- 03 6月, 2015 4 次提交
-
-
由 Aravind Gopalakrishnan 提交于
Use strings such as "hw" or "sw" to indicate the type of error injection to be performed. Current flags attribute derives the meanings of values that can be programmed into it from asm/mce.h. Moving to defined strings for the attribute allows this module to be self-sufficient and removes the dependency. Also, we can introduce new flags as and when needed without having to worry about conflicting with the flags already defined in asm/mce.h. Also, modify do_inject() to use the newly defined injection_type enum to figure out the injection mechanism we need to use Suggested-by: NBorislav Petkov <bp@suse.de> Signed-off-by: NAravind Gopalakrishnan <Aravind.Gopalakrishnan@amd.com> Cc: linux-edac <linux-edac@vger.kernel.org> Cc: mchehab@osg.samsung.com Cc: x86-ml <x86@kernel.org> Link: http://lkml.kernel.org/r/1433277362-10911-4-git-send-email-Aravind.Gopalakrishnan@amd.com [ Use strstrip() return value. ] Signed-off-by: NBorislav Petkov <bp@suse.de>
-
由 Aravind Gopalakrishnan 提交于
The number of banks for a given processor is encoded in MSR_IA32_MCG_CAP[7:0]. So obtain the value from that MSR and use it for sanity checking in inj_bank_set() instead of doing a family/model check. Signed-off-by: NAravind Gopalakrishnan <Aravind.Gopalakrishnan@amd.com> Cc: linux-edac <linux-edac@vger.kernel.org> Cc: mchehab@osg.samsung.com Link: http://lkml.kernel.org/r/1432753418-2985-3-git-send-email-Aravind.Gopalakrishnan@amd.comSigned-off-by: NBorislav Petkov <bp@suse.de>
-
由 Aravind Gopalakrishnan 提交于
inj_bank_get() is generic enough that we can use the MCE_INJECT_GET macro instead. No functionality change. Signed-off-by: NAravind Gopalakrishnan <Aravind.Gopalakrishnan@amd.com> Cc: linux-edac <linux-edac@vger.kernel.org> Cc: mchehab@osg.samsung.com Cc: x86-ml <x86@kernel.org> Link: http://lkml.kernel.org/r/1433277362-10911-2-git-send-email-Aravind.Gopalakrishnan@amd.comSigned-off-by: NBorislav Petkov <bp@suse.de>
-
由 Arnd Bergmann 提交于
The new x-gene EDAC driver incorrectly tried to figure out the version of one of its IP blocks by looking at the version of the CPU core, which is only vagely related. This removes the incorrect code and instead uses the version of the IP block in the compatible string where it belongs. Found using build testing on x86, which does not provide the arm64 cpuid interface. Signed-off-by: NArnd Bergmann <arnd@arndb.de> [ Changed subnode to "apm,xgene-edac-pmd-v2", adjusted check. ] Signed-off-by: NLoc Ho <lho@apm.com> Cc: devicetree@vger.kernel.org Cc: dougthompson@xmission.com Cc: ijc+devicetree@hellion.org.uk Cc: jcm@redhat.com Cc: linux-arm-kernel@lists.infradead.org Cc: linux-edac <linux-edac@vger.kernel.org> Cc: mark.rutland@arm.com Cc: mchehab@osg.samsung.com Cc: patches@apm.com Cc: robh+dt@kernel.org Link: http://lkml.kernel.org/r/3195065.IK73o60xya@wuerfelSigned-off-by: NBorislav Petkov <bp@suse.de>
-
- 31 5月, 2015 2 次提交
-
-
由 York Sun 提交于
Extend err_addr to cover 64 bits for DDR errors. Signed-off-by: NYork Sun <yorksun@freescale.com> Acked-by: NJohannes Thumshirn <morbidrsa@gmail.com> Cc: Mingkai.hu@freescale.com Link: http://lkml.kernel.org/r/1431425022-44766-2-git-send-email-Wenbin.Song@freescale.comSigned-off-by: Nsongwenbin <wenbin.song@freescale.com> Signed-off-by: NBorislav Petkov <bp@suse.de>
-
由 York Sun 提交于
Remove mpc83xx and mpc85xx as dependency. Signed-off-by: NYork Sun <yorksun@freescale.com> Acked-by: NJohannes Thumshirn <morbidrsa@gmail.com> Cc: Mingkai.hu@freescale.com Link: http://lkml.kernel.org/r/1431425022-44766-1-git-send-email-Wenbin.Song@freescale.comSigned-off-by: Nsongwenbin <wenbin.song@freescale.com> Signed-off-by: NBorislav Petkov <bp@suse.de>
-
- 30 5月, 2015 1 次提交
-
-
由 Borislav Petkov 提交于
<asm/edac.h> contains only the arch-specific scrubbing function and is thus not needed in edac_stub.c. Kill it. Signed-off-by: NBorislav Petkov <bp@suse.de>
