]> git.ipfire.org Git - thirdparty/kernel/stable-queue.git/commitdiff
4.14-stable patches
authorGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sat, 2 Mar 2019 08:36:36 +0000 (09:36 +0100)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sat, 2 Mar 2019 08:36:36 +0000 (09:36 +0100)
added patches:
mips-fix-truncation-in-__cmpxchg_small-for-short-values.patch
mm-enforce-min-addr-even-if-capable-in-expand_downwards.patch
mmc-sdhci-esdhc-imx-correct-the-fix-of-err004536.patch
mmc-spi-fix-card-detection-during-probe.patch
mmc-tmio-fix-access-width-of-block-count-register.patch
mmc-tmio_mmc_core-don-t-claim-spurious-interrupts.patch

queue-4.14/mips-fix-truncation-in-__cmpxchg_small-for-short-values.patch [new file with mode: 0644]
queue-4.14/mm-enforce-min-addr-even-if-capable-in-expand_downwards.patch [new file with mode: 0644]
queue-4.14/mmc-sdhci-esdhc-imx-correct-the-fix-of-err004536.patch [new file with mode: 0644]
queue-4.14/mmc-spi-fix-card-detection-during-probe.patch [new file with mode: 0644]
queue-4.14/mmc-tmio-fix-access-width-of-block-count-register.patch [new file with mode: 0644]
queue-4.14/mmc-tmio_mmc_core-don-t-claim-spurious-interrupts.patch [new file with mode: 0644]
queue-4.14/series

diff --git a/queue-4.14/mips-fix-truncation-in-__cmpxchg_small-for-short-values.patch b/queue-4.14/mips-fix-truncation-in-__cmpxchg_small-for-short-values.patch
new file mode 100644 (file)
index 0000000..5cb592e
--- /dev/null
@@ -0,0 +1,48 @@
+From 94ee12b507db8b5876e31c9d6c9d84f556a4b49f Mon Sep 17 00:00:00 2001
+From: Michael Clark <michaeljclark@mac.com>
+Date: Mon, 11 Feb 2019 17:38:29 +1300
+Subject: MIPS: fix truncation in __cmpxchg_small for short values
+
+From: Michael Clark <michaeljclark@mac.com>
+
+commit 94ee12b507db8b5876e31c9d6c9d84f556a4b49f upstream.
+
+__cmpxchg_small erroneously uses u8 for load comparison which can
+be either char or short. This patch changes the local variable to
+u32 which is sufficiently sized, as the loaded value is already
+masked and shifted appropriately. Using an integer size avoids
+any unnecessary canonicalization from use of non native widths.
+
+This patch is part of a series that adapts the MIPS small word
+atomics code for xchg and cmpxchg on short and char to RISC-V.
+
+Cc: RISC-V Patches <patches@groups.riscv.org>
+Cc: Linux RISC-V <linux-riscv@lists.infradead.org>
+Cc: Linux MIPS <linux-mips@linux-mips.org>
+Signed-off-by: Michael Clark <michaeljclark@mac.com>
+[paul.burton@mips.com:
+  - Fix varialble typo per Jonas Gorski.
+  - Consolidate load variable with other declarations.]
