac(a)gmx.de> Signed-off-by: Greg Kroah-Hartman <gregkh(a)linuxfoundation.org> --- drivers/gpu/drm/nouveau/nvkm/engine/disp/sorgf119.c | 1 + 1 file changed, 1 insertion(+) --- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/sorgf119.c +++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/sorgf119.c @@ -174,6 +174,7 @@ gf119_sor = { .links = gf119_sor_dp_links, .power = g94_sor_dp_power, .pattern = gf119_sor_dp_pattern, + .drive = gf119_sor_dp_drive, .vcpi = gf119_sor_dp_vcpi, .audio = gf119_sor_dp_audio, .audio_sym = gf119_sor_dp_audio_sym, Patches currently in stable-queue which might be from robdclark(a)gmail.com are queue-4.14/drm-nouveau-disp-gf119-add-missing-drive-vfunc-ptr.patch

7 years, 5 months

1
0
0 0

[Linux-stable-mirror] [PATCH 3.16-stable 0/5] Add PVCLOCK_FIXMAP user mapping

by Juerg Haefliger

Patch 3 fixes the userspace segfaults caused by the PVCLOCK_FIXMAP user mapping (which I've copied from the 3.2 kaiser patchset). I don't claim I fully understand this so the fix might be too broad. Andrea Arcangeli (1): x86/mm/kaiser: remove paravirt clock warning Juerg Haefliger (3): Revert "x86: kvmclock: Disable use from vDSO if KPTI is enabled" x86/kaiser: Add PVCLOCK_FIXMAP user mapping x86/kaiser: Fix segfaults caused by the PVCLOCK_FIXMAP user mapping Marcelo Tosatti (1): kvmclock: export kvmclock clocksource and data pointers arch/x86/include/asm/kvmclock.h | 6 ++++++ arch/x86/kernel/kvmclock.c | 9 +++------ arch/x86/mm/kaiser.c | 12 +++++++++++- 3 files changed, 20 insertions(+), 7 deletions(-) create mode 100644 arch/x86/include/asm/kvmclock.h -- 2.14.1

7 years, 5 months

1
5
0 0

[Linux-stable-mirror] v4.14.14 build: 0 failures 0 warnings (v4.14.14)

by Build bot for Mark Brown

Tree/Branch: v4.14.14 Git describe: v4.14.14 Commit: 9c0bf98471 Linux 4.14.14 Build Time: 108 min 51 sec Passed: 10 / 10 (100.00 %) Failed: 0 / 10 ( 0.00 %) Errors: 0 Warnings: 0 Section Mismatches: 0 ------------------------------------------------------------------------------- defconfigs with issues (other than build errors): ------------------------------------------------------------------------------- =============================================================================== Detailed per-defconfig build reports below: ------------------------------------------------------------------------------- Passed with no errors, warnings or mismatches: arm64-allnoconfig arm64-allmodconfig arm-multi_v5_defconfig arm-multi_v7_defconfig x86_64-defconfig arm-allmodconfig arm-allnoconfig x86_64-allnoconfig arm-multi_v4t_defconfig arm64-defconfig close failed in file object destructor: sys.excepthook is missing lost sys.stderr

7 years, 5 months

1
0
0 0

[Linux-stable-mirror] Patch "libnvdimm, btt: Fix an incompatibility in the log layout" has been added to the 4.9-stable tree