-
- 29 5月, 2015 1 次提交
-
-
由 Loc Ho 提交于
Add support for the APM X-Gene SoC EDAC driver. Signed-off-by: NLoc Ho <lho@apm.com> Acked-by: NArnd Bergmann <arnd@arndb.de> Cc: devicetree@vger.kernel.org Cc: dougthompson@xmission.com Cc: ijc+devicetree@hellion.org.uk Cc: jcm@redhat.com Cc: linux-arm-kernel@lists.infradead.org Cc: linux-edac <linux-edac@vger.kernel.org> Cc: mark.rutland@arm.com Cc: mchehab@osg.samsung.com Cc: patches@apm.com Cc: robh+dt@kernel.org Link: http://lkml.kernel.org/r/1432337580-3750-5-git-send-email-lho@apm.comSigned-off-by: NBorislav Petkov <bp@suse.de>
-
- 28 5月, 2015 1 次提交
-
-
由 Borislav Petkov 提交于
So first of all, this atomic_scrub() function's naming is bad. It looks like an atomic_t helper. Change it to edac_atomic_scrub(). The bigger problem is that this function is arch-specific and every new arch which doesn't necessarily need that functionality still needs to define it, otherwise EDAC doesn't compile. So instead of doing that and including arch-specific headers, have each arch define an EDAC_ATOMIC_SCRUB symbol which can be used in edac_mc.c for ifdeffery. Much cleaner. And we already are doing this with another symbol - EDAC_SUPPORT. This is also much cleaner than having CONFIG_EDAC enumerate all the arches which need/have EDAC support and drivers. This way I can kill the useless edac.h header in tile too. Acked-by: NRalf Baechle <ralf@linux-mips.org> Acked-by: NMichael Ellerman <mpe@ellerman.id.au> Acked-by: NChris Metcalf <cmetcalf@ezchip.com> Acked-by: NIngo Molnar <mingo@kernel.org> Acked-by: NRussell King <rmk+kernel@arm.linux.org.uk> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Doug Thompson <dougthompson@xmission.com> Cc: linux-arm-kernel@lists.infradead.org Cc: linux-edac@vger.kernel.org Cc: linux-kernel@vger.kernel.org Cc: linux-mips@linux-mips.org Cc: linuxppc-dev@lists.ozlabs.org Cc: "Maciej W. Rozycki" <macro@codesourcery.com> Cc: Markos Chandras <markos.chandras@imgtec.com> Cc: Mauro Carvalho Chehab <mchehab@osg.samsung.com> Cc: Paul Mackerras <paulus@samba.org> Cc: "Steven J. Hill" <Steven.Hill@imgtec.com> Cc: x86@kernel.org Signed-off-by: NBorislav Petkov <bp@suse.de>
-
- 03 5月, 2015 1 次提交
-
-
由 Thor Thayer 提交于
The SDRAM EDAC requires SDRAM configuration/initialization before SDRAM is accessed (in the preloader) and therefore before Linux is loaded. Having a module compile is not desired so force to be built into kernel. Signed-off-by: NThor Thayer <tthayer@opensource.altera.com> Cc: mchehab@osg.samsung.com Cc: Takashi Iwai <tiwai@suse.de> Link: http://lkml.kernel.org/r/1429308974-26380-3-git-send-email-tthayer@opensource.altera.comSigned-off-by: NBorislav Petkov <bp@suse.de>
-
- 27 4月, 2015 1 次提交
-
-
由 Stephen Rothwell 提交于
Signed-off-by: NStephen Rothwell <sfr@canb.auug.org.au> Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>
-
- 25 4月, 2015 2 次提交
-
-
由 Dmitry Torokhov 提交于
The new Atmel MXT driver expects i2c client's address contain the primary (main address) of the chip, and calculates the expected bootloader address form the primary address. Unfortunately chrome_laptop does probe the devices and if touchpad (or touchscreen, or both) comes up in bootloader mode the i2c device gets instantiated with the bootloader address which confuses the driver. To work around this issue let's probe the primary address first. If the device is not detected at the primary address we'll probe alternative addresses as "dummy" devices. If any of them are found, destroy the dummy client and instantiate client with proper name at primary address still. Signed-off-by: NDmitry Torokhov <dmitry.torokhov@gmail.com> Signed-off-by: NOlof Johansson <olof@lixom.net>
-
由 Hans de Goede 提交于