+Signed-off-by: Paul Burton <paul.burton@mips.com>
+Fixes: 3ba7f44d2b19 ("MIPS: cmpxchg: Implement 1 byte & 2 byte cmpxchg()")
+Cc: stable@vger.kernel.org # v4.13+
+Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
+
+---
+ arch/mips/kernel/cmpxchg.c |    3 +--
+ 1 file changed, 1 insertion(+), 2 deletions(-)
+
+--- a/arch/mips/kernel/cmpxchg.c
++++ b/arch/mips/kernel/cmpxchg.c
+@@ -54,10 +54,9 @@ unsigned long __xchg_small(volatile void
+ unsigned long __cmpxchg_small(volatile void *ptr, unsigned long old,
+                             unsigned long new, unsigned int size)
+ {
+-      u32 mask, old32, new32, load32;
++      u32 mask, old32, new32, load32, load;
+       volatile u32 *ptr32;
+       unsigned int shift;
+-      u8 load;
+       /* Check that ptr is naturally aligned */
+       WARN_ON((unsigned long)ptr & (size - 1));
diff --git a/queue-4.14/mm-enforce-min-addr-even-if-capable-in-expand_downwards.patch b/queue-4.14/mm-enforce-min-addr-even-if-capable-in-expand_downwards.patch
new file mode 100644 (file)
index 0000000..d0ac24d
--- /dev/null
@@ -0,0 +1,44 @@
+From 0a1d52994d440e21def1c2174932410b4f2a98a1 Mon Sep 17 00:00:00 2001
+From: Jann Horn <jannh@google.com>
+Date: Wed, 27 Feb 2019 21:29:52 +0100
+Subject: mm: enforce min addr even if capable() in expand_downwards()
+
+From: Jann Horn <jannh@google.com>
+
+commit 0a1d52994d440e21def1c2174932410b4f2a98a1 upstream.
+
+security_mmap_addr() does a capability check with current_cred(), but
+we can reach this code from contexts like a VFS write handler where
+current_cred() must not be used.
+
+This can be abused on systems without SMAP to make NULL pointer
+dereferences exploitable again.
+
+Fixes: 8869477a49c3 ("security: protect from stack expansion into low vm addresses")
+Cc: stable@kernel.org
+Signed-off-by: Jann Horn <jannh@google.com>
+Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
+Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
+
+---
+ mm/mmap.c |    7 +++----
+ 1 file changed, 3 insertions(+), 4 deletions(-)
+
+--- a/mm/mmap.c
++++ b/mm/mmap.c
+@@ -2348,12 +2348,11 @@ int expand_downwards(struct vm_area_stru
+ {
+       struct mm_struct *mm = vma->vm_mm;
+       struct vm_area_struct *prev;
+-      int error;
++      int error = 0;
+       address &= PAGE_MASK;
+-      error = security_mmap_addr(address);
+-      if (error)
+-              return error;
++      if (address < mmap_min_addr)
++              return -EPERM;
+       /* Enforce stack_guard_gap */
+       prev = vma->vm_prev;
diff --git a/queue-4.14/mmc-sdhci-esdhc-imx-correct-the-fix-of-err004536.patch b/queue-4.14/mmc-sdhci-esdhc-imx-correct-the-fix-of-err004536.patch
new file mode 100644 (file)
index 0000000..94f0269
--- /dev/null
@@ -0,0 +1,83 @@
+From e30be063d6dbcc0f18b1eb25fa709fdef89201fb Mon Sep 17 00:00:00 2001
+From: BOUGH CHEN <haibo.chen@nxp.com>
+Date: Thu, 28 Feb 2019 10:15:42 +0000
+Subject: mmc: sdhci-esdhc-imx: correct the fix of ERR004536
+
+From: BOUGH CHEN <haibo.chen@nxp.com>
+
+commit e30be063d6dbcc0f18b1eb25fa709fdef89201fb upstream.
+
+Commit 18094430d6b5 ("mmc: sdhci-esdhc-imx: add ADMA Length
+Mismatch errata fix") involve the fix of ERR004536, but the
+fix is incorrect. Double confirm with IC, need to clear the
+bit 7 of register 0x6c rather than set this bit 7.
+Here is the definition of bit 7 of 0x6c:
+    0: enable the new IC fix for ERR004536
+    1: do not use the IC fix, keep the same as before
+
+Find this issue on i.MX845s-evk board when enable CMDQ, and
+let system in heavy loading.