by gregkh＠linuxfoundation.org

This is a note to let you know that I've just added the patch titled libnvdimm, btt: Fix an incompatibility in the log layout to the 4.9-stable tree which can be found at: http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=sum… The filename of the patch is: libnvdimm-btt-fix-an-incompatibility-in-the-log-layout.patch and it can be found in the queue-4.9 subdirectory. If you, or anyone else, feels it should not be added to the stable tree, please let <stable(a)vger.kernel.org> know about it. >From 24e3a7fb60a9187e5df90e5fa655ffc94b9c4f77 Mon Sep 17 00:00:00 2001 From: Vishal Verma <vishal.l.verma(a)intel.com> Date: Mon, 18 Dec 2017 09:28:39 -0700 Subject: libnvdimm, btt: Fix an incompatibility in the log layout From: Vishal Verma <vishal.l.verma(a)intel.com> commit 24e3a7fb60a9187e5df90e5fa655ffc94b9c4f77 upstream. Due to a spec misinterpretation, the Linux implementation of the BTT log area had different padding scheme from other implementations, such as UEFI and NVML. This fixes the padding scheme, and defaults to it for new BTT layouts. We attempt to detect the padding scheme in use when probing for an existing BTT. If we detect the older/incompatible scheme, we continue using it. Reported-by: Juston Li <juston.li(a)intel.com> Cc: Dan Williams <dan.j.williams(a)intel.com> Cc: <stable(a)vger.kernel.org> Fixes: 5212e11fde4d ("nd_btt: atomic sector updates") Signed-off-by: Vishal Verma <vishal.l.verma(a)intel.com> Signed-off-by: Dan Williams <dan.j.williams(a)intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh(a)linuxfoundation.org> --- drivers/nvdimm/btt.c | 203 ++++++++++++++++++++++++++++++++++++++++++--------- drivers/nvdimm/btt.h | 45 +++++++++++ 2 files changed, 212 insertions(+), 36 deletions(-) --- a/drivers/nvdimm/btt.c +++ b/drivers/nvdimm/btt.c @@ -183,13 +183,13 @@ static int btt_map_read(struct arena_inf return ret; } -static int btt_log_read_pair(struct arena_info *arena, u32 lane, - struct log_entry *ent) +static int btt_log_group_read(struct arena_info *arena, u32 lane, + struct log_group *log) { - WARN_ON(!ent); + WARN_ON(!log); return arena_read_bytes(arena, - arena->logoff + (2 * lane * LOG_ENT_SIZE), ent, - 2 * LOG_ENT_SIZE); + arena->logoff + (lane * LOG_GRP_SIZE), log, + LOG_GRP_SIZE); } static struct dentry *debugfs_root; @@ -229,6 +229,8 @@ static void arena_debugfs_init(struct ar debugfs_create_x64("logoff", S_IRUGO, d, &a->logoff); debugfs_create_x64("info2off", S_IRUGO, d, &a->info2off); debugfs_create_x32("flags", S_IRUGO, d, &a->flags); + debugfs_create_u32("log_index_0", S_IRUGO, d, &a->log_index[0]); + debugfs_create_u32("log_index_1", S_IRUGO, d, &a->log_index[1]); } static void btt_debugfs_init(struct btt *btt) @@ -247,6 +249,11 @@ static void btt_debugfs_init(struct btt } } +static u32 log_seq(struct log_group *log, int log_idx) +{ + return le32_to_cpu(log->ent[log_idx].seq); +} + /* * This function accepts two log entries, and uses the * sequence number to find the 'older' entry. @@ -256,8 +263,10 @@ static void btt_debugfs_init(struct btt * * TODO The logic feels a bit kludge-y. make it better.. */ -static int btt_log_get_old(struct log_entry *ent) +static int btt_log_get_old(struct arena_info *a, struct log_group *log) { + int idx0 = a->log_index[0]; + int idx1 = a->log_index[1]; int old; /* @@ -265,23 +274,23 @@ static int btt_log_get_old(struct log_en * the next time, the