Scrubbing pages on initial balloon down can take some time, especially in nested virtualization case (nested EPT is slow). When HVM/PVH guest is started with memory= significantly lower than maxmem=, all the extra pages will be scrubbed before returning to Xen. But since most of them weren't used at all at that point, Xen needs to populate them first (from populate-on-demand pool). In nested virt case (Xen inside KVM) this slows down the guest boot by 15-30s with just 1.5GB needed to be returned to Xen. Add runtime parameter to enable/disable it, to allow initially disabling scrubbing, then enable it back during boot (for example in initramfs). Such usage relies on assumption that a) most pages ballooned out during initial boot weren't used at all, and b) even if they were, very few secrets are in the guest at that time (before any serious userspace kicks in). Convert CONFIG_XEN_SCRUB_PAGES to CONFIG_XEN_SCRUB_PAGES_DEFAULT (also enabled by default), controlling default value for the new runtime switch.
Unified · Split
| @@ -75,3 +75,12 @@ Contact: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> · 74 unmodified lines | ||
| 75 | 75 | Description: |
| 76 | 76 | Amount (in KiB) of low (or normal) memory in the |
| 77 | 77 | balloon. |
| 78 | + | |
| 79 | +What: /sys/devices/system/xen_memory/xen_memory0/scrub_pages | |
| 80 | +Date: September 2018 | |
| 81 | +KernelVersion: 4.20 | |
| 82 | +Contact: xen-devel@lists.xenproject.org | |
| 83 | +Description: | |
| 84 | + Control scrubbing pages before returning them to Xen for others domains | |
| 85 | + use. Can be set with xen_scrub_pages cmdline | |
| 86 | + parameter. Default value controlled with CONFIG_XEN_SCRUB_PAGES_DEFAULT. | |
| @@ -5000,6 +5000,12 @@ · 4999 unmodified lines | ||
| 5000 | 5000 | Disables the PV optimizations forcing the HVM guest to |
| 5001 | 5001 | run as generic HVM guest with no PV drivers. |
| 5002 | 5002 | |
| 5003 | + xen_scrub_pages= [XEN] | |
| 5004 | + Boolean option to control scrubbing pages before giving them back | |
| 5005 | + to Xen, for use by other domains. Can be also changed at runtime | |
| 5006 | + with /sys/devices/system/xen_memory/xen_memory0/scrub_pages. | |
| 5007 | + Default value controlled with CONFIG_XEN_SCRUB_PAGES_DEFAULT. | |
| 5008 | + | |
| 5003 | 5009 | xirc2ps_cs= [NET,PCMCIA] |
| 5004 | 5010 | Format: |
| 5005 | 5011 | <irq>,<irq_mask>,<io>,<full_duplex>,<do_sound>,<lockup_hack>[,<irq2>[,<irq3>[,<irq4>]]] |
| @@ -79,15 +79,19 @@ config XEN_BALLOON_MEMORY_HOTPLUG_LIMIT · 78 unmodified lines | ||
| 79 | 79 | This value is used to allocate enough space in internal |
| 80 | 80 | tables needed for physical memory administration. |
| 81 | 81 | |
| 82 | −config XEN_SCRUB_PAGES | |
| 83 | − bool "Scrub pages before returning them to system" | |
| 82 | +config XEN_SCRUB_PAGES_DEFAULT | |
| 83 | + bool "Scrub pages before returning them to system by default" | |
| 84 | 84 | depends on XEN_BALLOON |
| 85 | 85 | default y |
| 86 | 86 | help |
| 87 | 87 | Scrub pages before returning them to the system for reuse by |
| 88 | 88 | other domains. This makes sure that any confidential data |
| 89 | 89 | is not accidentally visible to other domains. Is it more |
| 90 | − secure, but slightly less efficient. | |
| 90 | + secure, but slightly less efficient. This can be controlled with | |
| 91 | + xen_scrub_pages=0 parameter and | |
| 92 | + /sys/devices/system/xen_memory/xen_memory0/scrub_pages. | |
| 93 | + This option only sets the default value. | |
| 94 | + | |
| 91 | 95 | If in doubt, say yes. |
| 92 | 96 | |
| 93 | 97 | config XEN_DEV_EVTCHN |
| @@ -14,6 +14,10 @@ · 13 unmodified lines | ||
| 14 | 14 | |
| 15 | 15 | #include <xen/interface/memory.h> |
| 16 | 16 | #include <xen/mem-reservation.h> |
| 17 | +#include <linux/moduleparam.h> | |
| 18 | + | |
| 19 | +bool __read_mostly xen_scrub_pages = IS_ENABLED(CONFIG_XEN_SCRUB_PAGES_DEFAULT); | |
| 20 | +core_param(xen_scrub_pages, xen_scrub_pages, bool, 0); | |
| 17 | 21 | |
| 18 | 22 | /* |
| 19 | 23 | * Use one extent per PAGE_SIZE to avoid to break down the page into |
| @@ -44,6 +44,7 @@ · 43 unmodified lines | ||
| 44 | 44 | #include <xen/xenbus.h> |
| 45 | 45 | #include <xen/features.h> |
| 46 | 46 | #include <xen/page.h> |
| 47 | +#include <xen/mem-reservation.h> | |
| 47 | 48 | |
| 48 | 49 | #define PAGES2KB(_p) ((_p)<<(PAGE_SHIFT-10)) |
| 49 | 50 | |
| @@ -137,6 +138,7 @@ static DEVICE_ULONG_ATTR(schedule_delay, 0444, balloon_stats.schedule_delay); · 87 unmodified lines | ||
| 137 | 138 | static DEVICE_ULONG_ATTR(max_schedule_delay, 0644, balloon_stats.max_schedule_delay); |
| 138 | 139 | static DEVICE_ULONG_ATTR(retry_count, 0444, balloon_stats.retry_count); |
| 139 | 140 | static DEVICE_ULONG_ATTR(max_retry_count, 0644, balloon_stats.max_retry_count); |
| 141 | +static DEVICE_BOOL_ATTR(scrub_pages, 0644, xen_scrub_pages); | |
| 140 | 142 | |
| 141 | 143 | static ssize_t show_target_kb(struct device *dev, struct device_attribute *attr, |
| 142 | 144 | char *buf) |
| @@ -203,6 +205,7 @@ static struct attribute *balloon_attrs[] = { · 60 unmodified lines | ||
| 203 | 205 | &dev_attr_max_schedule_delay.attr.attr, |
| 204 | 206 | &dev_attr_retry_count.attr.attr, |
| 205 | 207 | &dev_attr_max_retry_count.attr.attr, |
| 208 | + &dev_attr_scrub_pages.attr.attr, | |
| 206 | 209 | NULL |
| 207 | 210 | }; |
| 208 | 211 | |
| @@ -17,11 +17,12 @@ · 16 unmodified lines | ||
| 17 | 17 | |
| 18 | 18 | #include <xen/page.h> |
| 19 | 19 | |
| 20 | +extern bool xen_scrub_pages; | |
| 21 | + | |
| 20 | 22 | static inline void xenmem_reservation_scrub_page(struct page *page) |
| 21 | 23 | { |
| 22 | −#ifdef CONFIG_XEN_SCRUB_PAGES | |
| 23 | − clear_highpage(page); | |
| 24 | −#endif | |
| 24 | + if (xen_scrub_pages) | |
| 25 | + clear_highpage(page); | |
| 25 | 26 | } |
| 26 | 27 | |
| 27 | 28 | #ifdef CONFIG_XEN_HAVE_PVMMU |