+
+root@imx8mmevk:~# dd if=/dev/mmcblk2 of=/dev/null bs=1M &
+root@imx8mmevk:~# memtester 1000M > /dev/zero &
+root@imx8mmevk:~# [  139.897220] mmc2: cqhci: timeout for tag 16
+[  139.901417] mmc2: cqhci: ============ CQHCI REGISTER DUMP ===========
+[  139.907862] mmc2: cqhci: Caps:      0x0000310a | Version:  0x00000510
+[  139.914311] mmc2: cqhci: Config:    0x00001001 | Control:  0x00000000
+[  139.920753] mmc2: cqhci: Int stat:  0x00000000 | Int enab: 0x00000006
+[  139.927193] mmc2: cqhci: Int sig:   0x00000006 | Int Coal: 0x00000000
+[  139.933634] mmc2: cqhci: TDL base:  0x7809c000 | TDL up32: 0x00000000
+[  139.940073] mmc2: cqhci: Doorbell:  0x00030000 | TCN:      0x00000000
+[  139.946518] mmc2: cqhci: Dev queue: 0x00010000 | Dev Pend: 0x00010000
+[  139.952967] mmc2: cqhci: Task clr:  0x00000000 | SSC1:     0x00011000
+[  139.959411] mmc2: cqhci: SSC2:      0x00000001 | DCMD rsp: 0x00000000
+[  139.965857] mmc2: cqhci: RED mask:  0xfdf9a080 | TERRI:    0x00000000
+[  139.972308] mmc2: cqhci: Resp idx:  0x0000002e | Resp arg: 0x00000900
+[  139.978761] mmc2: sdhci: ============ SDHCI REGISTER DUMP ===========
+[  139.985214] mmc2: sdhci: Sys addr:  0xb2c19000 | Version:  0x00000002
+[  139.991669] mmc2: sdhci: Blk size:  0x00000200 | Blk cnt:  0x00000400
+[  139.998127] mmc2: sdhci: Argument:  0x40110400 | Trn mode: 0x00000033
+[  140.004618] mmc2: sdhci: Present:   0x01088a8f | Host ctl: 0x00000030
+[  140.011113] mmc2: sdhci: Power:     0x00000002 | Blk gap:  0x00000080
+[  140.017583] mmc2: sdhci: Wake-up:   0x00000008 | Clock:    0x0000000f
+[  140.024039] mmc2: sdhci: Timeout:   0x0000008f | Int stat: 0x00000000
+[  140.030497] mmc2: sdhci: Int enab:  0x107f4000 | Sig enab: 0x107f4000
+[  140.036972] mmc2: sdhci: AC12 err:  0x00000000 | Slot int: 0x00000502
+[  140.043426] mmc2: sdhci: Caps:      0x07eb0000 | Caps_1:   0x8000b407
+[  140.049867] mmc2: sdhci: Cmd:       0x00002c1a | Max curr: 0x00ffffff
+[  140.056314] mmc2: sdhci: Resp[0]:   0x00000900 | Resp[1]:  0xffffffff
+[  140.062755] mmc2: sdhci: Resp[2]:   0x328f5903 | Resp[3]:  0x00d00f00
+[  140.069195] mmc2: sdhci: Host ctl2: 0x00000008
+[  140.073640] mmc2: sdhci: ADMA Err:  0x00000007 | ADMA Ptr: 0x7809c108
+[  140.080079] mmc2: sdhci: ============================================
+[  140.086662] mmc2: running CQE recovery
+
+Fixes: 18094430d6b5 ("mmc: sdhci-esdhc-imx: add ADMA Length Mismatch errata fix")
+Signed-off-by: Haibo Chen <haibo.chen@nxp.com>
+Cc: stable@vger.kernel.org
+Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
+Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
+
+---
+ drivers/mmc/host/sdhci-esdhc-imx.c |    9 +++++----
+ 1 file changed, 5 insertions(+), 4 deletions(-)
+
+--- a/drivers/mmc/host/sdhci-esdhc-imx.c
++++ b/drivers/mmc/host/sdhci-esdhc-imx.c
+@@ -1077,11 +1077,12 @@ static void sdhci_esdhc_imx_hwinit(struc
+               writel(readl(host->ioaddr + SDHCI_HOST_CONTROL)
+                       | ESDHC_BURST_LEN_EN_INCR,
+                       host->ioaddr + SDHCI_HOST_CONTROL);
++
+               /*
+-              * erratum ESDHC_FLAG_ERR004536 fix for MX6Q TO1.2 and MX6DL