commit a39f46df ("toshiba_acpi: Fix regression caused by backlight extra check code") causes the backlight to no longer work on the Toshiba Z30, reverting that commit fixes this but restores the original issue fixed by that commit. Looking at the toshiba_acpi backlight code for a fix for this I noticed that the toshiba code is the only code under platform/x86 which unconditionally registers a vendor acpi backlight interface, without checking for acpi_video backlight support first. This commit adds the necessary checks bringing toshiba_acpi in line with the other drivers, and fixing the Z30 regression without needing to revert the commit causing it. Chances are that there will be some Toshiba models which have a non working acpi-video implementation while the toshiba vendor backlight interface does work, this commit adds an empty dmi_id table where such systems can be added, this is identical to how other drivers handle such systems. BugLink: https://bugzilla.redhat.com/show_bug.cgi?id=1206036 BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=86521Signed-off-by: NHans de Goede <hdegoede@redhat.com> Reviewed-and-tested-by: NAzael Avalos <coproscefalo@gmail.com> Signed-off-by: NDarren Hart <dvhart@linux.intel.com>
-
- 24 4月, 2015 13 次提交
-
-
由 Geert Uytterhoeven 提交于
If NO_DMA=y: drivers/built-in.o: In function `img_hash_write_via_dma_stop': img-hash.c:(.text+0xa2b822): undefined reference to `dma_unmap_sg' drivers/built-in.o: In function `img_hash_xmit_dma': img-hash.c:(.text+0xa2b8d8): undefined reference to `dma_map_sg' img-hash.c:(.text+0xa2b948): undefined reference to `dma_unmap_sg' Also move the "depends" section below the "tristate" line while we're at it. Signed-off-by: NGeert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>
-
由 Rafael J. Wysocki 提交于
If the special PRP0001 device ID is present in the given device's list of ACPI/PNP IDs and the device has a valid "compatible" property in the _DSD, it should be enumerated using the default mechanism, unless some scan handlers match the IDs preceding PRP0001 in the device's list of ACPI/PNP IDs. In addition to that, no scan handlers matching the IDs following PRP0001 in that list should be attached to the device. To make that happen, define a scan handler that will match PRP0001 and trigger the default enumeration for the matching devices if the "compatible" property is present for them. Since that requires the check for platform_id and device->handler to be removed from acpi_default_enumeration(), move the fallback invocation of acpi_default_enumeration() to acpi_bus_attach() (after it's checked if there's a matching ACPI driver for the device), which is a better place to call it, and do the platform_id check in there too (device->handler is guaranteed to be unset at the point where the function is looking for a matching ACPI driver). Signed-off-by: NRafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: NDarren Hart <dvhart@linux.intel.com>
-
由 Rafael J. Wysocki 提交于
acpi_scan_is_offline() may be called under the physical_node_lock lock of the given device object's parent, so prevent lockdep from complaining about that by annotating that instance with SINGLE_DEPTH_NESTING. Fixes: caa73ea1 (ACPI / hotplug / driver core: Handle containers in a special way) Reported-and-tested-by: NXie XiuQi <xiexiuqi@huawei.com> Reviewed-by: NToshi Kani <toshi.kani@hp.com> Cc: 3.14+ <stable@vger.kernel.org> # 3.14+ Signed-off-by: NRafael J. Wysocki <rafael.j.wysocki@intel.com>
-
由 Imre Deak 提交于
Due this typo we don't save/restore the GFX_MAX_REQ_COUNT register across suspend/resume, so fix this. This was introduced in commit ddeea5b0 Author: Imre Deak <imre.deak@intel.com> Date: Mon May 5 15:19:56 2014 +0300 drm/i915: vlv: add runtime PM support I noticed this only by reading the code. To my knowledge it shouldn't cause any real problems at the moment, since the power well backing this register remains on across a runtime s/r. This may change once system-wide s0ix functionality is enabled in the kernel. v2: - resend after a missing git add -u :/ Cc: stable@vger.kernel.org Signed-off-by: NImre Deak <imre.deak@intel.com> Tested-By: PRC QA PRTS (Patch Regression Test System Contact: shuang.he@intel.com) Signed-off-by: NRodrigo Vivi <rodrigo.vivi@intel.com> Reviewed-by: NMika Kuoppala <mika.kuoppala@intel.com> Signed-off-by: NJani Nikula <jani.nikula@intel.com>
