From 0d6526f82c3cdefcca47f73f5fc08dc6f335eac6 Mon Sep 17 00:00:00 2001 From: Tim Chen Date: Mon, 21 Sep 2026 17:37:22 -0700 Subject: sched/cache: Keep nr_pref_llc_running in the runnable domain, to fix LLC mis-scheduling bug alb_break_llc() decides whether to break LLC preference during active load balance. It does so by testing that every runnable fair task on the source rq prefers its LLC: env->src_rq->nr_pref_llc_running == env->src_rq->cfs.h_nr_runnable But the two counters cover different sets. nr_pref_llc_running is updated in account_llc_enqueue()/account_llc_dequeue(), next to cfs_rq->nr_queued, so it follows queued tasks. h_nr_runnable is updated in set_delayed()/ clear_delayed() and drops delay-dequeued tasks. So under DELAY_DEQUEUE, a preferring task that goes to sleep stays counted in nr_pref_llc_running while h_nr_runnable falls. The equality then breaks, alb_break_llc() returns false, and active balance is free to pull a task off its preferred LLC. Active balance only moves runnable tasks, and this is the only LLC check it consults: once the stopper runs, LBF_ACTIVE_LB skips the per-task test in can_migrate_task(). The runnable set is the one we want. Fix it on the counter side. A task should be counted in nr_pref_llc_running exactly while it is both queued on its preferred LLC (pref_llc_queued) and runnable (!sched_delayed). Define that membership once in task_pref_llc_runnable(), and adjust the counter only through pref_llc_running_inc()/pref_llc_running_dec() from the four sites that change either input: account_llc_enqueue(), account_llc_dequeue(), set_delayed() and clear_delayed(). Gating every update on the same predicate keeps the delay, wake and dequeue paths from double-counting or underflowing; see the comments at those sites for the ordering. nr_llc_running and sd->llc_counts are not touched and stay on queued semantics. Fixes: 714059f79ff0 ("sched/cache: Handle moving single tasks to/from their preferred LLC") Closes: https://lore.kernel.org/lkml/20260827135000.735138-1-zhanxusheng@xiaomi.com/ Reported-by: Zhan Xusheng Suggested-by: Chen Yu Signed-off-by: Tim Chen Signed-off-by: Peter Zijlstra (Intel) Signed-off-by: Ingo Molnar Reviewed-by: Kayra Cizmeci Cc: # v7.2.x Link: https://patch.msgid.link/06af61afedac32e6477f57feb4d658f6c411c3af.1790035273.git.tim.c.chen@linux.intel.com --- drivers/firmware/sysfb_simplefb.c | 140 ++++++++++++++++++++++++++++++++++++++ 1 file changed, 140 insertions(+) create mode 100644 drivers/firmware/sysfb_simplefb.c (limited to 'drivers/firmware/sysfb_simplefb.c') diff --git a/drivers/firmware/sysfb_simplefb.c b/drivers/firmware/sysfb_simplefb.c new file mode 100644 index 000000000..71f542e37 --- /dev/null +++ b/drivers/firmware/sysfb_simplefb.c @@ -0,0 +1,140 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * Generic System Framebuffers + * Copyright (c) 2012-2013 David Herrmann + */ + +/* + * simple-framebuffer probing + * Try to convert "screen_info" into a "simple-framebuffer" compatible mode. + * If the mode is incompatible, we return "false" and let the caller create + * legacy nodes instead. + */ + +#include +#include +#include +#include +#include +#include +#include +#include + +static const char simplefb_resname[] = "BOOTFB"; +static const struct simplefb_format formats[] = SIMPLEFB_FORMATS; + +/* try parsing screen_info into a simple-framebuffer mode struct */ +__init bool sysfb_parse_mode(const struct screen_info *si, + struct simplefb_platform_data *mode) +{ + __u8 type; + u32 bits_per_pixel; + unsigned int i; + + type = si->orig_video_isVGA; + if (type != VIDEO_TYPE_VLFB && type != VIDEO_TYPE_EFI) + return false; + + bits_per_pixel = __screen_info_lfb_bits_per_pixel(si); + + for (i = 0; i < ARRAY_SIZE(formats); ++i) { + const struct simplefb_format *f = &formats[i]; + + if (f->transp.length) + continue; /* transparent formats are unsupported by VESA/EFI */ + + if (bits_per_pixel == f->bits_per_pixel && + si->red_size == f->red.length && + si->red_pos == f->red.offset && + si->green_size == f->green.length && + si->green_pos == f->green.offset && + si->blue_size == f->blue.length && + si->blue_pos == f->blue.offset) { + mode->format = f->name; + mode->width = si->lfb_width; + mode->height = si->lfb_height; + mode->stride = si->lfb_linelength; + return true; + } + } + + return false; +} + +__init struct platform_device *sysfb_create_simplefb(const struct screen_info *si, + const struct simplefb_platform_data *mode, + struct device *parent) +{ + struct platform_device *pd; + struct resource res; + u64 base, size; + u32 length; + int ret; + + /* + * If the 64BIT_BASE capability is set, ext_lfb_base will contain the + * upper half of the base address. Assemble the address, then make sure + * it is valid and we can actually access it. + */ + base = si->lfb_base; + if (si->capabilities & VIDEO_CAPABILITY_64BIT_BASE) + base |= (u64)si->ext_lfb_base << 32; + if (!base || (u64)(resource_size_t)base != base) { + printk(KERN_DEBUG "sysfb: inaccessible VRAM base\n"); + return ERR_PTR(-EINVAL); + } + + /* + * Don't use lfb_size as IORESOURCE size, since it may contain the + * entire VMEM, and thus require huge mappings. Use just the part we + * need, that is, the part where the framebuffer is located. But verify + * that it does not exceed the advertised VMEM. + * Note that in case of VBE, the lfb_size is shifted by 16 bits for + * historical reasons. + */ + size = si->lfb_size; + if (si->orig_video_isVGA == VIDEO_TYPE_VLFB) + size <<= 16; + length = mode->height * mode->stride; + if (length > size) { + printk(KERN_WARNING "sysfb: VRAM smaller than advertised\n"); + return ERR_PTR(-EINVAL); + } + length = PAGE_ALIGN(length); + + /* setup IORESOURCE_MEM as framebuffer memory */ + memset(&res, 0, sizeof(res)); + res.flags = IORESOURCE_MEM; + res.name = simplefb_resname; + res.start = base; + res.end = res.start + length - 1; + if (res.end <= res.start) + return ERR_PTR(-EINVAL); + + pd = platform_device_alloc("simple-framebuffer", 0); + if (!pd) + return ERR_PTR(-ENOMEM); + + pd->dev.parent = parent; + + sysfb_set_efifb_fwnode(si, pd); + + ret = platform_device_add_resources(pd, &res, 1); + if (ret) + goto err_put_device; + + ret = platform_device_add_data(pd, mode, sizeof(*mode)); + if (ret) + goto err_put_device; + + ret = platform_device_add(pd); + if (ret) + goto err_put_device; + + return pd; + +err_put_device: + platform_device_put(pd); + + return ERR_PTR(ret); +} -- cgit v1.3.1