+-              * TO1.1, it's harmless for MX6SL
+-              */
+-              writel(readl(host->ioaddr + 0x6c) | BIT(7),
++               * erratum ESDHC_FLAG_ERR004536 fix for MX6Q TO1.2 and MX6DL
++               * TO1.1, it's harmless for MX6SL
++               */
++              writel(readl(host->ioaddr + 0x6c) & ~BIT(7),
+                       host->ioaddr + 0x6c);
+               /* disable DLL_CTRL delay line settings */
diff --git a/queue-4.14/mmc-spi-fix-card-detection-during-probe.patch b/queue-4.14/mmc-spi-fix-card-detection-during-probe.patch
new file mode 100644 (file)
index 0000000..817d1c7
--- /dev/null
@@ -0,0 +1,55 @@
+From c9bd505dbd9d3dc80c496f88eafe70affdcf1ba6 Mon Sep 17 00:00:00 2001
+From: =?UTF-8?q?Jonathan=20Neusch=C3=A4fer?= <j.neuschaefer@gmx.net>
+Date: Sun, 10 Feb 2019 18:31:07 +0100
+Subject: mmc: spi: Fix card detection during probe
+MIME-Version: 1.0
+Content-Type: text/plain; charset=UTF-8
+Content-Transfer-Encoding: 8bit
+
+From: Jonathan Neuschäfer <j.neuschaefer@gmx.net>
+
+commit c9bd505dbd9d3dc80c496f88eafe70affdcf1ba6 upstream.
+
+When using the mmc_spi driver with a card-detect pin, I noticed that the
+card was not detected immediately after probe, but only after it was
+unplugged and plugged back in (and the CD IRQ fired).
+
+The call tree looks something like this:
+
+mmc_spi_probe
+  mmc_add_host
+    mmc_start_host
+      _mmc_detect_change
+        mmc_schedule_delayed_work(&host->detect, 0)
+          mmc_rescan
+            host->bus_ops->detect(host)
+              mmc_detect
+                _mmc_detect_card_removed
+                  host->ops->get_cd(host)
+                    mmc_gpio_get_cd -> -ENOSYS (ctx->cd_gpio not set)
+  mmc_gpiod_request_cd
+    ctx->cd_gpio = desc
+
+To fix this issue, call mmc_detect_change after the card-detect GPIO/IRQ
+is registered.
+
+Signed-off-by: Jonathan Neuschäfer <j.neuschaefer@gmx.net>
+Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
+Cc: stable@vger.kernel.org
+Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
+Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
+
+---
+ drivers/mmc/host/mmc_spi.c |    1 +
+ 1 file changed, 1 insertion(+)
+
+--- a/drivers/mmc/host/mmc_spi.c
++++ b/drivers/mmc/host/mmc_spi.c
+@@ -1447,6 +1447,7 @@ static int mmc_spi_probe(struct spi_devi
+               mmc->caps &= ~MMC_CAP_NEEDS_POLL;
+               mmc_gpiod_request_cd_irq(mmc);
+       }
++      mmc_detect_change(mmc, 0);
+       if (host->pdata && host->pdata->flags & MMC_SPI_USE_RO_GPIO) {
+               has_ro = true;
diff --git a/queue-4.14/mmc-tmio-fix-access-width-of-block-count-register.patch b/queue-4.14/mmc-tmio-fix-access-width-of-block-count-register.patch
new file mode 100644 (file)
index 0000000..1de30c7
--- /dev/null
@@ -0,0 +1,80 @@
+From 5603731a15ef9ca317c122cc8c959f1dee1798b4 Mon Sep 17 00:00:00 2001
+From: Takeshi Saito <takeshi.saito.xv@renesas.com>
+Date: Thu, 21 Feb 2019 20:38:05 +0100
+Subject: mmc: tmio: fix access width of Block Count Register
+
+From: Takeshi Saito <takeshi.saito.xv@renesas.com>
+
+commit 5603731a15ef9ca317c122cc8c959f1dee1798b4 upstream.