-
由 Michel Thierry 提交于
WaIdleLiteRestore is an execlists-only workaround, and requires the driver to ensure that any context always has HEAD!=TAIL when attempting lite restore. Add two extra MI_NOOP instructions at the end of each request, but keep the requests tail pointing before the MI_NOOPs. We may not need to executed them, and this is why request->tail is sampled before adding these extra instructions. If we submit a context to the ELSP which has previously been submitted, move the tail pointer past the MI_NOOPs. This ensures HEAD!=TAIL. v2: Move overallocation to gen8_emit_request, and added note about sampling request->tail in commit message (Chris). v3: Remove redundant request->tail assignment in __i915_add_request, in lrc mode this is already set in execlists_context_queue. Do not add wa implementation details inside gem (Chris). v4: Apply the wa whenever the req has been resubmitted and update comment (Chris). Cc: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk> Signed-off-by: NThomas Daniel <thomas.daniel@intel.com> Signed-off-by: NMichel Thierry <michel.thierry@intel.com> Signed-off-by: NJani Nikula <jani.nikula@intel.com>
-
由 Dmitry Torokhov 提交于
The hardware, according to the specs, is limited to 256 byte transfers, and current driver has no protections in case users attempt to do larger transfers. The code will just stomp over status register and mayhem ensues. Let's split larger transfers into digestable chunks. Doing this allows Atmel MXT driver on Pixel 1 function properly (it hasn't since commit 9d8dc3e5 "Input: atmel_mxt_ts - implement T44 message handling" which tries to consume multiple touchscreen/touchpad reports in a single transaction). Cc: stable@vger.kernel.org Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk> Signed-off-by: NDmitry Torokhov <dmitry.torokhov@gmail.com> Signed-off-by: NJani Nikula <jani.nikula@intel.com>
-
由 Wolfram Sang 提交于
The driver uses pinctrl directly and thus should include the appropriate header. Sort the headers while we are here to have a better view what is included and what is not. Reported-by: NPascal Huerst <pascal.huerst@gmail.com> Signed-off-by: NWolfram Sang <wsa@the-dreams.de>
-
由 Wolfram Sang 提交于
Those symlinks are created for the mux_dev, so we need to remove it from there. Currently, it breaks for muxes where the mux_dev is not the device of the parent adapter like this: [ 78.234644] WARNING: CPU: 0 PID: 365 at fs/sysfs/dir.c:31 sysfs_warn_dup+0x5c/0x78() [ 78.242438] sysfs: cannot create duplicate filename '/devices/platform/i2cbus@8/channel-0' Remove confusing comments while we are here. Signed-off-by: NWolfram Sang <wsa+renesas@sang-engineering.com> Signed-off-by: NWolfram Sang <wsa@the-dreams.de> Fixes: c9449aff Cc: stable@kernel.org
-
由 Wolfram Sang 提交于
And sort them to prevent this from happening again. Reported-by: NWei Yongjun <yongjun_wei@trendmicro.com.cn> Signed-off-by: NWolfram Sang <wsa@the-dreams.de>
-
由 Charles Keepax 提交于
Commit 523c5b89 ("i2c: Remove support for legacy PM") removed the PM ops from the bus type, which causes the pm operations on the s3c2410 adapter device to fail (-ENOSUPP in rpm_callback). The adapter device doesn't get bound to a driver and as such can't have its own pm_runtime callbacks. Previously this was fine as the bus callbacks would have been used, but now this can cause devices which use PM runtime and are attached over I2C to fail to resume. This commit fixes this issue by marking all adapter devices with pm_runtime_no_callbacks, since they can't have any. Signed-off-by: NCharles Keepax <ckeepax@opensource.wolfsonmicro.com> Acked-by: NBeata Michalska <b.michalska@samsung.com> Signed-off-by: NWolfram Sang <wsa@the-dreams.de> Fixes: 523c5b89 Cc: stable@kernel.org