following logic works out to put this * (next) entry into [1] */ - if (ent[0].seq == 0) { - ent[0].seq = cpu_to_le32(1); + if (log_seq(log, idx0) == 0) { + log->ent[idx0].seq = cpu_to_le32(1); return 0; } - if (ent[0].seq == ent[1].seq) + if (log_seq(log, idx0) == log_seq(log, idx1)) return -EINVAL; - if (le32_to_cpu(ent[0].seq) + le32_to_cpu(ent[1].seq) > 5) + if (log_seq(log, idx0) + log_seq(log, idx1) > 5) return -EINVAL; - if (le32_to_cpu(ent[0].seq) < le32_to_cpu(ent[1].seq)) { - if (le32_to_cpu(ent[1].seq) - le32_to_cpu(ent[0].seq) == 1) + if (log_seq(log, idx0) < log_seq(log, idx1)) { + if ((log_seq(log, idx1) - log_seq(log, idx0)) == 1) old = 0; else old = 1; } else { - if (le32_to_cpu(ent[0].seq) - le32_to_cpu(ent[1].seq) == 1) + if ((log_seq(log, idx0) - log_seq(log, idx1)) == 1) old = 1; else old = 0; @@ -306,17 +315,18 @@ static int btt_log_read(struct arena_inf { int ret; int old_ent, ret_ent; - struct log_entry log[2]; + struct log_group log; - ret = btt_log_read_pair(arena, lane, log); + ret = btt_log_group_read(arena, lane, &log); if (ret) return -EIO; - old_ent = btt_log_get_old(log); + old_ent = btt_log_get_old(arena, &log); if (old_ent < 0 || old_ent > 1) { dev_info(to_dev(arena), "log corruption (%d): lane %d seq [%d, %d]\n", - old_ent, lane, log[0].seq, log[1].seq); + old_ent, lane, log.ent[arena->log_index[0]].seq, + log.ent[arena->log_index[1]].seq); /* TODO set error state? */ return -EIO; } @@ -324,7 +334,7 @@ static int btt_log_read(struct arena_inf ret_ent = (old_flag ? old_ent : (1 - old_ent)); if (ent != NULL) - memcpy(ent, &log[ret_ent], LOG_ENT_SIZE); + memcpy(ent, &log.ent[arena->log_index[ret_ent]], LOG_ENT_SIZE); return ret_ent; } @@ -338,17 +348,13 @@ static int __btt_log_write(struct arena_ u32 sub, struct log_entry *ent) { int ret; - /* - * Ignore the padding in log_entry for calculating log_half. - * The entry is 'committed' when we write the sequence number, - * and we want to ensure that that is the last thing written. - * We don't bother writing the padding as that would be extra - * media wear and write amplification - */ - unsigned int log_half = (LOG_ENT_SIZE - 2 * sizeof(u64)) / 2; - u64 ns_off = arena->logoff + (((2 * lane) + sub) * LOG_ENT_SIZE); + u32 group_slot = arena->log_index[sub]; + unsigned int log_half = LOG_ENT_SIZE / 2; void *src = ent; + u64 ns_off; + ns_off = arena->logoff + (lane * LOG_GRP_SIZE) + + (group_slot * LOG_ENT_SIZE); /* split the 16B write into atomic, durable halves */ ret = arena_write_bytes(arena, ns_off, src, log_half); if (ret) @@ -419,16 +425,16 @@ static int btt_log_init(struct arena_inf { int ret; u32 i; - struct log_entry log, zerolog; + struct log_entry ent, zerolog; memset(&zerolog, 0, sizeof(zerolog)); for (i = 0; i < arena->nfree; i++) { - log.lba = cpu_to_le32(i); - log.old_map = cpu_to_le32(arena->external_nlba + i); - log.new_map = cpu_to_le32(arena->external_nlba + i); - log.seq = cpu_to_le32(LOG_SEQ_INIT); - ret = __btt_log_write(arena, i, 0, &log); + ent.lba = cpu_to_le32(i); + ent.old_map = cpu_to_le32(arena->external_nlba + i); + ent.new_map = cpu_to_le32(arena->external_nlba + i); + ent.seq = cpu_to_le32(LOG_SEQ_INIT); + ret = __btt_log_write(arena, i, 0, &ent); if (ret) return ret; ret = __btt_log_write(arena, i, 1, &zerolog); @@ -490,6 +496,123 @@ static int btt_freelist_init(struct aren return 0; } +static bool ent_is_padding(struct log_entry *ent) +{ + return (ent->lba == 0) && (ent->old_map == 0) && (ent->new_map == 0) + && (ent->seq == 0); +} + +/* + * Detecting valid log indices: We read a log group (see the comments in btt.h + * for a description of a 'log_group' and its 'slots'), and iterate over its + * four slots. We expect that a padding slot will be all-zeroes, and use this + * to detect a padding slot vs. an actual entry. + * + * If a log_group is in the initial state, i.e. hasn't been used since the + * creation of this BTT layout, it will have three of the four slots with + * zeroes. We skip over these log_groups for the detection of log_index. If + * all log_groups are in the initial state (i.e. the BTT has never been + * written to), it is safe to assume the 'new format' of log entries in slots + * (0, 1). + */ +static int log_set_indices(struct arena_info *arena) +{ + bool idx_set = false, initial_state = true; + int ret, log_index[2] = {-1, -1}; + u32 i, j, next_idx = 0; + struct log_group log; + u32 pad_count = 0; + + for (i = 0; i < arena->nfree; i++) { + ret = btt_log_group_read(arena, i, &log); + if (ret < 0) + return ret; + + for (j = 0; j < 4; j++) { + if (!idx_set) { + if (ent_is_padding(&log.ent[j])) { + pad_count++; + continue; + } else { + /* Skip if index has been recorded */ + if ((next_idx == 1) && + (j == log_index[0])) + continue; + /* valid entry, record index */ + log_index[next_idx] = j; + next_idx++; + } + if (next_idx == 2) { + /* two valid entries found */ + idx_set = true; + } else if (next_idx > 2) { + /* too many valid indices */ + return -ENXIO; + } + } else { + /* + * once the indices have been set, just verify + * that all subsequent log groups are either in + * their initial state or follow the same + * indices. + */ + if (j == log_index[0]) { + /* entry must be 'valid' */ + if (ent_is_padding(&log.ent[j])) + return -ENXIO; + } else if (j == log_index[1]) { + ; + /* + * log_index[1] can be padding if the + * lane never got used and it is still + * in the initial state (three 'padding' + * entries) + */ + } else { + /* entry must be invalid (padding) */ + if (!ent_is_padding(&log.ent[j])) + return -ENXIO; + } + } + } + /* + * If any of the log_groups have more than one valid, + * non-padding entry, then the we are no longer in the + * initial_state + */ + if (pad_count < 3) + initial_state = false; + pad_count = 0; + } + + if (!initial_state && !idx_set) + return -ENXIO; + + /* + * If all the entries in the log were in the initial state, + * assume new padding scheme + */ + if (initial_state) + log_index[1] = 1; + + /* + * Only allow the known permutations of log/padding indices, + * i.e. (0, 1), and (0, 2) + */ + if ((log_index[0] == 0) && ((log_index[1] == 1) || (log_index[1] == 2))) + ; /* known index possibilities */ + else { + dev_err(to_dev(arena), "Found an unknown padding scheme\n"); + return -ENXIO; + } + + arena->log_index[0] = log_index[0]; + arena->log_index[1] = log_index[1]; + dev_dbg(to_dev(arena), "log_index_0 = %d\n", log_index[0]); + dev_dbg(to_dev(arena), "log_index_1 = %d\n", log_index[1]); + return 0; +} + static int btt_rtt_init(struct arena_info *arena) { arena->rtt = kcalloc(arena->nfree, sizeof(u32), GFP_KERNEL); @@ -545,8 +668,7 @@ static struct arena_info *alloc_arena(st available -= 2 * BTT_PG_SIZE; /* The log takes a fixed amount of space based on nfree */ - logsize = roundup(2 * arena->nfree * sizeof(struct log_entry), - BTT_PG_SIZE); + logsize = roundup(arena->nfree * LOG_GRP_SIZE, BTT_PG_SIZE); available -= logsize; /* Calculate optimal split between map and data area */ @@ -563,6 +685,10 @@ static struct arena_info *alloc_arena(st arena->mapoff = arena->dataoff + datasize; arena->logoff = arena->mapoff + mapsize; arena->info2off = arena->logoff + logsize; + + /* Default log indices are (0,1) */ + arena->log_index[0] = 0; + arena->log_index[1] = 1; return arena; } @@ -653,6 +779,13 @@ static int discover_arenas(struct btt *b arena->external_lba_start = cur_nlba; parse_arena_meta(arena, super, cur_off); + ret = log_set_indices(arena); + if (ret) { + dev_err(to_dev(arena), + "Unable to deduce log/padding indices\n"); + goto out; + } + ret = btt_freelist_init(arena); if (ret) goto out; --- a/drivers/nvdimm/btt.h +++ b/drivers/nvdimm/btt.h @@ -26,6 +26,7 @@ #define MAP_ERR_MASK (1 << MAP_ERR_SHIFT) #define MAP_LBA_MASK (~((1 << MAP_TRIM_SHIFT) | (1 << MAP_ERR_SHIFT))) #define MAP_ENT_NORMAL 0xC0000000 +#define LOG_GRP_SIZE sizeof(struct log_group) #define LOG_ENT_SIZE sizeof(struct log_entry) #define ARENA_MIN_SIZE (1UL << 24) /* 16 MB */ #define ARENA_MAX_SIZE (1ULL << 39) /* 512 GB */ @@ -44,12 +45,52 @@ enum btt_init_state { INIT_READY }; +/* + * A log group represents one log 'lane', and consists of four log entries. + * Two of the four entries are valid entries, and the remaining two are + * padding. Due to an old bug in the padding location, we need to perform a + * test to determine the padding scheme being used, and use that scheme + * thereafter. + * + * In kernels prior to 4.15, 'log group' would have actual log entries at + * indices (0, 2) and padding at indices (1, 3), where as the correct/updated + * format has log entries at indices (0, 1) and padding at indices (2, 3). + * + * Old (pre 4.15) format: + * +-----------------+-----------------+ + * | ent[0] | ent[1] | + * | 16B | 16B | + * | lba/old/new/seq | pad | + * +-----------------------------------+ + * | ent[2] | ent[3] | + * | 16B | 16B | + * | lba/old/new/seq | pad | + * +-----------------+-----------------+ + * + * New format: + * +-----------------+-----------------+ + * | ent[0] | ent[1] | + * | 16B | 16B | + * | lba/old/new/seq | lba/old/new/seq | + * +-----------------------------------+ + * | ent[2] | ent[3] | + * | 16B | 16B | + * | pad | pad | + * +-----------------+-----------------+ + * + * We detect during start-up which format is in use, and set + * arena->log_index[(0, 1)] with the detected format. + */ + struct log_entry { __le32 lba; __le32 old_map; __le32 new_map; __le32 seq; - __le64 padding[2]; +}; + +struct log_group { + struct log_entry ent[4]; }; struct btt_sb { @@ -117,6 +158,7 @@ struct aligned_lock { * @list: List head for list of arenas * @debugfs_dir: Debugfs dentry * @flags: Arena flags - may signify error states. + * @log_index: Indices of the valid log entries in a log_group * * arena_info is a per-arena handle. Once an arena is narrowed down for an * IO, this struct is passed around for the duration of the IO. @@ -147,6 +189,7 @@ struct arena_info { struct dentry *debugfs_dir; /* Arena flags */ u32 flags; + int log_index[2]; }; /** Patches currently in stable-queue which might be from vishal.l.verma(a)intel.com are queue-4.9/libnvdimm-btt-fix-an-incompatibility-in-the-log-layout.patch

7 years, 5 months

1
0
0 0

2025

2024

2023

2022

2021

2020

2019

2018

2017

Linux-stable-mirror