+
+In R-Car Gen2 or later, the maximum number of transfer blocks are
+changed from 0xFFFF to 0xFFFFFFFF. Therefore, Block Count Register
+should use iowrite32().
+
+If another system (U-boot, Hypervisor OS, etc) uses bit[31:16], this
+value will not be cleared. So, SD/MMC card initialization fails.
+
+So, check for the bigger register and use apropriate write. Also, mark
+the register as extended on Gen2.
+
+Signed-off-by: Takeshi Saito <takeshi.saito.xv@renesas.com>
+[wsa: use max_blk_count in if(), add Gen2, update commit message]
+Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
+Cc: stable@kernel.org
+Reviewed-by: Simon Horman <horms+renesas@verge.net.au>
+[Ulf: Fixed build error]
+Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
+Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
+
+---
+ drivers/mmc/host/renesas_sdhi_sys_dmac.c |    1 +
+ drivers/mmc/host/tmio_mmc.h              |    5 +++++
+ drivers/mmc/host/tmio_mmc_core.c         |    6 +++++-
+ 3 files changed, 11 insertions(+), 1 deletion(-)
+
+--- a/drivers/mmc/host/renesas_sdhi_sys_dmac.c
++++ b/drivers/mmc/host/renesas_sdhi_sys_dmac.c
+@@ -68,6 +68,7 @@ static const struct renesas_sdhi_of_data
+       .scc_offset     = 0x0300,
+       .taps           = rcar_gen2_scc_taps,
+       .taps_num       = ARRAY_SIZE(rcar_gen2_scc_taps),
++      .max_blk_count  = 0xffffffff,
+ };
+ /* Definitions for sampling clocks */
+--- a/drivers/mmc/host/tmio_mmc.h
++++ b/drivers/mmc/host/tmio_mmc.h
+@@ -286,6 +286,11 @@ static inline void sd_ctrl_write32_as_16
+       writew(val >> 16, host->ctl + ((addr + 2) << host->bus_shift));
+ }
++static inline void sd_ctrl_write32(struct tmio_mmc_host *host, int addr, u32 val)
++{
++      iowrite32(val, host->ctl + (addr << host->bus_shift));
++}
++
+ static inline void sd_ctrl_write32_rep(struct tmio_mmc_host *host, int addr,
+                                      const u32 *buf, int count)
+ {
+--- a/drivers/mmc/host/tmio_mmc_core.c
++++ b/drivers/mmc/host/tmio_mmc_core.c
+@@ -46,6 +46,7 @@
+ #include <linux/regulator/consumer.h>
+ #include <linux/mmc/sdio.h>
+ #include <linux/scatterlist.h>
++#include <linux/sizes.h>
+ #include <linux/spinlock.h>
+ #include <linux/swiotlb.h>
+ #include <linux/workqueue.h>
+@@ -761,7 +762,10 @@ static int tmio_mmc_start_data(struct tm
+       /* Set transfer length / blocksize */
+       sd_ctrl_write16(host, CTL_SD_XFER_LEN, data->blksz);
+-      sd_ctrl_write16(host, CTL_XFER_BLK_COUNT, data->blocks);
++      if (host->mmc->max_blk_count >= SZ_64K)
++              sd_ctrl_write32(host, CTL_XFER_BLK_COUNT, data->blocks);
++      else
++              sd_ctrl_write16(host, CTL_XFER_BLK_COUNT, data->blocks);
+       tmio_mmc_start_dma(host, data);
diff --git a/queue-4.14/mmc-tmio_mmc_core-don-t-claim-spurious-interrupts.patch b/queue-4.14/mmc-tmio_mmc_core-don-t-claim-spurious-interrupts.patch
new file mode 100644 (file)
index 0000000..751235e
--- /dev/null
@@ -0,0 +1,73 @@
+From 5c27ff5db1491a947264d6d4e4cbe43ae6535bae Mon Sep 17 00:00:00 2001
+From: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
+Date: Mon, 18 Feb 2019 20:45:40 +0300
+Subject: mmc: tmio_mmc_core: don't claim spurious interrupts
+
+From: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
+
+commit 5c27ff5db1491a947264d6d4e4cbe43ae6535bae upstream.