-
由 Wolfram Sang 提交于
My Pengutronix address is not valid anymore, redirect people to the Pengutronix kernel team. Reported-by: NHarald Geyer <harald@ccbib.org> Signed-off-by: NWolfram Sang <wsa@the-dreams.de> Acked-by: NRobert Schwebel <r.schwebel@pengutronix.de>
-
由 Wolfram Sang 提交于
My Pengutronix address is not valid anymore, redirect people to the Pengutronix kernel team. Reported-by: NHarald Geyer <harald@ccbib.org> Signed-off-by: NWolfram Sang <wsa@the-dreams.de> Acked-by: NRobert Schwebel <r.schwebel@pengutronix.de>
-
由 Dmitry Torokhov 提交于
master_xfer() method should return number of i2c messages transferred, but on Rockchip we were usually returning just 1, which caused trouble with users that actually check number of transferred messages vs. checking for negative error codes. Signed-off-by: NDmitry Torokhov <dmitry.torokhov@gmail.com> Signed-off-by: NWolfram Sang <wsa@the-dreams.de> Cc: stable@kernel.org
-
- 23 4月, 2015 2 次提交
-
-
由 Shobhit Kumar 提交于
For platforms that don't support DT, some early MFD modules can register lookup tables. Remove the __init annotation so that this works. This is similar to gpio_add_lookup_table() which allows late additions. CC: Samuel Ortiz <sameo@linux.intel.com> Cc: Linus Walleij <linus.walleij@linaro.org> Cc: Alexandre Courbot <gnurou@gmail.com> Cc: Thierry Reding <thierry.reding@gmail.com> Signed-off-by: NShobhit Kumar <shobhit.kumar@intel.com> Signed-off-by: NThierry Reding <thierry.reding@gmail.com>
-
由 Vinod Koul 提交于
HSU_DMA is selected by the HSU_DMA_PCI driver, this should be user selected so remove the user prompt for this Signed-off-by: NVinod Koul <vinod.koul@intel.com> Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>
-
- 22 4月, 2015 12 次提交
-
-
由 Ilya Dryomov 提交于
After the switch to blk-mq rbd_wq processes requests, not devices. Signed-off-by: NIlya Dryomov <idryomov@gmail.com>
-
由 Wolfram Sang 提交于
My Pengutronix address is not valid anymore, redirect people to the Pengutronix kernel team. Reported-by: NHarald Geyer <harald@ccbib.org> Signed-off-by: NWolfram Sang <wsa@the-dreams.de> Acked-by: NRobert Schwebel <r.schwebel@pengutronix.de> Signed-off-by: NWim Van Sebroeck <wim@iguana.be>
-
由 Wolfram Sang 提交于
My Pengutronix address is not valid anymore, redirect people to the Pengutronix kernel team. Reported-by: NHarald Geyer <harald@ccbib.org> Signed-off-by: NWolfram Sang <wsa@the-dreams.de> Acked-by: NRobert Schwebel <r.schwebel@pengutronix.de> Signed-off-by: NWim Van Sebroeck <wim@iguana.be>
-
由 Aaro Koskinen 提交于
Use fixed length string for register names. This saves 416 bytes in text size. Signed-off-by: NAaro Koskinen <aaro.koskinen@iki.fi> Reviewed-by: NGuenter Roeck <linux@roeck-us.net> Signed-off-by: NWim Van Sebroeck <wim@iguana.be>
-
由 Aaro Koskinen 提交于
Fix some trivial coding style issues to reduce noise from static analyzers. Signed-off-by: NAaro Koskinen <aaro.koskinen@iki.fi> Reviewed-by: NGuenter Roeck <linux@roeck-us.net> Signed-off-by: NWim Van Sebroeck <wim@iguana.be>
-
由 Aaro Koskinen 提交于
Convert OCTEON watchdog to WATCHDOG_CORE API. This enables support for multiple watchdogs on OCTEON boards. Signed-off-by: NAaro Koskinen <aaro.koskinen@iki.fi> Reviewed-by: NGuenter Roeck <linux@roeck-us.net> Signed-off-by: NWim Van Sebroeck <wim@iguana.be>
-
由 Michal Simek 提交于
Remove Kconfig dependency and enable driver for all ARCHs. Signed-off-by: NMichal Simek <michal.simek@xilinx.com> Reviewed-by: NGuenter Roeck <linux@roeck-us.net> Signed-off-by: NWim Van Sebroeck <wim@iguana.be>
-
由 Mathieu Olivari 提交于
MSM watchdog configuration happens in the same register block as the timer, so we'll use the same binding as the existing timer. The qcom-wdt will now be probed when devicetree has an entry compatible with "qcom,kpss-timer" or "qcom-scss-timer". Signed-off-by: NMathieu Olivari <mathieu@codeaurora.org> Reviewed-by: NStephen Boyd <sboyd@codeaurora.org> Acked-by: NGuenter Roeck <linux@roeck-us.net> Signed-off-by: NWim Van Sebroeck <wim@iguana.be>
-
由 Joe Perches 提交于
The seq_printf return value, because it's frequently misused, will eventually be converted to void. See: commit 1f33c41c ("seq_file: Rename seq_overflow() to seq_has_overflowed() and make public") Signed-off-by: NJoe Perches <joe@perches.com> Acked-by: Guenter Roeck <linux~roeck-us.net> Signed-off-by: NWim Van Sebroeck <wim@iguana.be>
-
由 Vinod Koul 提交于
DW_DMAC_CORE is slected by PCI or Platform driver, so this symbol shouldn't be user selectable, so remove the prompt Signed-off-by: NVinod Koul <vinod.koul@intel.com>
-
由 Chris Bainbridge 提交于
Use list_for_each_entry_safe for iterating because handler may be freed in the loop. BUG: unable to handle kernel NULL pointer dereference at 000000000000002c IP: [<ffffffff814d69c8>] acpi_ec_put_query_handler+0x7/0x1a Call Trace: acpi_ec_remove_query_handler+0x87/0x97 acpi_smbus_hc_remove+0x2a/0x44 [sbshc] acpi_device_remove+0x7b/0x9a __device_release_driver+0x7e/0x110 driver_detach+0xb0/0xc0 bus_remove_driver+0x54/0xe0 driver_unregister+0x2b/0x60 acpi_bus_unregister_driver+0x10/0x12 acpi_smb_hc_driver_exit+0x10/0x12 [sbshc] SyS_delete_module+0x1b8/0x210 system_call_fastpath+0x12/0x6a Signed-off-by: NChris Bainbridge <chris.bainbridge@gmail.com> Signed-off-by: NRafael J. Wysocki <rafael.j.wysocki@intel.com>
-
由 Eric Mei 提交于
When array is degraded, read data landed on failed drives will result in reading rest of data in a stripe. So a single sequential read would result in same data being read twice. This patch is to avoid chunk aligned read for degraded array. The downside is to involve stripe cache which means associated CPU overhead and extra memory copy. Test Results: Following test are done on a enterprise storage node with Seagate 6T SAS drives and Xeon E5-2648L CPU (10 cores, 1.9Ghz), 10 disks MD RAID6 8+2, chunk size 128 KiB. I use FIO, using direct-io with various bs size, enough queue depth, tested sequential and 100% random read against 3 array config: 1) optimal, as baseline; 2) degraded; 3) degraded with this patch. Kernel version is 4.0-rc3. Each individual test I only did once so there might be some variations, but we just focus on big trend. Sequential Read: bs=(KiB) optimal(MiB/s) degraded(MiB/s) degraded-with-patch (MiB/s) 1024 1608 656 995 512 1624 710 956 256 1635 728 980 128 1636 771 983 64 1612 1119 1000 32 1580 1420 1004 16 1368 688 986 8 768 647 953 4 411 413 850 Random Read: bs=(KiB) optimal(IOPS) degraded(IOPS) degraded-with-patch (IOPS) 1024 163 160 156 512 274 273 272 256 426 428 424 128 576 592 591 64 726 724 726 32 849 848 837 16 900 970 971 8 927 940 929 4 948 940 955 Some notes: * In sequential + optimal, as bs size getting smaller, the FIO thread become CPU bound. * In sequential + degraded, there's big increase when bs is 64K and 32K, I don't have explanation. * In sequential + degraded-with-patch, the MD thread mostly become CPU bound. If you want to we can discuss specific data point in those data. But in general it seems with this patch, we have more predictable and in most cases significant better sequential read performance when array is degraded, and almost no noticeable impact on random read. Performance is a complicated thing, the patch works well for this particular configuration, but may not be universal. For example I imagine testing on all SSD array may have very different result. But I personally think in most cases IO bandwidth is more scarce resource than CPU. Signed-off-by: NEric Mei <eric.mei@seagate.com> Signed-off-by: NNeilBrown <neilb@suse.de>
-