+
+I have encountered an interrupt storm during the eMMC chip probing (and
+the chip finally didn't get detected).  It turned out that U-Boot left
+the DMAC interrupts enabled while the Linux driver  didn't use those.
+The SDHI driver's interrupt handler somehow assumes that, even if an
+SDIO interrupt didn't happen, it should return IRQ_HANDLED.  I think
+that if none of the enabled interrupts happened and got handled, we
+should return IRQ_NONE -- that way the kernel IRQ code recoginizes
+a spurious interrupt and masks it off pretty quickly...
+
+Fixes: 7729c7a232a9 ("mmc: tmio: Provide separate interrupt handlers")
+Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
+Reviewed-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
+Tested-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
+Reviewed-by: Simon Horman <horms+renesas@verge.net.au>
+Cc: stable@vger.kernel.org
+Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
+Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
+
+---
+ drivers/mmc/host/tmio_mmc_core.c |   11 +++++++----
+ 1 file changed, 7 insertions(+), 4 deletions(-)
+
+--- a/drivers/mmc/host/tmio_mmc_core.c
++++ b/drivers/mmc/host/tmio_mmc_core.c
+@@ -688,7 +688,7 @@ static bool __tmio_mmc_sdcard_irq(struct
+       return false;
+ }
+-static void __tmio_mmc_sdio_irq(struct tmio_mmc_host *host)
++static bool __tmio_mmc_sdio_irq(struct tmio_mmc_host *host)
+ {
+       struct mmc_host *mmc = host->mmc;
+       struct tmio_mmc_data *pdata = host->pdata;
+@@ -696,7 +696,7 @@ static void __tmio_mmc_sdio_irq(struct t
+       unsigned int sdio_status;
+       if (!(pdata->flags & TMIO_MMC_SDIO_IRQ))
+-              return;
++              return false;
+       status = sd_ctrl_read16(host, CTL_SDIO_STATUS);
+       ireg = status & TMIO_SDIO_MASK_ALL & ~host->sdio_irq_mask;
+@@ -709,6 +709,8 @@ static void __tmio_mmc_sdio_irq(struct t
+       if (mmc->caps & MMC_CAP_SDIO_IRQ && ireg & TMIO_SDIO_STAT_IOIRQ)
+               mmc_signal_sdio_irq(mmc);
++
++      return ireg;
+ }
+ irqreturn_t tmio_mmc_irq(int irq, void *devid)
+@@ -727,9 +729,10 @@ irqreturn_t tmio_mmc_irq(int irq, void *
+       if (__tmio_mmc_sdcard_irq(host, ireg, status))
+               return IRQ_HANDLED;
+-      __tmio_mmc_sdio_irq(host);
++      if (__tmio_mmc_sdio_irq(host))
++              return IRQ_HANDLED;
+-      return IRQ_HANDLED;
++      return IRQ_NONE;
+ }
+ EXPORT_SYMBOL_GPL(tmio_mmc_irq);
index 7c61f40d2bba54b7a5a0977c3544b813fba9837c..9cf1d311499538a8bde94587154c163f9fc76128 100644 (file)
@@ -42,3 +42,9 @@ cfg80211-extend-range-deviation-for-dmg.patch
 svm-fix-avic-incomplete-ipi-emulation.patch
 kvm-nsvm-clear-events-pending-from-svm_complete_inte.patch
 powerpc-always-initialize-input-array-when-calling-epapr_hypercall.patch
+mmc-spi-fix-card-detection-during-probe.patch
+mmc-tmio_mmc_core-don-t-claim-spurious-interrupts.patch
+mmc-tmio-fix-access-width-of-block-count-register.patch
+mmc-sdhci-esdhc-imx-correct-the-fix-of-err004536.patch
+mm-enforce-min-addr-even-if-capable-in-expand_downwards.patch
+mips-fix-truncation-in-__cmpxchg_small-for-short-values.patch